Battery Life
December 2, 2024
I'm satisfied with the battery life
of most of my Apple gadgets. Except for
iPhone SE (2nd Generation), performance
per watt is really impressive. On iPhone SE,
low power mode is required to grind through
the day.
It builds habits. I ended up using
iPhone SE only for critical tasks. There
is no room for too much distraction.
Perhaps it's a good thing, but I hope
the next iPhone SE can provide better
battery life.
The lowest percentage I ever got
is 4% battery, not too bad. But occasionally
I have to charge iPhone SE during the day.
As the reader might guess, I don't travel
much, so it's fine.
M2 MacBook Air works surprisingly well.
On average I only need to charge it once
per week if used lightly. These days most
tasks are done on the iPad. MacBook Air
provides the most satisfactory experience
for me.
Efficient wireless charging is very
hard. Although it looks like the future,
we probably need significant innovations
to get there, while my power cable
connection works most of the time.
If there is a mode to dramatically
boost battery life on iPhone SE, even
with drastically limited idle functionalities,
I'd appreciate it very much. It will
allow a very thin profile.
Battery life is probably a decisive
factor for mobile phone purchase, but
vendors don't talk about it seriously.
I certainly didn't expect iPhone SE to
shape my habits. It turned out well,
but customers should be informed.
Social Media Influences
November 30, 2024
Many people believe in computable data.
What they observe is computational retrieval
of data. What they believe in is computational
generation of data. A piece of data can be
generated by probability processes and people
still think it's computational.
Social media made people stupid and
foolish, not because X/Twitter was powerful,
but because people were intellectually lazy.
So the case described in the previous paragraph
made it to the high order of science.
Yeah, ink-and-tape Turing machines are alive.
I have little interest in making people
see the truth, because people won't, but here
is a observation. Most people live in The Matrix,
their lives and power depend on it. The most
I can hope for is peaceful coexistence, rather
than a land of truth.
Don't think it's as simple as switching
from X/Twitter to Bluesky. Politicians are
the same. Economists are the same. Opinion
leaders are the same. It will get better only
after people become intellectually scrupulous.
On that, much effort is required.
Propagandists don't understand the war
on truth is waged in people's minds, rather
than a technological anomaly. For centuries,
the same thing happens. It's just it happens
through digital technologies today. Of course,
what else can it be, dial-up telephone?
The purported solution proposed by
propagandists to break up leading technological
companies isn't going to solve the problem.
We will discover we get a much weaker and
much worse Silicon Valley with the same
dysfunctional politicians.
The tradition of the shocked is to
go berserk. That's what propagandists are.
Rather than learning how to make better use
of technology, they trash it. But, we believe
trashing digital technologies critical to
people's lives is stupid and foolish.
You have been warned.
Tip of a Iceberg
November 26, 2024
As the blog grows, it's good time to inform
the reader the style of its posts. Unlike many
dedicated bloggers, the posts I wrote are quite
short. The purpose is to facilitate efficient
exchange of materials, rather than a comprehensive
exposition. Here I illustrate how to read the
tip of a iceberg.
I outlined HTML5 postcards
in a paragraph. Here we expand it a little, just
to show how much content it covers.
HTML5 can fit multiple screens. With well-designed
templates, postcards can look great on iPhone,
iPad, and Macs. As content can be dynamic, protections
like sandboxing are recommended for privacy
and security.
Down to elements. We may use Clips, iMovie,
or Final Cut Pro to prepare a video. Use Apple
Music to add soundtracks. Use Pixelmator to
draw graphic with personal style. Use Camera
to add photos. AI can help with text and
images, both content and style.
There can be stores and apps that sell
templates for convenient use. Or the user may
craft dedicated templates in professional apps.
Assembling elements and templates together,
we may preview postcards in multi-screen
simulator.
The process can go on and on until one
gets a perfect postcard. Then we may send
it via Messages. If you like it, you may
easily store the postcard in Photos for
future reference. HTML5 dynamic content
can be very engaging.
Postcards are enabled by our unified
scheme for mobile, cloud, and AI. Here we
won't go deep into the technology, but
emphasize that the choice of HTML5 is for
cross platform compatibility, so that even
Android users can view postcards in browsers.
Alright, the previous paragraphs are
a little expansion of our postcard idea.
As the reader might suspect, much content
in this blog is of such nature. It's possible
and rewarding for a imaginative person
to see through the implications of our
posts. Posts are very short for efficient
exchange, but the possibilities are endless.
That's the style here.
No Need for Nullstellensatz
November 24, 2024
For AG beginners, the way scheme theory
does away with Nullstellensatz may seem magical.
Here we explain the workings behind based on
our previous post
for modern AG motivation.
In classical AG, for a algebraically
closed field k, we have AG dictionary that
morphisms between affine varieties correspond
to k-algebra homomorphisms between coordinate
algebras.
The map between varieties is determined
point by point by pulling back maximal ideals
via k-algebra homomorphism. Here Nullstellensatz
is used to guarantee maximal ideals correspond
to points.
For schemes, we replace maximal ideals with
prime ideals. Since the pullback of a prime
ideal is always a prime ideal, there is no
need for Nullstellensatz to guarantee the
correspondence.
From categorical point of view, cohomology
theory can be defined without reference to
base field k, preserving the AG dictionary
is all we need. This allows us to work with
integers as well.
Google AI Disadvantage
November 22, 2024
Microsoft Ignite 2024 is really worth
watching. With Microsoft's focus on corporate
world, and Apple's focus on personal computing,
Google's a-bit-here-a-bit-there approach risks
losing its AI edge. Microsoft Office together
with Copilot is heavy weight, is Google prepared?
While there are debatable subtleties, Excel
with Copilot is a wonder. Experts might argue
whether data can reasonably gauge truth, but
using Excel and Copilot to perform data analysis
can be very helpful.
Better, Microsoft has advanced developer tools
to integrate AI into all sorts of apps. Google
lacks powerful PCs, which might cost it the game.
As illustrated in our unified scheme for
mobile, cloud, and AI, AI without some device
paradigm is very much less useful. Google can
be a AI company, but can it survive integrated
attack from Microsoft and maintain leadership?
Of course, it's possible to integrate
corporate and personal computing to find a way
out. But it requires efforts beyond
a-bit-here-a-bit-there approach. My feeling is
that Google doesn't take both worlds seriously
enough. Even with Apple's AI backwardness,
Google doesn't show significant advantage.
It's time to devise a workable strategy, Google!
Taiwan's Opportunity
November 19, 2024
DOJ is trying to force Chrome sale. It's
a huge opportunity for Taiwan. While many
companies have no interest in Chrome, Taiwan
can greatly benefit from global software
penetration and influence. There are few
apps with such global reach that Taiwan
can capitalize upon. It would be a great
addition to TSMC from visibility point of
view. Everybody knows TikTok is from PRC.
Now they would know Chrome is from Taiwan.
As USA politicians repeatedly ignored
our warning in previous posts and
book in progress,
it's really possible that their misinformed
decisions can ruin the Silicon Valley.
Our recommendation is therefore Taiwan
ownership of these pieces so that they
can work together under the hood. It's
not brilliant as a whole, but Taiwan
can use it.
Photo Albums
November 17, 2024
When Apple introduced iPhoto, Steve Jobs
amazed the crowd with the ease to create and
order a sophisticated print photo album. As
more and more people switch to sharing photos
online via social networks, Apple's edge
is quickly eroded.
But just as our unified scheme for mobile,
cloud, and AI may support digital
postcards
via Messages, it can support digital photo
albums as well. The bottleneck is that
modern EPUB format is very low quality.
While there is no obstacle for a open
format for books with rich graphics and
dynamic content, nobody is developing for it.
HTML5 within Messages might be a short-term
compromise for compelling digital photo
albums. Of course, privacy and security need
to be taken care of via sandboxing, etc.
The point is that we don't need to accept
Facebook aesthetics, but can create postcards
and photo albums according to personal taste.
Along with signal processing for music production
and medical devices, photo albums are another
example of the versatility of our scheme.
We will talk more under favorable conditions.
Public Contacts
November 15, 2024
A nice feature of the iPhone is that
callers are identified if they happen to be
in Contacts. Often, it means peace of mind
knowing who is calling. However, it's impractical
to build a comprehensive list of contacts
personally. We need a cloud service for
verified public contacts.
The Phone app may display who is calling
based on public contacts. Users may easily
add public contacts to their own Contacts app.
As the source updates information, the user's
Contacts app automatically updates via cloud sync.
Clearly, it looks similar to Maps. Currently,
we add locations to library, rather than easily
access them in Contacts. Public contacts may
provide a link between Maps and Contacts, so when
the government calls you back, you know where to go.
The possibility to search through public
contacts can save a lot of time. We may devise
digital business cards for exchange. It also makes
finding a graphic designer nearby much easier,
as many people don't register on Maps.
The problem is, of course, verification.
Powerful checks need to be put in place to
prevent fraud. Not to mention privacy and
security. But, if done right, it may inform
us whose call to pick up, and who is a
spammer. When government calls you, you
have a choice what to do.
Subtlety
November 12, 2024
There are many accusations around Apple
failing to be innovative. Blogs like Daring
Fireball countered with product launches like
Apple Watch and Vision Pro. Here we wish to
discuss another point, that the critics lack
the taste to appreciate subtlety.
Take a look at Camera Control. Critics
dismissed it as a disappointment. While people's
fingers are quite different, I can inform that
Camera Control worked perfectly on my first try.
It solves problems with the double role of volume button as
shutter button that feels like a hack.
It's just people have different fingers.
Let's turn our attention to software distribution.
Apps used to be sold in boxes via retail stores.
In such a way, iPhone would be severely limited.
But, of course, Apple invented App Store. Is it
too difficult for critics to recognize the significance
of such innovation?
Subtlety, especially incremental updates, can
make or break a product. It can not all be
revolutionary product launches, but must make
a lot of room for improvements so that products
can stay alive. If Apple launched first generations
after first generations, other competitors would
drive out Apple with much better follow-ups.
We need to appreciate subtlety.
These days, Macs are comparable with workstations.
The improvements under the hood are huge, but
critics only see fashionable GUI, and complain why
Apple didn't preserve Aqua. Well, I still
remember OS X Tiger freezing in front of me,
which is much more unlikely on modern macOS.
Similarly, people appreciate drop-resistant
iPhone design, including Ceramic Shield. Can
critics reasonably argue it's nothing? Hardware
is extremely hard. I bet critics have no idea
of the physics of Ceramic Shield.
Even little changes from isolated song
purchases to library subscriptions reflect Apple
innovating to survive the market. These are
subtleties that allow users to easily enjoy
gigantic collections of
music, without extensive trial and error.
It's easy to spell AI and Private Cloud
Compute, but implementation is much harder.
Critics may not be satisfied with the amazing
speed Apple delivered these products, but
they are foundations upon which future device
design depends on. Don't be surprised if
critics fail to predict Apple revolutionizing
music production.
Computers and Robots
November 10, 2024
Some economists are skeptical of computers
but bullish on robots. They don't understand
good robots build upon good computer control.
You can't build good robots by ditching good
computers.
Economists are quite serious about their
own views that often diverge exponentially from
reality. They mistook a little wisdom for grand
understanding like physics.
Even if computers are not as likable as
robots to economists, isn't it people's freedom
to build computers without the fear of being
unproductive?
Suppressing computers in order to boost
robots is not going to work. Suppressing AI
automation in order to boost robots is not
going to work, either.
If economists don't understand this,
they shouldn't be taken too seriously. The
Biden administration followed economists'
doctrine to defeat. Please learn something.
Apple Intelligence Waitlist
November 7, 2024
The reason I haven't tried out Apple Intelligence
is not the waitlist, but the consideration that
early Apple software tends to be buggy and AI can
do a lot of things to the computer. The waitlist
description said it's in beta. I have no spare
computer to act as beta tester.
In early days of iCloud, data loss was common.
I really fear Apple Intelligence may do the same.
But yeah, as everybody is still pondering the
consequences of USA Presidential Election, let's not
talk too much. After all, it's much more important
to find the right way forward.
Motivation for Modern AG
November 5, 2024
Classical algebraic geometry builds on
k-algebra homomorphisms and the AG dictionary.
From categorical point of view, it's convenient
to consider a wider class of algebras, that of
Z-algebra. To preserve the AG dictionary, we
take prime ideals rather than maximal ideals.
Then we get affine schemes. As classical objects
are obtained by gluing affines, we also glue
affine schemes with sheaves. The result is
the category of schemes. Notice that Z-algebras
are naturally identified with commutative
rings, so the usual literature hides the
algebra structure, which might be a little
bit strange for beginners.
The Problem with Google Search
November 3, 2024
Today I stumbled upon a great collection of
Apple keynotes
given by Steve Jobs via Google Search.
It's great that Google Search knows the web
site, but some reflection suggests trouble
for Google. The web site's collection is vastly
more expansive than what I got from searching
YouTube.
That is, Google's search department gave
terrible list of results based on keywords.
The fact that people can easily visit the
web site, but can not get the list of videos
directly within Google/YouTube Search means
Google's ranking algorithm is very shaky.
We don't know what makes Google so. But,
it's reasonable to expect the same thing
happens to other search keywords. I can imagine
a much better ranking algorithm may drive
Google Search out of business. Please, Google,
I should be able to get the list of all
Steve Jobs keynotes within search.
Notifications
November 2, 2024
Notifications were introduced as a means to
deliver updates without enabling background
processes. Some notifications are very useful,
like those used by AirTag, but I must say that
many modern uses of notifications are extremely
spammy. Facebook constantly reminds you there
is a new post. YouTube suggests (ir)relevant
videos all the time. Some even use notifications
for advertisements. While free speech is precious,
the mechanism to raise engagement is really toxic.
MacBook Pro Workflow
October 31, 2024
There was a time when everyday tasks required
MacBook Pro. It's primarily due to many opening
apps and windows. Nowadays, MacBook Air is so
good that even pro-level works can be done. So,
MacBook Pro really is for pros.
Graphics, app development, and academia
are the areas MacBook Pro workflow currently
thrives most, but many pro uses are prohibited
due to lack of software, or even simply a printer
driver.
As pro works are often dictated by workflow,
one can not choose MacBook Pro exclusively for performance,
but must consider compatibility issues. Autodesk
works better under Windows, so it's over.
Can Apple convince Autodesk to provide parity?
The question is mostly beyond us, but ecosystem
certainly is hard to build. Perhaps it should not
be Apple's focus, but given we can view 3D
models together in Vision Pro, it's a pity not
to have Autodesk onboard.
As people can see, in the video introducing
M4 MacBook Pro models, the workflow offered by
Apple Intelligence is for researchers to enhance
abstract and create illustrations with AI. These
are not central to pro research work. There is
much to do.
In areas where Apple does relatively well, like Final
Cut Pro and Logic Pro, the workflow experience
is excellent, and MacBook Pro is of great
value. Perhaps it's just me, but I still
hope one day MacBook Pro can find broader
audience.
Mac mini and Labs
October 30, 2024
As Apple released new Mac mini, social media
focused on the design of its power button. We wish
to offer a broader view. The Mac lineup is
tilted in favor of Mac mini if Mac Studio doesn't
receive stronger updates, as Mac mini is quite
configurable and offers similar I/O.
The modular design of Mac mini makes it suitable
for integration with other devices, like in labs.
With smaller dimensions, Mac mini can easily fit
into lab instruments. The problem is, of course,
why would labs choose Mac mini rather than PC?
Although Mac mini has superior hardware, its
development environment lacks something fundamental
like integrated developer cloud. As many data
operations are better performed on the cloud,
labs might choose Microsoft for productivity.
Libraries like AI tools also matter. Researchers
often choose Wolfram Language over Swift so that
they don't need to reinvent the wheel. If Apple
really wants to break into technical computing,
its libraries must greatly expand.
If the developer package can be delivered,
Mac mini can play a significant role in places like
labs. Considering its performance and price, it
will be the most cost-effective solution in the
market, as long as development cost can be lowered
to Microsoft level.
Certainly, Mac mini is very friendly for
personal use, just like iMacs. What we talked about
here is opportunity to expand its uses. As Apple
walks deeper into AI, the benefit of building up
technical computing support is apparent. Apple
may not choose the path, but it's a possibility.
Apple iMac and Household Battery
October 29, 2024
Apple did a solid iMac update. I still remember
considering adding a UPS to office iMac years ago
to avoid data loss due to power outage. These days,
there are too many devices to cover, so the better
way out is to add a household/office battery. During
power outage, ideally there should be options which
modules to retain power, so that critical devices
like Macs and Starlink can continue to work,
while air conditioners
can wait a while. As extreme weather becomes more
and more common, we need sustainable development
as well as mitigation measures.
Outsider
October 27, 2024
I'm a outsider. There are lots of advantages
being a Apple insider, but the contributions from
outsiders are non-trivial. Insiders have all the
gathered intelligence, but outsiders have their
own unique experiences.
Compared with Microsoft and Google, I find
Apple is much more willing to listen to outsiders
and make adjustments, rather than throwing whatever
technology they have at the user.
Most recent encounter is the option to turn
multitasking on and off on the iPad. There are
people living on multitasking, and almost all
media outlets ruled in favor of it, but Apple
still make the option available so people have
choices.
I happen to be the guy who turns multitasking
off.
It's really a honor to be part of the
product formation process. I understand there
is something awkward telling engineers what to
improve, but should be no problem if engineers
receive most of the credit. On the other hand,
I get a better product!
Apple is a highly collaborative company,
not limited to insiders, but outsiders as well.
They have the genius to make right decisions,
which is very rare. Outsiders can comfort
themselves that Apple will make good judgment
based on suggestions.
Perhaps I will never have a chance to be
a Apple insider, but I'm happy. Apple respects
customers and users. The thing is, I probably
will still shamelessly speculate on future technology
to see if anyone makes it.
The Value of String Theory
October 26, 2024
String theory has many social media critics.
Some critics support other theories of everything
against string theory. They think a theory of everything
is merely a theory that makes quantum mechanics
and general relativity work together. They can't be
more wrong. It only takes a simple question of
the existence of dark matter to invalidate naive
theories of everything. String theory may not
be right, but it tried to predict new physics,
beyond making QM and GR compatible. If a theory
of everything doesn't solve the dark matter
problem, it will be truly laughable. On the other
hand, string theory actually tried to predict
dark matter, which is honorable effort. We will
only point out that string theory critics mostly
consist of those people incompetent in string
theory. That's it.
Family Cloud
October 19, 2024
Family computing features like Family Sharing
and iCloud Shared Photo Library are examples
of family cloud. While some believe family computing
should only be a thin layer on top of personal
computing, there are reasons to build family
computing into a distinct category.
Family computing is about groups of people,
which is fundamentally different from personal
computing. Take Calendar for example. We typically
subscribe to others' shared calendars, read only,
but family calendar should be automatically shared
across family members and editable.
It's obvious we need sophisticated family
cloud to properly serve families. Consider
family shopping. Clearly we may not want kids
to over-spend on credit cards. Who can make
purchase and how much needs to be a control.
It might sound little, but the whole thing
is more complicated than personal computing.
Since family structure can change, the
family cloud needs to address it. Perhaps it's
better to start simple. A person can only
belong to a family at a time, and data can
only be available to members. I don't quite
believe in family emails, but family
telephone sounds reasonable.
As noted
in previous posts, a lot of things are
indeterminate, but I bet success belongs
to which corporation can ship it simple.
Don't underestimate the challenge. Making
family cloud simple is very hard, especially
considering integration with personal computing
devices. Here we only explored a little.
Hope there is chance to say more.
Apple Business Connect
October 17, 2024
Companies can
run
marketing campaigns
on Apple Business Connect. It would be nice
if people could subscribe to updates, just
like social networks. We don't check a list
of stores regularly, but can instantly view
timely posts from the stores we follow.
Suppose a cafe offers limited-time membership card
discounts. If the information can reach followers
effortlessly, the campaign is more likely to
succeed. Currently, people have to arrive by
accident on Apple Maps to know it, which
is unsatisfactory. Also, posts
are better than emails, because followers
can maintain privacy. That said, I still
can not claim my business on Apple Business
Connect, and would very much appreciate help.
A Preview of Some Novel Gadgets
October 15, 2024
Gadgets are not vanilla Turing
machines, but devices with
hardware/software/services integration,
like Multi-Touch sensors. Although
many gadgets have comparable computational
power, their integrated characteristics
can be very different.
Therefore, devices for family computing
can not simply copy personal computing
devices. We
speculated a little.
Perhaps eventually form factors will change,
but the need to address families as groups
of people, designs for home uses, and
interoperability with personal computing
devices will remain.
Vision Pro and smart glasses are great
for home uses. As previous posts indicated,
when we service a garden, there may be many
unexpected plants, and headsets provide
instant identification without the need to
bring out the iPhone. These devices should
work with family computing as well, like
browsing family calendars.
Very likely, everything works beautifully
together. We don't need to worry about
killing iPhones.
To get a idea of family computing, look
at how we do online family shopping. These
days, people visit Amazon on iPad to place
orders, but before buying, we chat and
research about products with family members.
Wouldn't it be nice if everyone can look
at a big screen simultaneously and discuss?
On the mobile end, Vision Pro and smart
glasses are convenient when we don't need keyboard
and iPhone, but heavy with AR. Inspect a nuclear
facility to determine its safety. Learn about
a particular corner of the city. Or just
relax and enjoy recording restaurant food
clips.
Possibilities are endless. Not to mention
people will invent new instruments to work
with these devices. Remote robots? No problem.
How about household batteries? When we lose
power outside, there is still electricity to
sustain life. With privacy and security,
it can be great help, unless authorities like
governments inject a trojan horse, of course.
Membership
October 14, 2024
These days, people manage all kinds of
memberships on mobile devices. We have
accounts for banks, restaurants, and
periodicals, etc. Unlike Apple Wallet,
there is no app to collect memberships,
so we end up messing around.
Some restaurants use LINE to offer
membership discounts, coupons, points,
etc. It's quite clean and convenient,
but you must have LINE installed.
Some banks use apps, but instead
of offering web apps for those without
installed apps, they redirect users
all over around, ending up using,
say, Apple Messages.
Some periodicals use Sign in with
Apple. Then people manage memberships
with subscriptions in iOS.
Overall, it's easy to forget where
to look for membership information.
For example, LINE doesn't carry chat
records onto clean new devices, so
people may not find it. It's a small
problem, but nobody really provides
a solution.
Paywall
October 12, 2024
People who like paywalls often think
information carries a price and free stuff
is marketing material. As a advocate of
open research, I can provide a counterexample.
I publish research for free because
anyone can examine it so that verification
can be more rigorous. It's not out there merely
to gain readership. As my finance doesn't
depend on publication, I don't have to
flatter the audience.
Indeed, paywalled material may contain
marketing materials to flatter the audience
so the publication can sell better.
I'm not against paywalls, but do want
to emphasize that paywalls don't necessarily
guarantee information quality. While my
research posts on
linguistics
and
economic geography
are not paywalled, they are not worse
than paywalled academic papers.
Finally, I want to emphasize that
there are extremely valuable materials
that are kept secret rather than published,
paywalled or not. People might not be
able to establish TSMC by reading all
paywalled papers. That's real market.
Meta and Linux
October 10, 2024
Today I would like to talk about Meta
and Linux, but let's talk about the value
of theoretical speculation first. I know
people like John Gruber do not like speculation,
but I also know Alan Turing speculated about
modern computers long before their existence.
Because transistors were invented much
later than Turing-completeness, Turing
had no choice but to speculate, rather than
construct modern computers. Gruber would
belittle Turing, but without Turing's vision,
the modern computer revolution would lack
significant motivation.
It's because Turing, rather than Gruber,
said computers can
perform intelligent tasks that people rush
to make fast computers happen. So here
we are.
Turing isn't a product developer like
Apple, but is no less great.
Now back to Meta and Linux, if Meta
really embraced Linux, it could mount a
serious challenge to Apple and Google.
The difficulty is ecosystem, but under
current situation, it's possible to find
valuable devices to build and eventually
recover Apple/Google/Microsoft.
Gruber would say copycat, but would
be entirely unfair if Meta truly mastered AI.
Who says UI is the only thing that matters?
I can understand Gruber's desire to be
the Darth Vader of Silicon Valley, but
I prefer a better future for mankind.
Of course, Meta might not want to make
Linux mainstream. Then we have nothing to
say. It's good to have Apple and Google
around, but living Linux life with Pine64
isn't bad either. The choice is up to
you, Meta. Open source innovation!
Google
October 9, 2024
This year's Nobel Prizes honored Google,
which politicians are trying to shut down or
break up. Excellence deserves destruction,
while mediocrity deserves power. Please don't
laugh, Apple fans. The next might be Apple.
After USA government ruins the Silicon Valley,
maybe we will discover PRC technological
superiority, and switch from export controls
to import controls.
Apple Critics
October 7, 2024
As much as I admire Steve Jobs, I do
enjoy the opinion of serious Apple critics,
not like John Gruber or Tim Sweeney, but
like Yann LeCun. Yann rightly criticized
Apple's AI backwardness and lack of generous
support for all kinds of serious research.
Unsurprisingly, Gruber's defense is that
Apple is not backward, but preferred being behind
rather than ahead, and Siri sucks.
There really is a problem. While I agree
that Meta and Google alike are not as good
as Apple at turning research
and technologies into products,
the lame defense provided by Apple elites
doesn't help. What if Apple permanently stayed
behind in AI? It's not as if others can't
catch up on iPhone.
Instead of welcoming people of all
backgrounds, Apple elites like to draw a
sharp line to declare they are Apple, you
are not. This sort of exclusiveness may
work in fashion or media, but to tell the
truth, I can happily live with Linux,
and use no Apple products.
I use Apple products because I appreciate
Apple's genius and hard work, but Yann LeCun
is also right that Apple failed to win over him.
John Gruber might wish to say Yann LeCun is un-Apple,
but that intended insult would be a compliment,
badge of honor. Come on, drop all advanced AI
in Apple products.
Daring Fireball is worth reading, but
serves as a Apple apologist all too often.
Occasionally, it revolted against Apple to
favor power users instead of ordinary people.
I don't understand why there can not be
options that suit both purposes, but
do understand that it doesn't help against
DMA coup.
That said, I hope no companies are trying
to become the opposite of Apple. There is
nothing wrong being un-Apple, but Apple does
have a lot to learn from. And yes, I will
remain a regular Daring Fireball reader,
but please don't create a caste system
and leave me no choice but to embrace Linux.
Silicon Valley Dragged Down
October 5, 2024
As the current
political climate
continues, Silicon Valley will face difficult
political challenges. We outline two disaster
scenarios. Hope they won't happen.
First, Trump becomes POTUS and shuts down
Google, threatening retaliation against all
critics. The Biden administration is already
considering breaking up Google.
Second, Harris becomes POTUS and treats
the Valley like everything Elon Musk. While
the Valley is much more diverse, politicians
paint in big brushes. The left already
threatens to arrest Elon.
The disaster scenarios are made possible
by propaganda like The Tech Coup. Considering
both parties wanted to destroy Google for
not licking politicians' boots, be cautious
what makes a politician.
Debate Bot
October 3, 2024
Yann LeCun is a intelligent researcher,
but he is also very talkative. In his domain
of expertise, he is truly formidable, but
otherwise his understanding is quite spotty.
He is building superhuman AI. I suggest
building Yann LeCun debate bot to debate
Yann LeCun. If the bot can not defeat
Yann LeCun, it's not superhuman intelligence.
If the bot defeats Yann LeCun, we know
he is not that intelligent. Either way,
there is something to learn.
Government
October 2, 2024
Newsom made a wise decision regarding
AI regulation. I suppose it's good time to
talk about government. Governments can do
a lot of good things, but they tend to ruin it.
What does that mean?
Let's talk about public digital library.
As elucidated in our book
in progress, government efforts can provide
equality of opportunity. So, why not elect the
radical left to do it?
First, the radical left is more interested
in fraudulent antitrust than libraries. It's
better for them to make everybody equally poor
rather than to help the poor.
Second, when the radical left obtains
government power, its instinct demands the
promotion of radical left ideas, so public
libraries are not designed for the people.
Third, the radical left is very quick
to turn against each other for his/her own
sake. By the time they reach power, they
have become a power animal incapable of
common good.
The message is clear. Government can do
a lot of good things, but be careful who you
vote for. Make demands and oust politicians
if they don't serve the common good. Study
the world so you have good judgment.
NFT and Data
September 30, 2024
Blockchain found significant applications
in Bitcoin. The left was shocked, so they
trashed the technology. To some degree, the
left can be considered successful as the price
of many NFTs dropped to zero.
However, there is something fundamentally
unsound to trash technology based on some of
its applications. Here we propose a possible
use case for NFT-related technologies:
immutable data.
The problem of authenticity is critical
for many activities. Take science for example.
The experimenter may provide the theorist with
data. How does the theorist verify these raw
materials?
Inability to verify authenticity can lead
to fraudulent conclusions. Don't laugh at
social sciences. Perhaps perfected NFT technology
can provide trusted source of data that may
stand the test of time.
There are lots of people working in DeSci.
Aside from being freed from academia, we feel
the need to use blockchain to improve technical
processes in addition to social institutions.
Freedom is the freedom to say two plus two
equals four.
Headsets
September 27, 2024
Like Apple Vision Pro, Meta Orion is
a groundbreaking product. But, basic questions
still need to be answered. Are UI elements
as stable as those in visionOS? Does it
solve virtual keyboard? Is dropping VR a
good decision compared with spatial computing?
These are pretty tough questions. It
is related to whether Orion can become iPhone
killer. If Orion can't beat Vision Pro,
and Vision Pro can't beat iPhone, the
answer is negative.
Certainly, Orion can be useful in gardening
just like Apple Vision Pro with plant identification
capabilities powered by AI. However, can it
match the graphics of visionOS? I'm certain
postcards can work well in visionOS, but
Orion looks pretty shaky at present.
All will settle in time. Here we have
several pretty interesting possibilities
for useful headsets. Maybe I'm old, but
I do not sense them as iPhone killers.
Happy coexistence is what I feel most
reasonable.
FTC on Privacy
September 23, 2024
Recently, FTC expressed their concerns
over privacy on social networks. While I
agree that privacy is a big issue, FTC's
understanding of the problem is likely
flawed.
It has to do with the distinction between
privacy as keeping data private and privacy
as knowing what you subscribe to.
According to the former, Facebook shouldn't
use surveillance AI on events to provide more
informative suggestions. According to the latter,
private events are private, public events are public.
Clearly we may want surveillance AI to suggest
public events instead of browsing them all.
Understanding aside, the administration's record
on privacy isn't good. Insecure technology was
forced into Apple Messages. It makes me wonder
whether FTC's concern over privacy is genuine
or a power play aiming at playing the Valley and
voters for fools.
Please read previous posts and our
book
in progress for more detail. At present,
please note that the biggest surveillance
enterprise is the government, but FTC
only criticized corporations. Only a bit
hypocritical?
Giants
September 22, 2024
Today, X/Twitter served me a post
belittling Steve Jobs while praising
Dennis Ritchie. I can imagine a much
better world where C and Unix are
widely appreciated, but the post misjudged
these giants. True, C and Unix helped
Steve Jobs a lot. On the other hand,
without Steve Jobs, we'd probably still
stuck with Microsoft. Can we stop
judging giants by popularity?
AI on iPad
September 21, 2024
Apple Intelligence was announced without
a world model. It primarily works with personal
intelligence, which is great for iPhone. But,
iPad is much more content heavy, so we
speculate that a world model is suitable for
powerful iPad AI.
Take a relatively trivial example, like news.
I tried to google what important events happened
today. It gave me a list of newspaper articles
and websites, which is not helpful. ChatGPT gave
me Stephen King's birthday, which is definitely
designed for his fans.
It would be nice to have a good news outlet
rather than depending on ever-trolling media. The
challenge is huge, but it might be more useful
than Elon Musk's X. While I do agree that legacy
media misbehaved a lot, X really looks like
a spurious replacement.
News AI requires real-time world model. It
might just be implemented in a search engine,
but conversational style can help with content
summary so that there is no need to browse
page by page. ChatGPT isn't quite there yet,
but good math is already a good thing.
At present, I find little reason to upgrade
iPad. People understand iPhone much better
than iPad, which explains why iPad often lags
iPhone releases. This year, although I hope
base model iPad will receive Apple Intelligence
treatment, there is much to be done to make
iPad stand out.
Hardware
September 20, 2024
Daring Fireball's iPhone 16
review
focused on Camera Control. We wish to say
that the entire hardware package is so
amazing that few can appreciate the full
depth of the update. Although I will wait
for iPhone SE, iPhone 16 lineup really is
a bargain.
Future of the iPhone
September 13, 2024
As the success of the iPhone continues,
more and more politicians wish to
put their own ass on it. Today we address
a central issue critical for iPhone's success
but often misunderstood. It's about integration.
For those who think integration is merely
assembling things together, iPhone will become
more successful after breaking up Apple so that
each component can be perfected through competition.
Unfortunately, integration is much more
complex. A classic example is Maxwell equations
for electromagnetism. Politicians think
Maxwell merely summarized earlier experiments.
The truth is, Maxwell invented a framework
so that earlier results became special cases.
In the process, Maxwell added numerous
corrections, theoretical notions, and mathematical
tools. The end result not only unified previous
experiments, but also made revolutionary
predictions which resulted in relativity
and gauge theory.
Apple introduced Camera Control a few
days ago, while Google, with unsatisfactory
integration with fragmented Android, can
only improve camera in Made by Google.
Politicians don't understand, but that's
the kind of integration iPhone thrives upon.
It's pretty clear Camera Control isn't
merely assembling parts together. The parts
were imagined, designed, and produced from
ground up. It's not possible for many separate
companies to unify vision and execution.
Breaking up Apple means no integral innovations
in the future.
Microsoft and Intel didn't deliver the iPhone,
but Apple did. The same holds for iPad, Apple
Watch, and many more to come. Politicians are
trying to ruin Apple once and for all.
They might succeed. But, please understand,
we will remember Apple's genius, and possibly
only politicians' ass.
Black Swans
September 5, 2024
If you model a fat-tailed distribution
with a Gaussian, it's easy to be surprised
by 10-sigma events, or black swans. However,
even fat-tailed distributions are not required,
for a simple change of distributions can give
10-sigma relative to the original.
The latter is the case for water levels,
which are definitely bounded, rather than
fat-tailed.
More generally, there is also the possibility
of measure-zero events, which are not
discernible in distribution. For example, if we draw from a
deck of cards repeatedly and only get finite
number of jokers in infinite draws, jokers
are of probability zero. This is where
non-parametric statistics can be useful,
though not the naive empiricism Taleb criticized.
Overall, the field is still very under-developed.
Noise can lead to black swans in a chaotic fashion.
Few people are prepared for noise. If it happens at
a critical junction, consequences can be serious.
Yet people normatively think in terms of convergence
and equilibrium. Good luck!
Explorers
September 4, 2024
Every decent undergraduate probability
textbook features the central limit theorem.
Some proofs required the existence of fourth
moment. Bright students might wonder whether
there is a world where the central limit theorem
doesn't hold. Then the professor would
extinguish their speculation.
Such is the background for the rise
of Taleb. Taleb took the possibility seriously
and explored the consequences of fat tails.
Many regarded Taleb as the most brilliant
thinker of our time.
While there is no doubt about Taleb's
contributions, it's quite easy to see that
he is a passionate explorer surrounded by
idiots, rather than a giant. As every early
explorer of a rich field, discoveries led
to dubious theories for future generations
to rectify.
To a much more severe extent, the same
happened to Stephen Wolfram, who Taleb
recently described as Newton of our age.
It reflects the suppression of certain
explorations in scholarly studies, especially
in the academia. It doesn't require Newton's
genius to overcome suppression, but
surrounded by idiots, Wolfram certainly
looks like Newton.
To make the situation worse, academics
criticized Wolfram's personality, rather
than his dubious theories. It all makes
the academics look much more like Soviet tyrants.
While the final judgment isn't available
yet, I fully appreciate Wolfram's discoveries
and try to make the
matter more satisfactory.
Perhaps academics should learn something.
Time Dependence
September 1, 2024
While Yann LeCun's trust in economists has
been unwavering, Taleb is much more skeptical.
Indeed, economists' misuse of statistics is
so rampant that the profession is fraudulent.
However, Taleb's treatment of black swans
as tail events is also questionable. Typically,
Taleb ignored the possibility that probability
changes with time.
As we mentioned before,
bookstores were impacted by Amazon only fairly
recently. Amazon's rise due to the Internet
changed the probability distribution of bookstores.
Black swans once impossible became possible.
There is good reason to change mindset upon
circumstances, as indicated above. However, it's
not Krugman's doctrine, which amounts to that
truth doesn't matter, as long as you are good
at changing positions.
Please refer to the section on noise in
our book in progress for more
detail.
Generalized Algorithms
August 30, 2024
People have been using the word algorithm
vaguely for centuries. Turing defined it as
a Turing machine that halts on all inputs.
The entirety of computational complexity theory
was built upon it.
It's therefore natural that our
relative complexity theory requires
a generalization of algorithms. While exact
definitions are more precise, it's better just
to think of it as a recipe for operating a
machinery here.
The generalization is not without past
attempts. Since the arrival of parallel computing,
it became clear that thinking in terms of
a Turing machine was not enough. But the
proposed definitions were all related to
Turing machines, and not very general.
A benefit of thinking in terms of
generalized algorithms is that geometrical
processes find natural expressions.
Ribbon
concordance is a machinery to produce more
and more complex knots, for example. A
algorithm can mean to derive a knot with
a Morse function.
As we have seen in relative complexity
theory, computational complexity is not
a satisfactory characterization of real
world complexities. By introducing generalized
algorithms, we can talk about the complexity
of real world processes in very specific
language. It's a rich new field of study.
Privacy
August 29, 2024
The left regards privacy as keeping data
private rather than knowing what you subscribe
to. Therefore, surveillance is evil for the
left. The left is unsound. Please refer to
the section on privacy in our
book in progress.
Determinacy and Indeterminacy
August 27, 2024
Since outcomes of a indeterministic process
can always be regarded as deterministic in the
rear mirror, arguing about determinacy versus
indeterminacy is unfruitful. But, whether a
sequence is computable or whether a theory has
particular structure
is of interest. Determinacy doesn't imply
computability, contrary to many people's
beliefs. Computable reals are of measure zero.
Public Opinion
August 17, 2024
The early stage of the digital revolution
took place in the age of neoliberalism.
With relatively little restrictions, Silicon
Valley revolutionary companies were allowed
to realize their dreams. The scenario is
about to change.
In response to Asian development model,
USA embraced industrial policy. Government
sought to play a major role in shaping the
economy. The problem is, the Asian success
was due to competent politicians, which many
believe is not the case in USA.
Copying the Asian model risks undermining
USA.
If the political climate continues, Silicon
Valley needs to change its mindset. Freedom
is no longer guaranteed. Corporations need
to persuade voters to elect reasonable politicians
if the Valley is not to be destroyed.
It's that serious. We've seen bunch of
ridiculous policies and regulations.
The most recent hit on the Valley is
the threat to break up Google, a vote
against integration. Some wise men already
wrote that without integration, there
would be no iPhone. There is value in
disintegration, but the neoliberal case
is to let market judge which approach
works, while current USA government
wishes to decide with (idiotic) politicians'
foresight.
One doesn't have to accept neoliberalism
to appreciate the dynamism that freedom brings
to the Silicon Valley, which is currently
under severe attack from USA politicians.
Voters are deceived. The Valley remains
largely silent. Traditional lobbying likely
will backfire. Will Apple begin to consider
they need to persuade the public to
elect reasonable politicians?
Mining Robots
August 12, 2024
In the name of humanism, the left opposed
AI automation that may displace employment.
They wanted more human casualties instead of
mining robots.
Clearly, AI mining robots can improve
productivity as well as saving lives. All we
need to do is to help displaced workers
find more secure jobs.
The economics of the left is anti-human.
The reason why we are not seeing
waves of mining robots is not entirely due
to the left, however. There is also the
fact that miners' lives are too cheaply
priced.
We need to provide proper incentives for
the AI transformation, so that mining robots
can help us live better lives. Let AI do
the laundry, and you do the poetry.
Evidence
August 2, 2024
Everybody knows Taleb disliked the word
evidence. He used black swans as a example
against evidentiary science. However, in
the case of black swans, we knew they exist
because of their evidentiary existence.
The real point is no amount of white swan
occurrences can justify the non-existence of
black swans. That is, it's about our logic
applied to evidences, rather than evidences
themselves. One particular case is the
problem of induction. The fallacy resides
with people's reasoning, not observed objects.
The real lesson is that people's reasoning
regarding evidences is often seriously flawed.
Competition Myth
July 31, 2024
In the pursuit of antitrust, the left
claimed that competition is for greater
resilience. They took bookstores for example.
Bookstores were highly competitive and
therefore resilient to the extent that
bunch of them were slaughtered by Amazon.
We are left with the question how the
left were so wrong.
Don't laugh. The problem is not merely
the fact that competition doesn't imply
greater resilience. It's about the problem
of induction.
Induction draws conclusions based on
available data in order to gauge the
truth. Day after day, sunrise happened,
therefore induction says that sunrise
will happen as usual.
Unfortunately, induction is often
a terrible gauge of truth. For centuries,
induction suggested Newtonian mechanics,
only to be refuted by relativity and
quantum mechanics.
Economists thought competitive bookstores
are resilient because they persisted for
centuries. Then Amazon came, familiar?
Yet politicians are dominated by some
defunct economist's thoughts.
There is merit in competition. There
is virtue in resilience. But, the claim
that competition is for greater resilience
is political bullshit. Think of it, what
if Amazon is a Chinese company? What to make
of USA bookstores? I hope politicians'
misguided efforts can stop, or USA will
be weaker instead of stronger.
Food
July 26, 2024
There are plenty of criticisms about
digital technologies, but today I'm going
to share a upside: food. These days I eat
a lot better and in far greater variety than
when I was a kid. The reason is informational.
First, there are local food blogs and
Instagram that document food explorations.
With engaging multimedia and inviting texts,
these channels build credibility through
good taste, as readers can try it and decide.
Second, Google Maps provides current
updates regarding restaurants and cafes.
Through crowdsourcing, menu, dishes, and
the environment are instantly accessible.
Better, searching for local intrigues is
no problem.
Third, everybody can contribute and
share. I use InFrame to make collage and
Darkroom to tone the final image, all of
which work seamlessly between iPhone and
iPad. If Apple wants to make Maps better
for food, at least let people share info.
It was close to impossible
for me as a kid to explore the entire
city and find new places to try. With
digital technologies, both quality and
variety improve. To be sure, I learned
through experience, but the same can be
taught in schools. There are lots of
upsides using digital technologies.
Let's discover.
WWDC24, Part 7
July 22, 2024
WWDC24 brought massive OS releases. It's
impossible and not our intention to cover
all aspects of it. To conclude the series,
we talk about philosophy, which is what,
I believe, separates Apple and Google apart.
The different takes
on personalization provide a clue. Apple
develops technologies that grow with users,
while Google throws whatever technology
they have at users, hoping they would
appreciate Google's might.
For Google, it's AI first, not AI that
serves humanity.
Certainly, Google has its fans, but I
appreciate Apple's approach more. The iPad
is a tablet extremely useful for any age,
while Google tablets are not intuitive enough
for toddlers.
Google's defense is that people use computers
because of search. See?
While elitist tendencies have more grip
on Apple these days, Apple is still much more
considerate for ordinary users, which is why
I believe the DMA coup is utterly disastrous.
The reader may consult my
book in progress.
Looking forward, our hope is that Google
will pay more attention to good design, while
Apple will abandon elitist attitude. There
are much more people out there who appreciate
Apple products that are not John Gruber.
And Google really needs to rework Search
AI Overview. So, that's for WWDC24!
WWDC24, Part 6
July 18, 2024
USA government forced Apple to support RCS
in iOS 18. At first glance, it's good news, but
as everything concerning modern digital technology
politics, there is
significant drawback doing so.
Unsurprisingly, RCS is
terribly
insecure. Considering politicians considered
requiring backdoors in digital communication,
RCS is merely a logical consequence.
Clearly, there are better ways to regulate
messaging apps. Government may demand a open
standard for secure messaging, or simply give
people a choice to use Google messaging in iOS,
but inline with DMA coup, everything is
Apple's fault.
We can say Apple is too complacent regarding
politics. However, the fault is politicians'
ignorance of digital technology, and in some cases,
malicious exploitation. Now people have no
choice but to accept the presence of insecure RCS.
Does the problem of green bubbles warrant
compromising users' communication security? Or
should we develop a secure open standard? I'm
sure the latter is superior, but people must
wake up to the understanding of political manipulation
and start to demand truly considerate regulations.
WWDC24, Part 5
July 11, 2024
Apple showcased code generation in WWDC24.
At basic level, it's very useful and time-saving.
Students can now quickly complete school programming
projects by tapping into its power. However,
as wisdom suggests, subtleties are of tremendous
importance when we use it for real.
A primary issue is algorithms. As everybody
knows, there are countless algorithms all performing
the same or similar tasks. How did Apple Intelligence
decide which algorithm to use? Should it be the fastest?
Or use the least memory? Or parallel? We believe
it will be much more helpful if Apple Intelligence
can handle algorithmic subtleties.
Another consideration is refactoring. We wish
to write good code, by some standard of good. Can
Apple Intelligence comply with the standard and
produce coherent organization? Whether it's new
code or re-organizing existing code, can it
be generated in a specific style?
One of the most challenging tasks in programming
is debugging. Can Apple Intelligence examine
code and compare with function description to identify
bugs in code? This feature is very hard to implement,
as code generators can check outputs rigorously
to improve until no bugs exist, if the feature
is present.
Before these issues are resolved, the world
still needs plenty of traditional programmers,
whose judgment is paramount to successful development.
But, suppose Apple Intelligence matures one day.
Imagine how much effort can be automated and
time freed for really difficult and profound
production decisions. We shall see.
WWDC24, Part 4
July 7, 2024
VisionOS 2 addressed several past
issues,
including keyboards and precision. The solution
is that you just use ordinary keyboards. It's
workable, but clearly few people carry keyboards
around inspecting a building, so the use cases
are limited. Perhaps speech to text can help.
Since I never used Vision Pro, speculation
is all I offer.
Some people are disappointed that
Vision Pro is more of a enterprise product, rather
than a household wonder. Spatial computing is at
very early stage, so nobody really knows where
the future goes. Nonetheless, interesting apps
like plant identification for gardening may
merge both enterprise and household needs.
As for connecting people, Persona may be
less attractive than straightforward message
exchange. I can live with social networks
without expressions on a avatar. Content
matters more, it seems. Two people on Vision
Pro can readily view/manipulate and communicate about
3D models, which is already impressive.
Finally, our major speculation is that
pro apps can make or break Vision Pro as a
high-end product. It requires sophistication
to justify beyond personal entertainment.
What are the pro apps? Well, perhaps virtual
architecture for the LHC, much more informative
Maps augmenting the surroundings, or material
gallery for industrial deisgn may serve
as a guide.
WWDC24, Part 3
July 4, 2024
Apple Books is a app that served me
since Day 1. It hosts important materials
for study and research. Although it has been
indispensable, its development was very much
a drag. Apple announced Books is accessible
by Apple Intelligence in WWDC24. I hope it's
a good restart.
We must face the fact that Apple Books
faced regulatory retaliation in USA. Lawsuits
might have made Apple reluctant to develop the
app. That said, here we enumerate some basic
objectives that should be improved.
File management in Apple Books is a mess.
It not only eats up my iPad storage, but refuses
to free up space after local copies are removed.
I was forced to reinstall the app. The UI for
categorization is not good, either.
Thumbnails for books can not be enlarged,
which results in users guessing the content
of research papers. I simply don't know which
PDF document is the right paper by browsing the
bookshelf. Search doesn't
help. Most search functionality is useless.
To be realistic, there is incentive for
Apple to keep stuff back to allow a prosperous
app ecosystem, but Apple Books really is
backward. There are thousands of professional
features that can be added to a research app,
but Apple Books only needs to get the basics right.
As I have no idea how deep Apple Intelligence
is able to improve Books, it remains to be seen
if reading can be made more pleasant with features
such as summary. However, the app definitely
needs to progress, not to mention to offer
advantages
compared with printed books.
WWDC24, Part 2
June 27, 2024
There are numerous details we will talk about
later, but today, I wish to use the opportunity
to address a very narrow view of research and
product development that came from the academia.
Some academics draw a sharp line between research
and product development, claiming the latter isn't
the former, and in some cases, even regard industries
as applied academia.
These people don't know about IC, which was
both outstanding research and innovative product
development.
Take a step further. People may watch Jensen
Huang's keynote and get a picture how GPUs can
accelerate AI and sciences, similar to LHC. That
is, products can serve as great research tools.
Not only so, the invention of the iPhone not
only employed sciences, but itself became a object
of serious study, the origin of countless research
papers.
By now, the reader should see a highly productive
relationship between product development and research,
academia and industries. It's foolish to denigrate
the scientific potential of product development. It's
arrogant to regard industries simply as applied
academia.
To offer a little connection between WWDC24 and
Jensen Huang's keynote, we speculate that many AI
appliances can readily employ the iPhone and
Private Cloud Compute to accomplish sophisticated tasks
without designing a new machinery from ground up,
like signal processing in music production, or
medical devices. It's a rich research area.
The reason why industries are often overlooked
has something to do with private information, as
industrial secrets don't appear on published papers.
Another reason is that industrial synthesis is often
misunderstood as nothing new.
We wish to build a healthy relationship between
product development and research, academia and
industries, that don't belittle one another.
WWDC24 showcased Apple's innovation in privacy plus
AI. It's clearly brilliant product development as
well as the beginning of future research. Don't
be fooled by vanity.
WWDC24, Part 1
June 22, 2024
I haven't finished watching the videos yet,
so here are the first impressions. Given the
expectations prior to WWDC24 and Apple's
AI weakness, it's clear that Apple should deliver
basic stuff with superior privacy first.
Inline with our suggestions for intelligent
app interoperability with personal focus under
algorithmic privacy, Apple did roll out pretty
solid basic stuff. Still, it's very important
to evaluate Apple's implementation, as many
rivals like Google botched the game.
For some insight, we may compare Apple
and Google's takes on personalization. Google
did personalized chatbot. Apple used personal
context to deliver relevant information across
apps. The distinction is significant. In our
opinion, Google really squandered its leadership.
On the other hand, Jensen Huang's outlook
for AI is very different from Apple's. While
many commentators may argue which company is
right, we believe both of them have their
valid cases. It all boils down to the fact that
Apple focused on consumers, NVIDIA on industries.
Before we dive deeper, it's useful to
remind people that AI is still in experimental
stage. Therefore, the performances are hardly
conclusive. We will see long and winding road ahead,
so please be patient. At present, we can fully
appreciate considerate incremental updates.
Economic Nationalism
June 6, 2024
We will soon talk about WWDC24 and Jensen
Huang's keynote in Taiwan, but here is a
reminder that Jensen Huang definitely knows
better than Taiwan government, so he offered
to develop AI in Taiwan with USA support,
rather than Taiwan economic nationalism.
Our suggestion
is working out!
Google Search AI Overview
May 29, 2024
It's characteristic of Google not to take
our caution seriously.
Now they released what Paul Krugman called worse
than useless. Don't say I didn't warn you!
Clusters of Thoughts
May 25, 2024
At some point, it became clear that both
Stephen Wolfram and John Gruber needed to be
seriously examined. They represent schools of
thought popular to followers and serve as
their guardians. Wolfram is the advocate for
computational reductionism, while Gruber is
the archbishop of everything Apple.
Wolfram literally trashed geometric complexity
with his declaration that dimension doesn't matter
in complexity theory, which can be found in numerous
official sources. His treatment of geometry lacks
the understanding of a working mathematician.
A knot can be represented in a 3D grid, but
on a 1D Turing machine, the knot inevitably loses
adjacency connectedness. Wolfram's defense is that
adjacency can be recorded symbolically like pointers.
Therefore, connectedness is restored! Ask anybody
whether they wanted to be disassembled onto a
Turing machine tape and to stay connected symbolically.
More seriously, if Wolfram figured out how
to explain geometry with computation, he should
be able to explain the fine strucutre constant.
But, clearly he avoided the supremely important
open problem in physics, and moved on to a theory
of everything more ambitious than string theory.
In Wolfram's theory, Trump's victory was simply
a computational consequence of some simple origin
of the universe.
As for Gruber, it's understandable there has
to be a evangelist defending Apple. But Gruber
tried to be the evangelist on everything Apple.
As Gruber gave ground in DMA and antitrust issues,
virtually there is no prominent voice to shape
popular opinion in favor of considerate design
for ordinary people, rather than power users.
Meta has Yann LeCun, and Apple has John Gruber.
I know it's harsh, but Gruber's Apple priority
list is telling, which says Apple serves Apple first.
Users and developers are not on the first tier.
It's monopoly-style speak, not something you
hear from a innovative company. But, clearly
Gruber doesn't care to prove Apple is in oligopoly
position, rather than monopoly position.
That is to say, if someone strongly influenced
by Gruber decided Apple is a monopoly in lawsuit,
don't be surprised.
We didn't elaborate on Wolfram's theory of
biological evolution or Gruber's defense of
Apple's AI backwardness, not because they are
unimportant to discuss, but because those takes can not
be taken seriously. It's enough to shed light
on the general characteristics of their fallacies.
It's more satisfactory to consider relative
complexity than computational reductionism,
and to consider the benefit for humanity than
Apple leaks.
Created Life
May 19, 2024
While it's likely that current life on Earth
descended from natural selection, we suspect life
can be created, and thus biological evolution
in its current form is a incomplete theory.
It's taboo for biologists to discuss alternatives
to natural selection, but the possibility is there.
Tales of the Digital Revolution
May 15, 2024
I'm writing a book about the digital revolution
for laymen. Considering original materials may
soon be independently discovered, the
book (PDF) in progress
is posted here, so that accusations of plagiarism
can get lost.
Hypothesis Tests
May 12, 2024
Hypothesis tests are logically unsound,
but under suitable conditions, errors may be
controlled. However, these conditions, like
i.i.d., are often difficult to verify in
practice, and thus naive application of hypothesis
tests may result in gigantic errors. We have
already said this countless times. A simple test
for i.i.d. Bernoulli can be fraudulent if i.i.d.
isn't checked, because other combinations of
random variables may also result in the same
distribution. Beware!
ChatGPT Uses
May 6, 2024
As a emerging technology, AI has its
hits and misses. ChatGPT has been criticized
for various errors. While there is no
defense for its failures, we do find some
good uses. Here we discuss how to use
ChatGPT to find out whether a media
figure is a crook.
Media figures won't label themselves as
crooks. At first try, one might expect
ChatGPT to check logic and facts to reveal
them as such, but ChatGPT is too polite to
offend these propagandists.
Instead, one interrogate ChatGPT regarding
media figures' historical records, and find
out when ChatGPT gets apologetic. As ChatGPT
says something like there is a difference
between ideas and perception of reality or
something bullshit, one knows the media
figure is a crook.
It may very well not be ChatGPT's most
powerful uses, but the ability of AI to
parse through huge amount of records is
worth something. Considering there are too
many crooks in the media, naive people can
use ChatGPT to defend themselves against
propaganda. We will discuss more as
technology advances.
Smart Glasses
April 29, 2024
In the heated debate for AI-powered
gadgets, some claimed the superiority of
AR smart glasses over the iPhone. We
believe smart glasses UI is for very
specific population, while iPhone UI is
for the masses. Therefore, smart glasses
are not going to be iPhone killer, not
even in 10 years.
Complexity in Algebraic Geometry
April 28, 2024
While it's obvious how
relative complexity
can apply to gadgets or physics, it's very
much a research topic for algebraic geometry.
Our sense is that since a partial order system
is a category, it should find suitable place
in abstract algebraic geometry in the form
of categories. I'm working on it, and welcome
suggestions for a non-trivial theory.
Humane AI Pin
April 16, 2024
There is a fashion trend in media to
root for iPhone killers, iPad killers,
Mac killers, etc. Humane bought into it,
and developed AI Pin, without reading
our protest that not every technology is
Internet Explorer.
Then, Humane repurposed AI Pin as
a convenient wearable, but we already
have the Apple Watch. Certainly, Humane
can use some good judgment.
TV and Family Computer
April 15, 2024
Steve Jobs once imagined TV integrated
with iCloud under a simple user interface
is what TV needed. Since then, most people
still view watching movies as the primary
purpose of TV. It's very difficult to
envisage TV otherwise. It seems hopelessly
dumb.
Home automation came. TV became a hub
for appliances, but the whole idea is still
dumb. What's the problem?
We believe the issue lies with the lack
of efforts to build TV as a family computer.
It's not just about games, but relationships
with people/families.
TV can be home telephone, can help family
shopping, can facilitate holiday travel
planning, can watch your back while you are
not at home. It's just nobody is developing
for it.
With proper UI, we may organize a party
event on TV, contact all participants, arrange
party-time entertainment music, so on and so
forth. The advantage of TV is the natural family
connection, rather than being solely a personal
computing device.
As Apple released Vision Pro, TV's status
as the superior home entertainment gadget
is utterly shattered. Without family computing,
will it be shadowed indefinitely?
Complexity Theory
April 13, 2024
Our goal is to produce a complexity theory
that properly addresses the role played by
computation, statistics, and geometry. It's not
a trivial task. As a first step, a outline of
relative complexity is provided here to generalize
Turing machines, so that computational complexity
becomes a special case.
We postulate that complexity depends on what
cognitive instrument is at work. A task that appears
complex to a Turing machine may not be complex
to a probability process at all. A task that can
not be performed by a machinery is called too
complex with respect to it. We always speak of
complexity with respect to a machinery.
Machineries may be put into a partial order
system according to what tasks they can perform.
So, we can say that a Turing machine with a
stochastic register is more complex than the Turing
machine part, because there are tasks beyond
computation.
Logical depth can be put into this picture
by considering performing computation on a Turing
machine within certain number of steps. Kolmogorov
complexity can be put into this picture by
considering what output a Turing machine can
generate given certain length of input.
This theory can handle the following important
case. Theories with different geometrical structures
may be Turing equivalent, but some are chiral, while
others are not. If we make the distinction between
these machineries, we may say chirality is too
complex for a simple universal non-chiral
cellular automaton.
This amounts to the fact that Wolfram's classification
of processes into computational machineries with
equivalent computational power is too crude.
They tell no difference between chiral and non-chiral
theories.
Furthermore, we define easy as simple to do,
hard as complex to do. A icon simple to recognize
may be complex to produce, which reflects the
more vague experience that simplicity isn't always
easy to achieve. Clearly, we need more dimension
theory to account for it in our complexity theory.
All the essential materials here are put
forward in a form or another long ago. Since
recently there is renewed public interest in
complexity theory, we say it again so that
the topic may be more broadly understood and
confusion may be avoided. We are tired of
unnecessary and unjust damages from public
misunderstanding.
Probability
April 12, 2024
It might appear that our complexity theory
contradicts with the result that probability
offers no better efficiency than usual Turing
machine programs. Here we discuss how such
contradiction doesn't arise.
The result is for problems that can be
solved by Turing machines, which is quite limited.
A Turing machine can not generate a sequence
with Kolmogorov complexity O(n), while a probability
process can. No contradiction occurs because
we are not talking about a computable problem.
More generally, suppose a person throws a
fair coin. Without observation, we can compute
the probability 1/2, but for a observer, certainty
can be achieved. That is, observation provides
information beyond computation. This is one of the
most common uses of probability, rather than
enhancing the efficiency of algorithms.
Therefore, there is a role for probability
in complexity theory, contrary to Wolfram's
claims. It's just that we are not talking
about fast algorithms. Hope this post offers
some clarity for laymen, because the damage
from confusion can be huge.
Antitrust
April 4, 2024
Recently, the Biden administration announced
huge boost for No. 1 market monopoly, Microsoft,
by pursuing antitrust agenda against Apple. Perhaps
they really hate Steve Jobs. We will see.
Sheaf Cohomology on Finite Topological Spaces
April 2, 2024
Here we postulate that the sheaf cohomology
of the constant sheaf on
finite topological spaces resembling simplicial
complexes can be computed via Leray's theorem.
The key is to show that a simplex has no cohomology,
which could be proved via flasque sheaves and short
exact sequences. Thus, the sheaf cohomology of
these spaces can be related to the usual algebraic
topology of simplicial complexes.
Apple's Developer Cloud
March 11, 2024
There are many approaches to AI. Apple
probably will release GenAI similar to
ChatGPT, Firefly, Gemini, etc. The difference
is that Google and Microsoft have infrastructure
products for customers to build AI themselves,
while Apple's developer cloud is pretty weak.
In our opinion, ChatGPT won't dramatically
boost productivity, but making a wide array
of AI products available will. It's better to
have differentiation based on expertise, rather
than relying on a chatbot. Hospitals might
buy AI products without using ChatGPT.
Currently, Microsoft and Google have the
advantage of possessing tools for hospitals
and medical companies to build AI on their
own infrastructure. It's wise for Apple to
follow suit, as far as productivity is concerned.
More generally, Apple can pay more attention
to developer cloud that will enable a complete
solution with AI integration. Backend is messy,
but if Microsoft and Google deliver, Apple can
only sell interface devices, a pity.
If Apple decides to limit its own scope,
a interface device to AI is fine, but then
the company risks being sidelined, not to
mention the benefits a unified scheme for
mobile, cloud, and AI can provide. The question
is whether Apple is willing to empower developers,
or merely rely on Microsoft and Google to
do the job.
VR Postcards
March 10, 2024
It's Yana Rumyantseva's birthday today.
She is the best
artist
I've ever seen in my
limited time. I lost her information long
ago. This is a post remembering her.
It's clear that GenAI may
produce artwork for a
flat panel, as well as Apple Vision Pro.
Since generating VR postcards by hand
isn't easy, it's logical to ask AI for
help. We believe AI will deliver.
Postcards can include AR elements
to resemble a physical postcard. Then
the viewer can launch into a whole
new VR world to see its contents.
It can be amazing postcard experience.
The main challenge is the authoring
environment. Macs can not do VR, while
Vision Pro is awkward for postcard edits.
Perhaps it's best to have a set of professionally
designed templates to choose from.
Most probably, people record spatial
videos on iPhone or Vision Pro, and
add them to well-designed templates
together with a message. Then let
AI generate a world view for background.
Yana Rumyantseva's taste is more like
a elf. She might not like the technology
presented here, but anyway, it would be good
to have a chance to send her VR postcards.
Wish her a happy birthday.
UI Certainty and AI
March 5, 2024
There are
attempts
to kill apps with AI. While AI is
transformational, these attempts likely
misunderstood what apps are. Apps are
not just bunch of code to be executed,
but a user interface to humans.
The problem with typical AI user
interface is that it lacks
certainty. I
used Calendar and Notes to schedule tasks
everyday, and there is no problem with UI,
but if I needed to ask AI to set and view
tasks, let alone organization, I would
be pretty pissed off.
It's simply much easier to just tap
and launch apps to manage tasks than asking
AI.
In our view, AI isn't a user interface
to replace apps, but a infrastructure project
to support apps. Imagine generating VR
postcards with AI and view them on Apple
Vision Pro. It's awkward to ask AI to
check postcards, but a postcard icon
will do fine.
In other words, AI makes apps more
powerful. We can use AI to decorate
videos, which is tremendous improvement
to video editing, all within video
editing apps. We believe this is proper
use of AI, rather than going without apps,
because there are other operations within apps
that are much easier without AI.
Of course, people might disagree, but
the point is that markets will be our
final judgment. We will see if AI kills
apps or empowers them.
Window to the World
February 24, 2024
The iPad has been a powerful gadget
turning people into couch potato. For some,
it's a depressing fact and they lost faith
in the gadget. For us, it's also a depressing
fact, but we believe it reflects people's
inappropriate use of technology.
The problem is that it's relatively
easy to watch movies on iPad, rather than
to use it as a enlightenment. The iPad
can be a window to the world, but people
have to know how.
Here we take a simple example of coffee.
Search can provide information on coffee shops,
beans, and flavors. Maps can provide guide
to nearby cafe. Social networks let users
join coffee lovers around the world to
learn and grow.
Similar uses can be applied to gardening,
algebraic geometry, economics, or whatever.
The point is to resist the temptation of
degeneration. We need to have a school class
teaching how to properly use gadgets, or
risk turning future generations into
couch potato.
Geometry and Complexity
February 16, 2024
We elucidated how computational thinking
drops
topological structures before, but didn't
emphasize its significance. One of the most
important implication is related to the Standard
Model. There are computational models with
imposed symmetry that can not handle chirality.
These models can be made Turing-complete, and
thus capable of performing all computations.
Therefore, we are forced to accept that
the ability to handle chirality has to do
with symmetries, rather than computation. That
is, geometry affects complexity, in addition
to computation. Wolfram's attempt to reduce
complexity to computation is unsound.
Internet
February 10, 2024
Democrats' attitude toward digital technologies
has been contradictory. They worry about Apple's
power over the App Store, while Krugman declared
the Internet a disappointment. As everybody knows,
the App Store is built upon the Internet, so
Democrats are saying Apple's App Store is too
powerful, but the Internet sucks. If it sounds
ridiculous, you are not alone.
Imagine a world without the Internet. Instead
of Googling through papers, we rely on limited
paper records in local library. Instead of
building AI in the cloud, we run AI locally so
it can not take over the world. Instead of
communicating with loved ones via FaceTime,
we see distant loves only once or twice per year.
What a wonderful world!
This wonderful world is what Krugman would
suggest because the Internet sucks. It sucks
so much so that the iPhone should not exist,
since it's a breakthrough Internet communicator.
I'm sure there are still environments that
can fulfill Krugman's utopia. Let Democrats
follow their economics sage.
Trump raised considerable political capital
through the Internet. It must leave Democrats
with the wisdom that the Internet should be
suppressed, like some other countries did.
The tragedy is that there are very few who
don't buy Trump and are trying to remedy the Internet
through digital civilization, instead of
restricting the technology.
We don't know if Trump will prevail, but
Democrats' utopia with Krugman's foundation
is the best antithesis to the Silicon Valley.
If USA undermined the Valley, other inferior
countries would take up the opportunities to
go ahead and innovate. It's a great industrial
policy doing the rest of the world great favor.
So why not?
Vision Pro Spatial Navigation
February 3, 2024
Apple Vision Pro is clearly a first
generation product, groundbreaking and
remarkably limited. Some limitations,
like isolation and weight, might take
much time to overcome, but others, like
what we are going to talk about, may
be overcome soon with great benefits.
Although Apple deprecated VR in favor
of AR in the introduction of spatial
computing, there is actually a VR issue
in Vision Pro: spatial navigation. In
AR, one simply places apps around in front
of the physical background. In VR,
it becomes necessary to navigate the
3D landscape, sans physicals.
We believe it's beneficial to do
VR justice in spatial computing. Gestures
allowing users to move around in 3D
landscape can enable numerous apps,
like gaming and model evaluation.
There is no need for keyboard in this
case, and it can help Vision Pro
go beyond watching videos.
The use for gaming is obvious, so
we talk about model evaluation here.
When designing machines and houses,
there is often
the need to look inside and inspect.
On Macs and touch devices, this can
be cumbersome. However, with proper
spatial navigation, Vision Pro can
be put to great productive use,
sans keyboard.
Whether VR is worth it is subject
to Apple's decisions, but we feel
it can make the product much better.
Like the Apple Pencil, some controllers
may help along with gestures. It's
important not to judge Vision Pro by
revenue and profit in any time soon,
but good VR already has a niche base.
Worth a try.
CarPlay Privacy
January 31, 2024
Privacy is a central concern for Apple.
While there are numerous compromises, there
is no doubt that Apple takes privacy seriously.
Here we wish to raise a small issue regarding
CarPlay that may be improved.
The problem is the absence of a CarPlay
toggle to enable/disable CarPlay. It's imaginable
that when people give control of their car
to another person, they don't necessarily
want to transfer the control and information
of their iPhones.
Recently, I went to car maintenance
and felt uncomfortable for mechanics to
have access to my iPhone via CarPlay.
All I wished for is a CarPlay toggle
in iOS.
Certainly, this is a small request, but
it can do a lot of favor in addition to
battery saving. Of course, these days
everybody is talking about Apple Vision Pro,
and we will take a look at the gadget
soon.
CarPlay Toggle
January 25, 2024
Apple CarPlay is a nice feature as
I listen to music while driving alone.
However, it drains battery and there is
no MagSafe charger for iPhone SE. Therefore,
I really appreciate Apple can place a
CarPlay toggle in iOS to enable/disable
CarPlay so that battery life is under control.
Rising Tide
December 30, 2023
2023 is almost over. 2024 is almost here.
Famous people, like Bill Gates, have already
published their outlook for the next year.
Therefore, we feel it's appropriate to look
into the future, but this time, the remaining
21st century.
For learned people, 20th century is undoubtedly
the century of physics. Not only does the subject
deliver unprecedented breakthroughs into the workings
of fields, but can be regarded as the pinnacle of
human intellect.
Unfortunately, the brightest moment so far for physics
has already passed. In the near future, there
exists experimental barriers to understanding
high energy phenomena. Wise men desperately
search for new outlets for their own creativity.
Judging from the developments of early
21st century, it's easy to see the supposed
glory of finance was totally misleading, and
the Silicon Valley proved to be the new home
for creatives. The question is whether the
on-going digital revolution will rival the
achievements of 20th century physics.
The digital revolution enjoyed widespread
support not seen in physics. From academia to
markets, far more talented people joined
the adventure than physics ever had. While
physics Nobel pretty much summarized the
advancements in the field, the digital revolution
is too great for even the Turing Award to capture.
The existence of iPhone isn't predicted
by physics, as it's independent of the Standard
Model. Revolutionaries in the digital age
aren't merely carrying out derivations from physics.
The barrier to entry is comparatively low for
computer industries. Everybody can have their try.
It's not unreasonable to speculate the 21st
century will be the century of the digital revolution.
However, the digital revolution was largely
advanced by practitioners and scattered theorists.
While the future is bright, there is a need to
employ the physics model of comprehensive
theory-practice collaboration. It's unwise to
rely on historical accidents to advance the field.
Our complexity study
and proposal for a unified scheme for mobile,
cloud, and AI are a first try.
From my personal perspective, it's also
critical to build a digital library of algorithms
at spectrum-dev. As computers become more
and more popular, they should not appear as
black boxes to determined learners.
Future-gazing is remarkably difficult.
That's why there is a place for science fiction
to explore ideas. But, by engaging in real
speculation, the reward is far greater.
Although we possess limited foresight,
the practice of sticking one's neck out
may sharpen our senses as well as provide
valuable hypotheses to be confirmed or
refuted. Happy New Year!
Hostility from Apple Elite
December 23, 2023
It's Christmas weekend, a good time
for reflection. Despite AI underdevelopment,
Apple's record has been impressive. A related
scene is the rise of Apple elite. One can
easily spot them, for their purpose is to
predict Apple's trajectory, rather than
proposing something beneficial to humanity.
There would be no problem if Apple elite
restrained themselves to their purpose. However,
Apple elite often exhibits unpleasant traits
that trash disadvantaged people. It's similar
to class warfare. Apple elite reserves Apple
to themselves, but we feel it's better if
Apple can benefit humanity.
There are plenty of attacks from Apple elite.
When we revealed the nature of mobile gadgets
as a tool for modern surveillance, Apple elite
proudly denounced the possibility and regarded
the issue largely as a Google defect. But,
Daring Fireball has already published several
pieces on phone surveillance, and it's not
Google-specific.
Advanced readers may wish to consult the
book, Underground Empire, for more information.
The presence of surveillance control from
Washington alike predates mobile gadgets.
It's a superpower, as Krugman's introductory
article on Foreign Affairs demonstrated.
Apple elite is more naive than a regular
Krugman reader.
Another Apple elite attack is focused on
our formulation of a unified scheme for
mobile, cloud, and AI. They think it's
un-Apple. However, as our purpose is to benefit
humanity rather than to predict what Apple's
plan is, we feel it's of minimal importance
whether Apple, Google, Microsoft, or whoever
delivers such benefits.
Here we make clear the benefits of such
a unified scheme. Take postcards
for example. Mobile makes sending and receiving
HTML5 postcards in Messages easy. Cloud keeps
large files, like video, on demand and
maintains history. AI helps text and image
generation for postcard content. They
work seamlessly together, but Apple elite
wish to tear it apart.
The case can not be more obvious if we
consider books. Mobile makes reading books
enjoyable. Cloud keeps library and syncs
across devices. AI helps readers dig deeper
into content. The transformational power
of such a unified scheme can improve our
civilization at large. Yet, it's un-Apple
for Apple elite.
Wish you all a Merry Christmas. For
those disadvantaged, we feel your pain
and hope the new year will be better.
For Apple elite, may we ask a favor that
people should be treated well? We live
in a complex world that appreciates
simple UI, and our collective future
is full of challenges. We need all
the blessings. Do not divide the world
into haves and have-nots.
AI on Small Screens
December 17, 2023
Although it's obvious for devices
like the Apple Watch that AI is critical,
how to do it right is extremely challenging.
A aspect is the small screen size that
requires condensed information.
On the iPad, ChatGPT's full response
can be easily presented and scrolled through.
On the Apple Watch, however, there is
no such screen real estate. Speech doesn't
seem like a universal replacement for screen.
Thus, a much more concise response method
is needed.
Perhaps it will involve a chain of
interactions, with each step gathering
small pieces of information, like finding
restaurants nearby, then showing open
hours, then providing map location, and
so on.
Of course, Apple Watch already does
a bunch of AI, like double tap, crash
detection, fitness, etc. The point is,
for ChatGPT-like apps to work, there
has to be a revolutionary response method
suitable for small screens.
Postcards
December 9, 2023
I miss postcards. It's a subtle way of
deliberate communication with personal style.
Social networks killed postcards with real-time
interaction and video, but the format feels
markedly postmodern, not a continuation of
the postcard tradition.
There is no doubt we can not bring back
traditional postcards at scale, but we can make
live social communication more deliberate with
personal style.
Collage apps are a good start. However,
more interactivity and dynamic content are
much needed for expression. The problem is
social networks don't allow HTML5 freedom,
but limit the style with default boxes.
I'm not sure whether Apple is bold enough
to allow HTML5 freedom in Messages app, but
it can be much more than Live Stickers. The
disadvantage of Mail is lack of social
organization, and digital postcards are best
presented with social networks.
These are little things, but they can
express lots of love. The cynical drive in
Silicon Valley has done away with
many such subtleties. However, if we don't
want to become Microsoft, some taste has
to be restored.
Artwork Review
December 8, 2023
When Apple launched the Freeform app, the
default scenario is probably brainstorming,
but we find other uses plausible, namely,
artwork review. We haven't tried it, but feel
that in the simple case of a commissioned image,
Freeform app can serve as a great feedback tool
between artists/designers and clients.
Artists/designers typically create images
in professional apps like Photoshop. Usually,
clients don't have these apps. The conventional
process is to get feedback via email or social
networking. There are clearly huge disadvantages
like lack of annotation, organization, and
real-time participation.
All of these drawbacks can be solved via
Freeform artwork review process. However,
Freeform isn't designed for this specific purpose,
it's imaginable that improvements can be made
to address the artist/designer-client relationship.
For example, can messaging be integrated?
Take a step further and consider review process
in general. It's desirable that images, videos,
sounds and so on can all be presented in suitable
form for review, like at which time a frame needs
a edit, or what instrument needs to stand out more.
At minimum, there should be a timeline.
We think of this use because we consistently
used the iPad for artwork
reviews, and find Freeform can be a great addition.
App interoperability likely will help the entire
workflow instead of doing everything in Freeform.
We know it's unintended use, but remain hopeful
something can be done with it.
App Collaboration
December 5, 2023
When collaboration is desirable, people
are tempted to use such app functionalities.
The easiest way is to co-edit a document on
the cloud. Recently, Apple introduced co-editing
of local documents via native apps. Security
concerns aside, there are actually cases
where app collaboration is undesirable.
In my experience, a important consequence
of writing one's own article is a consistent tone.
App collaboration can easily turn a document
into dissonance. In such cases, traditional
review process often works better than
arbitrarily adding others' contribution.
It's easy to imagine app collaboration
may thrive where personal style doesn't matter
much, like code. Git is a very successful example,
but it takes monster merges to truly deliver
benefits, not to mention the complexity of
modifying histories. Can apps like Pages
implement such features for workplace?
We are not against app collaboration,
but do think it takes good design to work well.
Perhaps it's prudent to start with the
Reminders app. A shared list of Reminders
can be used to organize family activities
like holiday parties. It's much less complex
than office documents, so let's see.
Time Machine
December 2, 2023
The biggest reason I used Apple products
isn't that they are fancy and shiny. It's
because they are considerate. A great example
is Time Machine, used for automatic backup.
But, as cloud popularity grows, people gradually
forget Time Machine.
That is, until they need it.
As Apple ID can be stolen, cloud storage
is not safe. It's recommended that people
also keep local copies of their machines.
For Macs, it can be done, but to backup
iCloud Drive, files still need to be downloaded
first. There is no automatic backup for cloud
files. It's understandable that cloud storage
may be too large for local backup, but
automatic should at least be a option.
For iDevices, there is only despair.
Although restoring from local backups might
not be a good idea for iDevices, there should
be a mechanism for maintaining local backups
for files. Some files are simply too precious.
One may argue that Apple's focus is elsewhere.
However, being considerate can really make a
difference. Hope there will soon be a way to
secure access to files, in case Apple ID
is stolen.
Interface to Knowledge
November 30, 2023
Recently, there are numerous people referring
to GenAI as a interface to knowledge. It's clearly
untrue, but there is more to
the issue than occassional nonsense
from GenAI. It has to do with the distinction
between knowledge and wisdom.
By knowledge we mean statements that can
be rigorously proven to be true. It might
sound very broad, but the truth is that
we don't even know whether usual axiomatic
systems for natural numbers are consistent,
and therefore can not even regard
arithmetic deductions strictly as knowledge.
Nonetheless, we use these axiomatic systems.
This is wisdom, which is much harder to define,
but in practice works. With this distinction,
it soon becomes clear that knowledge engines
are very limited.
We've talked about
limitations of knowledge
engines, or interfaces to knowledge before.
Here we point out that knowledge engines are
unsound even for knowledge production.
It has to do with the
difference
between rational sciences and empirical sciences.
For rational sciences, theory is knowledge.
For empirical sciences, theory is speculation
before being confirmed or falsified by evidences.
Knowledge engines can be quite useful for
rational sciences, but in the production of
empirical knowledge, speculation has to be
integrated into the process, and much speculation
is destined to be wrong. A interface to pure
knowledge does not include speculation and
may not be that useful for empirical sciences.
The desire to label GenAI as a interface to
knowledge stems from the undisputed superiority
of knowledge. As we have seen, it can be hugely
misleading as well as counterproductive. There
is no doubt we need information systems based
on wisdom in addition to knowledge. Can we set
our eyes beyond limited knowledge engines?
Digital Revolution
November 25, 2023
Predicting the future is hard. Wise men
subject predictions to qualifications. A
great example is Steve Jobs, who famously
changes his mind upon circumstances. Here
we talk about the opposite of Steve Jobs,
a foolish universal prediction
that the digital revolution is ending as of 2018.
Harvard Business Review has the fortune
of publishing a
article
with a idiotic title that claims digital
revolution will end. Of course it will end.
Even idiots know that. But then the article
reveals itself that the true thesis is that
digital revolution is ending as of 2018. Ha!
With all the buzz coming out of Silicon
Valley, the HBR article can not be more wrong.
Therefore, it's beneficial to investigate
the article's fallacies.
The article's misjudgment is based on
fraudulent inference from three observations.
First, Moore's Law is ending. Second,
the existence of no-code platforms democratized
tech. Third,
gadgets appear stagnant in terms of functionalities.
The observations are not entirely true, but
the gigantic mistake is to make the assertion
that digital revolution is ending because of them.
For digital revolution to end, all transformational
technologies have to moderate, but the article
clearly missed a great deal of technologies.
The difference between a wise man and a fool
is to tell the value of such claims in real time.
At least, a commoner should postpone judgment
as history unfolds. While many scientists downplay
the role of logic in arguments, it's obvious
that they are destined to be forgotten.
There is a Planck quote.
Collaboration
November 23, 2023
It can be phenomenal collaborating with
great people. The problem is, of course,
there are not that many great people. I've
found collaboration can result in disasters
for countless reasons, and tend to be more
cautious. Apple can easily find talents,
but I often need to fend off assholes.
If positive externality is not within reach,
then at least avoid negatives. Hope people
find better judgment regarding collaboration.
Tim Cook's Philosophy
November 20, 2023
Tim Cook did a very well-presented
interview
with Dua Lipa. They discussed very important
questions, and Tim Cook sincerely answered
according to his beliefs. It's recommended
to watch the interview.
Here we wish to add some points, not
to argue with Tim Cook, but to provide
another perspective.
The most important topic is, of course,
climate. Although Apple did very well and
Tim Cook can be proud of it, climate is
unfortunately a global collective action
problem, while Apple is a local solution. It's
crucial to recognize we are running out of
time and sound economic measures are needed
to address climate if we want to avoid
unpleasant outcomes.
And there is equality. The way Tim Cook
used the word actually means alleviation,
especially regarding poverty. His answer is
remarkably first world thinking, which is
suitable given his nationality, but third
world countries probably need to build a
sound economy along the way, which Tim Cook
didn't emphasize.
The topic of education is controversial
among learned people. Enlightenment is
probably better than education, because
we've seen many pedagogical
disasters,
like generative grammar scientists. The
lesson is that education may teach bullshit
and falsehoods. Enlightened students should
avoid such education.
AI responsibility is a hot topic. While
many people feel the need for regulation,
the central problem is actually good regulation
versus bad regulation, which most people
ignore. To regulate AI is easy, but to do
well is hard. For example, should self-driving
cars protect driver first or pedestrians first?
As readers may notice, interview isn't the
best place to deliberate on issues that matter.
Perhaps Tim Cook just casually illustrated
his views, but it's almost certain that refinements
are required for soundness. We hope readers
look deeper into the questions and arrive at
better conclusions.
Taiwan AI Policy
November 19, 2023
Economic nationalism dominated policy
thinking in Taiwan. Presidential candidates
wanted to pour national resources into
national AI private/public corporations/startups
in the hope that they will be the next TSMC.
Are they credible?
The most unconvincing part of their agenda
is that they never find any key person capable
of such AI feat.
It is therefore our suggestion that instead
of economic nationalism, Taiwan should welcome
Silicon Valley giant corporations to develop
AI at the island, rather than relying on its
own limited expertise. Will politicians listen?
Optimization
November 18, 2023
A common misunderstanding about optimization
is to regard it as a business problem. Actually,
optimization is a integral reformulation of
differential evolution, commonly found in sciences.
Under suitable conditions, the minimal action
principle recovers mechanics. That is to say
the Lagrangian method transformed Hamiltonian
mechanics into a optimization problem!
The roots of the misunderstanding can
be traced to the greedy method. Greedy method
is a naive optimization procedure that
finds locally optimal options and hopes
they lead to globally optimal outcomes.
Clearly, greedy method may fail for many
problems. But the failure of a naive procedure
does not imply that the whole idea of optimization
is unsound.
Business people usually optimize according
to the greedy method and fail. Unusual Titans
see the whole picture and optimize globally
and succeed. But optimization is not limited
to business. It's a philosophy that leads
to many intelligent applications, because
it may be very hard to solve differential
evolution, but relatively easy to find
approximate minimum of the integral.
Typical academics have low opinion for
optimization, for they think it's just
incremental iteration. They can not be
more wrong. Optimization is about the
big picture, not merely local improvements.
There are places where iterations may
converge to optimal outcome, but generally
optimization is a global problem.
Engineers may tackle optimization
better than scientists, because in practice,
experiments can give outcomes very hard to
derive from theory. Engineers devise
reliable experiments, while scientists
compute in vain. In many cases, catching
a ball through observation is better than
solving fluid dynamics computationally.
It's impossible to do justice to
optimization in a few paragraphs, but
we believe there is much wisdom to be
learned. Here, we postulate that AI can
be made more intelligent if we understand
more about optimization. There are people
against intelligence, but we will see
who wins.
The Turing Blessing
November 14, 2023
A argument
against AI automation is
the Turing Trap, which might as well be
labeled
as the Turing Blessing. We won't argue
against the Turing Trap, since economists'
words are highly speculative and uncertain,
but offer some clarity in the debate.
Automation doesn't always cost jobs.
It's easy to come up with examples like
spreadsheet apps that automate calculations.
Economists might shout that it's augmentation,
but the reality is it's both automation and
augmentation. Spreadsheet apps automate
boring parts of the job so that time is
freed for more productive work.
Linking automation with HLAI is highly
spurious. Since no human beings can produce
sophisticated semiconductor chips by hand,
machine automation used by TSMC shouldn't
be likened to HLAI. Automation doesn't
always imitate human abilities.
It's good to have AI augmenting human
work, but to use the assistant model is
quite limited. If economists can only
dream of AI assistants, but not AI
accelerator physics, there is no doubt
the analysis of AI augmentation is
likely meh.
Most importantly, economists' speculative
and often dubious assertions about AI
should not guide AI policy. We've seen
regulations based on computational complexity
that may become hindrance rather than
safety measures. AI is too important
to leave to economists.
There is enormous amount of work
to be done regarding AI and society.
Here we merely clarify some points,
but if even supposedly smart people
get these points wrong, there is reason
to put much money and effort in AI
research so that we can make sure
it benefits humanity.
Downstream Media
November 11, 2023
There is value and wisdom in Apple-related blogs,
but they really need to get rid of the condescending
attitude toward ordinary people. It doesn't require
much intellect to appreciate Apple products, and
the appearance of a Apple elite is not that attractive.
Do not divide the world into haves and have-nots,
wise men said so.
Multiway Systems
November 5, 2023
In Stephen Wolfram's original approach,
he made no significant distinction between
computation and multicomputation. We pointed
out that while computable enumerations give
Kolmogorov complexity log(n), multicomputation
can give n, so that the result can be made
consistent with standard quantum mechanics.
Wolfram updated his view later, but didn't
make it clear that log(n) is not standard
quantum mechanical prediction. Here we wish
to emphasize that computation and multicomputation
are different in terms of Kolmogorov complexity,
even though they may be Turing equivalent.
There are serious consequences from the
distinction, including explaining what's wrong with
Wolfram's dubious early judgment that
probabilistic AI isn't successful. We
feel it's better to tell the public here.
Connectedness
November 4, 2023
Stephen Wolfram's radical ideas might be
flawed, but they deserve examination. As noted
before, the notion of
complexity can not be satisfactorily explained
by computation alone, contrary to Wolfram's
claims.
Recently, Wolfram Physics Project tried to
deduce the universe from computational thinking.
While there are very dubious general claims,
here we restrict ourselves to technical details
to see how the build-up might be problematic.
The idea to build a discrete universe isn't
new, as many scientists, like Penrose,
have tried with some
success. To this extent, Wolfram Physics Project
is also successful, but the more aggressive
claims that relate generative graph theory to
computation are spurious.
In graph theory, the idea of connectedness
is topological, whereas in computational representation
of graphs, they are just symbols. It's like
a lego house can be disassembled, but the
topology isn't preserved. Although generative
graphs and Turing machines can be Turing-equivalent,
they are not topologically equivalent.
That is to say, computational thinking
drops additional topological structures.
In fact, even in the setting of generative
graph theory, the scale of the theory isn't
derivable from Wolfram Physics Project, which
may cause serious issues elucidated in our
previous note.
Hopefully, we will discuss other fallacies
in this blog later.
Now and Then
November 3, 2023
The last Beatles song, Now and Then, is a reminder
that love is in the little things and how technology
can make it possible.
Macs
November 2, 2023
The most discussed topic of the Mac
event is of course Shot on iPhone. Irony
aside, the Mac update carried a few implications
for consumers, beyond the performance of M3 series.
Touch Bar is officially gone. It's expected
apps that support Touch Bar will drop development
on the technology soon.
The entry-level laptop is now without question
the MacBook Air, which is still very capable.
For me, without a fan is a win.
If there is ever a need to go MacBook Pro,
the M3 option is far less attractive than the
M3 Pro and M3 Max. Choose wisely.
The performance of M3 Max is rivaling
M2 Ultra, which is only available on desktops.
It's a good time to go mobile.
The M3 iMac update is long overdue. While
it's debatable whether all-in-one is still a
popular option, iMacs remain convenient
for businesses like hotels.
The event is short, but very focused. However,
the maturity of the Mac means far less surprises.
Can we architect benefits other than better chips?
Algebraic Topology with ChatGPT
October 31, 2023
Sheaf cohomology on finite topological spaces
seems to be a interesting computational problem.
I asked ChatGPT, and it suggested using differential
forms, which is quite outlandish, as expected.
Of course, why not try Wolfram|Alpha? Aha, it can
not even correctly interpret the objective! Although
Wolfram loved to advertise the superpower of Wolfram
technologies, reality is far less pretty. Beware
of crap propaganda!
Intelligence and Content Pool
October 27, 2023
Recently, there is concern that GenAI
is squeezing content pools like Stack Overflow,
which, in turn, might hurt the data sources
used in GenAI. While the disequilibrium economics
is clear, we believe both GenAI and content
pools will find their own places in time,
if managed properly.
What GenAI replaced is the part of content
that's derivable from the pool. It's natural
for content pools to take a hit upon the launch
of GenAI. However, since new content isn't
necessarily derivable from old content, GenAI
won't kill well-managed content pools that value
logically independent propositions.
A better outcome is that content pools
host propositions not found in GenAI, and
then GenAI learns from the content pool
to update itself. It's advantageous that
GenAI and content pools are integrated to
maintain balance.
As everybody knows, better outcomes are
not necessarily the result of market forces,
contrary to standard economics. Therefore,
the economic problem remains to address the
concern mentioned at the beginning. There is
hope, but we need to find a way.
Krugman in Bikini
October 22, 2023
Adobe Firefly has been fun to play with.
However, it has moral principles that prohibit
the generation of certain images. I tried the
prompt, Krugman in Bikini, and it refused to
output. Funny.
The prompt, Steve Jobs funeral, does pass
the moral check. Adobe Firefly returned completely
counter-factual images that might feel offensive.
It's a wonder what the morals are.
It seems that one of the advantages of
using Adobe Firefly is to quickly prepare
illustrations for magazines. Although the
results can be counter-factual, it may
pass as artistic inspirations.
Alas, eventually people with sufficient
resources or access can obtain images for
Krugman in Bikini. So, beware of images.
Design and Fashion
October 21, 2023
Apple UI design isn't impeccable, but
it's vastly better than rivals, which is why
the lock screen on macOS Sonoma is worrisome.
Not about wallpapers, but the organization of
UI elements is alarming.
First, profile photograph is located at
the bottom together with password box. Presumably,
this gives a better view of the wallpaper.
However, it does look a little bit eccentric.
Second, a very small icon for power button is on the
top-right with a menu for selection. It used to
take 1-click to power down a Mac. Now it takes
2-clicks. And the icon is quite obscure.
Third, it does seem that Apple is determined not to
add widgets to the lock screen, or the
sole focus on wallpaper and time is dubious.
Information is deliberately left out to make
space for a recurring theme.
People might argue the Sonoma lock screen
is in fact quite fashionable, given Windows
precedents. Our view is that sacrificing good
design for fashion isn't worth it.
Logic in Sciences
October 20, 2023
To justify the reductionist agenda, Steven
Weinberg used the arrows of explanation to
try to reduce every science to physics. It's
beneficial to give a logical analysis of such
attempts, so that its prospects will follow.
From logical point of view, the Standard Model
can not predict whether Biden or Trump won
the USA Presidential Election, because whoever
won would be consistent with particle physics.
Thus, logical analysis gives a negative answer
to the initiative to reduce every science to
physics. Moreover, by organizing scientific
statements into logical chains, we may explore
sciences beyond physics, which may lead to
a unified approach to sciences.
Wisdom
October 16, 2023
Recently, I saw a X/Twitter post portraying
academics as Jedi, and industries as Storm Troopers,
and wondered whether they have ever met Sith Lords,
who are often very wise, but just regard ordinary
people as something to be exploited for profit.
The misunderstanding is a result of
distinct wisdom structures. Academics publish
wisdom to gain status. Industries reserve wisdom
for themselves, because leaks are often harmful.
Academics aren't necessarily more intelligent or
wiser than industries, but simply more publicly
visible.
Trust me. There are plenty of Jedi and Sith
Lords in industries.
It's not just that academics are often
detached from the real world, but there are
very wise men who don't care about Nobel Prizes.
Academics who thought industries are applied
academia are terribly misguided.
I believe there is no need to name them.
Advantages of Digital Books
October 15, 2023
The preference for printed or digital books
depends very much on taste and user experience.
There is no need to argue with someone with a
different opinion, but we document some advantages
of digital books here, in case they aren't clear.
Printed books occupy much space. Digital books
are much more economical. It's not merely convenience,
but a decisive distinction that human wisdom may be
better preserved in digital format. Most books
aren't classics that deserve to be printed again
and again, but offer valuable insights for the
right person.
Printed books use index for search purpose.
Digital books can be made fully searchable, including
images and videos. It might sound small, but if
we consider searching through huge volume of
references, the digital format is much more
practical and comprehensive.
Printed books can not be easily updated.
Digital books can be revised in time to meet
evolutionary pressure, like programming books.
In a transformative period, there is no end
to revisions. The digital format not only saves
space, but also delivers to wider audiences on time.
For serious readers, the advantages of
digital books can not be overlooked. During
my graduate study, vast amount of materials
are downloaded from the Internet, most of which
can not afford to be printed. Hope people
learn to appreciate it.
Challenges for Search Engines
October 10, 2023
Wolfram|Alpha fails as a general purpose
replacement for Google. Its main critique of
Google is a very general defect of all current
search engines, that they do not output knowledge.
While the conception of
a knowledge engine is likely limited and flawed,
it's true that Google has lots of problems.
Such sentiment is most obvious when we
consider how many people want AI to replace
Google. AI is still a very early stage technology.
In its current form, it may generate bullshit,
make up pseudofacts, and be utterly misleading.
The main announcement from Google regarding
AI for search is to summarize search results,
not promising.
Privacy is another issue, often brought
up by Apple fans. While it's not a critical
flaw, and a privacy focused Google is not
a much better Google, the user experience
of search engines really needs a lift.
Being asked to provide location on the web
is very unpleasant.
If information is a proper goal for
a search engine, Google must overcome the
desire to favor business over substance.
There are countless instances that a much
lower ranked web site provided much better
content than the site at the top. Google
should value content accuracy
with the help of AI instead of following
human herding.
The solution to these problems is likely
very complex, and by no means easy, but
it's possible that a much better search
engine with informative AI can drive out
Google. The most apparent recent innovation in
search engines is advertisement, which
is telling. Let's see if someone succeeds
where Wolfram|Alpha fails.
AI Strategies
October 6, 2023
Each Silicon Valley giant has its own
AI strategy. For certain companies,
it means more research and public participation.
For others, it means more AI-integrated apps.
It's not difficult to see the challenges.
Should it be one AI for many applications, or
one AI for each application? Should it be general
purpose AI with separate applications, or special
purpose AI for specific applications?
It's possible that giant corporations settle
these questions behind closed doors. While
expert opinions are valuable, the matter
with AI is that it's still rapidly evolving.
How to choose wisely among strategies given
limited resources?
Making mistakes in AI strategy may cost
leadership and reputation, eventually with
market response. It's a true test of wisdom.
We will see which CEOs are wiser than others
soon.
Privacy on Macs
September 30, 2023
Apple used to market the iPhone as a
much more private and secure platform than
Macs. Although it makes business sense
as iPhone is much more popular, I find
Apple's approach worrisome for Macs.
It has to do with setups.
Like most
people, I don't have more than one
smartphones. As a result, my iPhone is
used for private matter, school, and
work simultaneously. It never feels like a
very private object for me, especially
considering the need to show it to somebody
else on occassion.
On the other hand, ideally I would
employ a Mac exclusively for private matter,
with a separate Apple ID. It's a computer
I never have to show to the public. However,
since it's less private and secure than the
iPhone, the setup is flawed.
I wish there is a private mode on Macs,
so that the setup can work. Furthermore,
if Macs can be easily compromised and everything
syncs under Apple ID, wouldn't it be a
security hole for everything on iPhone
as well?
Reading Images
September 29, 2023
AI has been successful in image classification.
Although probably no one knows exactly how AI
did its work, there is good reason to use it.
Here, we propose a challenge for AI to read
images in cultural context. AI can identify
dogs and horses in Tarkovsky's films, but can
it grasp their meaning? To be fair, it's a highly
subjective matter, but there should be some
interpretations that feel sound rather than
bullshit. More generally, can we make a good
poet out of AI?
The iPad
September 21, 2023
As readers probably know, we are
great proponents of the iPad, even as
the mainstream
fails
to see its value. Not only so, we figured
the iPad can be embedded
in a bigger scheme
unifying mobile, cloud, and AI to deliver
dramatic improvement to our civilization.
The problem with the iPad lies deep
within Apple. Judging from recent developments
of the device, Apple didn't quite know
what it's headed to. That's why features
came to iOS first, then iPadOS. We believe
it's a terrible mistake. As post-iPhone era
looms large, there is a need to treat
each category on its own merits, rather than
a shadow of the iPhone.
The iPad has been good for business use.
Businessman can read newspaper and respond to
queries via messaging simultaneously. Yet the
claim that it's not a productivity powerhouse
misdirects its usage. It's not meant for
replacing Macs, but I'd rather comfortably
reply to a query on the iPad, rather than
sit idle in front of a Mac for hours.
The iPad has been good for learning and
research. People no longer need to carry
paper references, and with access to the cloud,
can search for relevant content conveniently.
Apple focused on Apple Pencil for notetaking.
While valuable, its scope is quite limited.
Apple has to take entire learning and research
process into account, rather than provide
piecemeal solutions.
The iPad has been good for exploring
the world. From browsing the web to finding
events to attend, the device is just right.
However, profit motives behind social networks
turned the device into a cesspool of
misinformation and garbage. It's a serious question
whether we can sort Facebook events by time.
The problem lies with Facebook, not the iPad.
The considerations above are very general
purpose, enough to establish the iPad as a
daily gadget. Our fear is that Apple is growing
more and more conventional and finally
consumed by mainstream media. Steve Jobs
wisely stated that Apple has only scratched
the surface of the iPad. Hope Apple still remember.
Apple Watch and AI
September 15, 2023
The iPhone event has passed, and the
upgrades Apple Watch received are solid.
Here we won't repeat announced features,
but give a perspective on the development
of Apple Watch.
From the beginning, Apple Watch relied
heavily on AI, especially biometrics.
Double Tap is a nice improvement.
However,
one wonders whether Apple has a thorough
AI strategy. Hints on Siri capabilities
seem to suggest that Apple is looking into
conversational AI on Apple Watch. It's
important, but does Apple already figure
out how to implement features?
Since nobody outside knows the answer,
we won't provide one. It's clear how
conversational AI can be critical considering
the small screen size of Apple Watch.
The hard part is to do it well.
Hope Apple will deliver.
TeX App Bundle
September 8, 2023
As everybody placed their focus on
the upcoming iPhone event, I wish there
is a standard TeX app bundle on Macs.
PRC Youth Unemployment
September 4, 2023
Unemployment has been a key element
in socialist critique of capitalism. It's
huge waste for PRC to allow high
youth unemployment. True, measures
like consumption vouchers are mediocre,
but PRC government should not ignore the fact
that cynical adversaries are happy to see
PRC failures. With the wisdom of Keynes,
PRC government can provide useful work for
young people, like teaching in rural areas,
or devise grand projects that will benefit
all population, like building public digital
libraries. Avoid avoidable misery.
Categorical View of Cohomology
September 3, 2023
A benefit of categorical view of cohomology
is that given suitable equivalence of categories,
cohomology is then automatically identical.
There is no need to compute and compare cohomologies
of algebraic varieties and those of corresponding
schemes. Thus, the embedding of varieties into
schemes is not only a enlargement, but preserves
important algebraic structures.
The Value of Category Theory
August 26, 2023
One of the most poorly taught lessons
in abstract algebra is category theory.
It appears to be useless when introduced,
and lacks motivation.
Here we provide a clue regarding the
value of category theory. Abstract algebra
can be roughly described as what can be said
given relations between objects, without
knowledge of the inner workings of these objects.
Category theory gives such abstract algebra
for relations described by morphisms.
Grothendieck's abelian category paper
is a demonstration that cohomology can be
defined from the structure of morphisms
between objects, regardless of whether these
objects are abelian groups, modules, or
sheaves.
It remains a mystery why category theory
works amazingly well in algebraic geometry,
but lacks traction in other areas. Perhaps
future study will reveal.
Post-iPhone Era
August 25, 2023
As the iPhone loses novelty, people began
to speculate about post-iPhone era. There are
a few serious proposals. One is AI, led by nVidia.
Another is spatial computing, led by Apple.
Our view is that these proposals are fragments
of a scheme unifying mobile, cloud, and AI.
We have elaborated
benefits of such a
unified scheme, whose conceptual existence
we elucidated long ago. Here we explain
the superiority of the scheme compared with
fragmented approaches proposed by nVidia and
Apple.
The advocacy of AI largely surrounds its
content generation capabilities, most of
which are just bits of information. These
bits are valuable, but without the
mobile-cloud paradigm, their usefulness
is limited. Mobile allows us to use AI
anytime anywhere. Cloud allows us to use
AI much more resourceful than the
local model.
Spatial computing is exemplified by
Apple Vision Pro. Although the headset
will never replace the iPhone, spatial
computing is likely a new frontier worth
exploring. Apple's strategy is to shift
some focus from the iPhone to this new
frontier. The obvious future is integrating
spatial computing into the unified scheme,
rather than making Vision Pro the
flagship product.
We didn't provide many details here,
but the advantage of a scheme unifying
mobile, cloud, and AI is undeniable.
While Apple probably thought it's
impossible to surpass the iPhone revolution,
the reality likely will prove otherwise.
The question is whether it will still
be Apple.
Lebesgue Method in Finance
August 18, 2023
Here we propose a method of financial
investment where the long-run trend of the asset is
upward, but short-run fluctuations
are problematic. It's named after Lebesgue,
not related to measure theory, but rather,
the idea of computing Lebesgue integral
with partitions on the y-axis instead of x-axis.
To illustrate with a example,
the asset can be taken as a good index fund.
First, take a arbitrary entry point and choose
a appropriate step size on asset price. Buy
1 unit at the entry point and 2n
units if the price goes from n-1
times the step size lower to n times
the step size lower. If the price goes to
a step size higher than the entry point,
reset the entry point.
It's a strategy to buy at low prices given
fluctuations and uncertainty. Instead of buying
at regular or random time intervals, we buy
according to prices. Sales can be conducted
similarly in reverse. We recommend this method
as a good strategy
for managing assets.
How to Read a Book with AI
August 12, 2023
The advancement of technology often
begins with technical considerations, but
successful consumer technologies are designed for
humans. There is no doubt AI has been a
technical achievement, but how to make
AI into a consumer technology remains murky.
Other than obvious threats such as
surveillance, there is another obstacle
for mature AI consumer applications, that is,
wisdom versus bullshit. It should act as
a bicycle for our minds, but all too often,
corporations opt for cognitive bombardment.
Here we propose a challenge for civilized
AI, to help people on reading. Serious reading
involves understanding, analysis, and synthesis.
In each stage, AI can help with meaning, reference,
and organization. It's plausible that well-designed
AI can have significant positive impact on
people's reading.
However, giant corporations are busy
building personalized chatbots, rather than
figuring out how to help people. The excuse
is business, but we believe civilized AI
will be better rewarded in the end.
May wisdom prevail over bullshit.
A Bicycle for Our Minds
August 5, 2023
Today, we wish to raise a important question.
From the age of Steve Jobs to Tim Cook, computers
went from a bicycle for our minds to a device for
mindless scrolling. What happened?
The answer is still unknown, but we will attempt
to give a few sketches in future posts.
Wisdom Online
July 22, 2023
Blogs were a great platform for forming
wisdom online. Records are kept. References
are extensive. Arguments are well-formed.
From small food blogs to economic debates,
blogs carry the weight of ideas. The
problem is blogs are in decline.
The decline of blogs is accompanied by
the rise of social networks. At first glance,
it's simply a shift in platform, but careful
consideration reveals one serious concern.
Social network posts are generally meant to
be forgotten. Hot takes get tremendous visibility
while tightly woven expositions are soon
neglected.
People began to aim for constantly at top
of their game, rather than to deliberate
a sound perspective for the long-term.
While many economists laugh at the possibility
of social networks intensifying extremism,
we believe the nature of social networks has
much to do with the rise of Trump. It's just
nobody wants to talk about it.
Therefore, the social network transformation
has big impact on wisdom online. Can wisdom
regularly fend off bullshit in hot takes?
Can social network algorithms be made more
reasonable given profit motives? Can people
be trusted to validate social posts? The
answers are unclear.
Activities
July 1, 2023
Isolation is a nasty issue in contemporary
digital life. Arguably digital entertainment
and other mechanisms have turned people into
sorts of couch potato, with fewer close friends
than before.
Digital transformation is inevitable. A
natural question is to ask if isolation is
inherent, or rather reflects our inappropriate
use of digital technologies.
We believe isolation isn't inherent, and
offer a case for consideration.
Recently, we began to plan a gathering
to meet new friends. A topic is chosen,
about graphic design. Keynote on Mac is
used to prepare the presentation, with
content from the web.
A nice place to host the gathering is
found on Google Maps. The event will
be posted on social networks to seek participants.
Without Keynote and the web, the presentation
wouldn't exist. Without Google Maps, it's
hard to find a nice place for the gathering.
Without social networks, people interested
in the topic would have a difficult time discovering
the event.
Digital technologies can connect people,
but people have to overcome couch potato mentality
to build meaningful relationships.
The problem is proper use of digital technologies,
not technology itself.
People can not be expected to discover
proper use of digital technologies themselves.
A approach similar to Apple's introduction of
iLife is appropriate to teach people how
not to be a slave of technology.
Our hope is that a digital civilisation
will finally be built, so that technologies
can be used for greater good. Governments,
markets, and societies need to take the issue
seriously, or the shadow of The Matrix
will prevail.
WWDC23, Part 4
June 30, 2023
To us, the most prevalent question about
OS is unanswered in WWDC23, of sideloading.
Apple has made arguments around privacy and
security to counter the sideloading regulation.
Another point is worth mentioning. Sideloading
prevents Apple from taking responsibility of
apps on user's iDevices. Malicious content
is user's own business. It's unfortunately
a step backwards.
Back to WWDC23, the first thing we wish
to talk about is content creation. Apple
seems to follow the practice of providing
users with the most capable tools for form,
rather than substance. Final Cut Pro and
Logic Pro are excellent editing tools, but
users need to bring substance.
Compared with Google, which used AI
to generate substance for user's prompts,
Apple's approach is less controversial,
but demands more creativity on user's side.
It's imaginable that Apple appeals to
creative pros, while Google aims at laymen.
Second, we wish to talk about personalization.
This is a area where Apple truly beats Google.
NameDrop, Journal, and Live Stickers help users
express themselves, rather than talk to a
personalized chatbot. It's not about the focus
on AI, as Journal and Live Stickers both
utilized it, but about uses of AI.
The Journal app is a example of the
mobile-cloud paradigm, which allows users
to take action whenever inspiration strikes.
On the other hand, Google's personalized
chatbot is likely a ramification rather
than a feature.
Third, we wish to talk about business uses.
As everybody knows, Apple's approach to
business use is quite artisanal, with
features here and there, but no coherent design.
It's a pity, as iPad proves
exceptional in handling business.
Google's generative AI approach to business,
however, is likely to generate some controversies
as well. We already know that ChatGPT makes up
pseudofacts, and thus without rigorous checking,
may result in disasters. The road to
useful AI has begun, but there is still
much to do before AI can serve us reasonably well.
Although there is much to talk about,
we feel it's better to leave it to other
blog posts, rather than try to squeeze
into the WWDC23 series. WWDC23 may seem
small without Vision Pro announcement,
but it's really not the case, as Apple
expressed its philosophy very clearly,
albeit with some deficiencies. So,
that's it for WWDC23.
WWDC23, Part 3
June 23, 2023
To appreciate OS features Apple announced
at WWDC23, a comparison between Apple and Google
is welcome. Apple and Google are not only
market competitors, but proponents of very
different philosophy.
Apple tends to help
users create, connect, and entertain themselves,
while Google uses its giant data system and AI to
filter information, generate content, and
form decisions.
As readers might have suspected, I used
both approaches. To some extent, one might
argue Apple or Google is better, but a more
satisfactory way is to complement each other.
Apple devices generally offer better
experiences, while Google cloud services
give more detailed information.
Philosophical differences lead
to different perspectives on AI. As Google
used AI as writing aid, Apple provided
Live Stickers for more animated communication.
Usually, Google has better data, while
Apple excels at graphics.
It's in this perspective we will provide
a view of OS announcements at WWDC23, and
compare them with Google I/O. Our view
will not give a comprehensive list of
features, but hopefully may spark some insights.
More to come.
Focus
June 21, 2023
Steve Jobs famously described that focus
is about saying no to numerous other good ideas.
It requires that one picks carefully what to
focus on. Often, people have no foresight
what to focus on, but do have several good ideas.
We feel sharing these out-of-focus ideas is
a good thing.
As a researcher, my purpose is to study
derived algebraic geometry
as a continuation of
complexity theory,
and to build a digital library of algorithms
at spectrum-dev.
From time to time, other ideas came to me, but it's
clear my limited time can not be allocated
to them. Sharing them is good, because others
might find their own inspired paths.
WWDC23, Part 2
June 11, 2023
The most important Mac announcement is
of course the complete transition to Apple
Silicon. Not so noticeable is the more finely
meshed Mac lineup. There are much finer gradations
for choosing a Mac than ever before,
and each product is of meaningful differentiation.
13-inch MacBook Pro and M1 iMac are probably
the weakest in the lineup. They still beat the
competition, but certainly may benefit from
reconsideration and upgrade. But, please resist
the temptation of touchscreen Macs.
15-inch MacBook Air is amazingly capable
at the price point. We expect many creative
workflows to utilize the increased screen size
and stay cost-effective.
Mac Studio and Mac Pro are built with
excellent efficiency. They might not be the
most powerful monster desktop computer,
but definitely the most enjoyable.
Apple didn't build specific gaming desktops,
but gamers can be hugely rewarded buying these.
Mac announcement at WWDC23 is quite brief.
There is not much marketing pitch involved.
It would be nice if Apple web site can provide
useful information for users in addition to
technology overview, like use cases.
So, this is for Macs. More to come.
WWDC23, Part 1
June 8, 2023
Apple Vision Pro is a dreamlike product.
It's great to see Apple put much care into
challenging headset
issues. Although
concerns remain, unpleasant headset characteristics
have been ameliorated. We don't have access
to the device, so impressions are the only
thing we can offer.
Readers are encouraged to read Daring Fireball's
account
first. The role of AI biometrics,
although Apple didn't mention
it, is obviously present in Vision Pro's UI.
The degree Apple mastered the technology is
astounding, as Gruber mentioned no latency
and no rendering instability along with
excellent graphics.
Although VR pioneered immersive experiences,
it's spatial computing that finally removed the
barrier between computer graphics and outside
world, in 3D. The technology harbors great
applications, like real time AI analysis
of objects, useful for museum navigation.
Apple didn't mention much, but the prospect
is clear.
For brevity, well-documented features
are omitted here, but we raise some additional
issues instead, so that one day the device
may mature. Hope this serves readers.
First, there needs to be a solution to
virtual keyboard. Physical keyboard allows
users to type without looking. Touchscreens
are typically small enough so that looking
at virtual keyboard and content simultaneously
is possible. With vast view offered by Vision Pro,
however, using virtual keyboard while looking
at distant content can be a real pain.
Second, the desire to replace Macs with
Vision Pro might have to be curbed. Macs
are based on precise cursor system for UI.
Imprecise eye-tracking and gestures are
no replacement, but a complementary feature.
Use Macs to build architecture, and Vision
Pro to evaluate.
Third, despite many criticisms of the
separate battery, it's necessary to keep
weight off user's head, but general ergonomics
definitely can be improved further. Can
it be a helmet? Can the battery be wearable?
These considerations are crucial if activities
that require agility, like gaming, demand
quick motion with Vision Pro intact.
Overall, Apple Vision Pro is a very
nice product with plausible use cases,
and some issues. We didn't dig into the
technology yet, and this is only Part 1
of WWDC23, so more to come if time permits.
Foundation for Mobile Gadgets
May 30, 2023
AI is taking over the news. Apple is
curiously missing. While AI features in
mobile gadgets have been long with us,
there is a view that they are simply
add-ons, rather than something essential.
Digital Revolution isn't easy, and Apple
should not be blamed for not seeing the
foundational role of AI in mobile gadgets,
but a little explanation helps.
As readers probably know, from the right
start, we actually derived mobile gadgets
from AI. The reason is that mobile gadgets
are convenient interfaces to information.
In a age of information deluge, it's inevitable
that computers are required to process information
for us, that is, AI.
People could imagine how shallow our
mobile gadgets would be without AI. Pretty
UI isn't replacement for lack of search,
maps, and biometrics. To Tony Fadell,
AI is the
future
of smartphones. To us, AI is at the heart
of mobile gadgets from the beginning.
Readers don't have to watch The Matrix
to feel how essential AI is.
We wish to say it again that Apple should
not be blamed for not seeing the essential
role of AI, but a little change in perspective
definitely will guide us through the dramatically
transformative period ahead.
Parallel Computing and Automation
May 27, 2023
From the perspective of time complexity
as n tends to infinity, parallel speedup
is only a constant factor. It's not trivial,
but pales before improvements over orders
of n.
However, in practice, since n is typically
bounded by a fixed number, the importance of
parallel computing begins to emerge. If medical
diagnosis can be processed 10 times faster, it
might mean life or death.
Take radiology for example. There are lots
of independent images for processing. Do it
sequentially is considerably slower than do it
parallel. Wouldn't it be nice if AI can
quickly run a analysis with multi-core GPU?
It's increasingly clear that parallel
computing is going to be with us for very long
time, but its benefits are not fully appreciated
and therefore not deeply utilized. Hope the
onset of AI will pave the way for fruition.
Taiwan Strategy in AI Race
May 20, 2023
After metaverse flop, Taiwan tried
to establish itself as a AI island.
Terry Gou placed sole emphasis on AI,
after previously described our age as
characterized by AI and 5G.
It's clear Taiwanese mostly don't
know what they are talking about.
More importantly, Taiwan's traditional
focus on hardware actually points toward
a future of AI backwardness, rather than
leadership.
Of course, Taiwan can make new arrangements
for AI, but I doubt the island's politicians
are smart enough for the task, or its people
wise enough to see the technical difficulties
ahead.
UI Programming
May 19, 2023
There was a time when UI programming
was best done with WYSIWYG. It's a
paradigm where the viewport is fixed.
Flash championed the era with versatile
graphics tools.
When multi-screen reality arrived,
the WYSIWYG paradigm faced enormous challenge.
It's impractical to design for each dimension
separately. Therefore, we see the rise of
programmable UI, specifying UI with code directly.
It's pretty certain that fixed viewport
is dead forever. Companies should invest
in programmable UI.
However, modern
app programming didn't adopt UI-logic
separation, which caused a lot of confusion.
There is a dated view that somehow UI
should still be done visually to avoid
excessive bugs.
We believe the right way forward is
to infuse clarity into UI code. How to
achieve this is a very open problem.
Is it declarative programming? Is it
a markup language? We don't know, but
programmable UI is the future.
Bookmark
May 17, 2023
A bookmark I added today points to a strange
URL. Is the computer hacked?
Doing Business on iPad
May 6, 2023
People have been skeptical of the iPad,
thinking it's a entertainment channel, rather
than a productivity tool. As a result, Apple's
efforts on iPad emphasized productivity tools.
Here we provide another view. Doing business
on iPad can be more important than productivity.
Traditionally, business communication involves
phones and computers. Phones are not as versatile
as computers, and computers are not as convenient
as phones.
As you might have guessed, iPad doesn't have
these drawbacks. Doing business on iPad is
incredibly comfortable. In fact, we used iPad
to recruit all our freelancers.
Receiving proposals, reviewing artworks,
instant messaging all work effortlessly
with or without Apple Pencil or keyboard.
It's a mystery why there are so few discussions
about iPad for business. Is it due to
feeling unprofessional?
Back to Office
May 1, 2023
Recently, we
talked
about music
in the workplace and related applications,
but didn't offer a use scenario.
It's useful to consider a small studio.
At present, there are lots of separated
devices, including hubs, telephones, music
players, etc. Lack of integration is a
problem. A missed office phone call doesn't
show up on iPhone. There is no cloud-synced
business contacts app. Meetings are
recorded without automatic cloud note-taking.
In a small studio, a single device
with touchscreen remote control can
solve all the problems. More, meetings
can be seamlessly scheduled on a unified
work calendar with all attached files
instantly available via AirPlay.
Cloud functionalities easily let participants
review previous meetings. Not to mention
no clutter of office space.
Security cameras, firewalls, and VPN
can all be suitably integrated. With simple
UI, there is no better way to access the
office remotely. It's up to serious business.
The list of features can go on and on.
While it's entertaining to explore them,
there is no replacement for a real product.
We will return to the topic if anything
relevant hits the market.
Deep Technical Design
April 29, 2023
There is a deep connection between design
and technology at Apple that no other company
can replicate. Design without technology
can not accomplish serious feats, while technology
without design is inhumane and waste.
A example is provided by MacBook Air.
Apple
tried
to design a thinner MacBook Air
with Intel CPU without much success. Then
Apple Silicon finally made it feasible.
A improved form accompanying M2 not only
instilled beauty, but also delivered
tremendous comfort.
Microsoft, lacking both design and
technology, is no match. Windows laptops
are bulkier and generate excessive heat.
The reason why Microsoft survived is the
dominance of Office and its slightly incompatible
Mac version. If Office for Mac worked seamlessly
with Windows counterparts,
there would be far fewer Windows installations.
AI, Empiricism, and Judicial System
April 27, 2023
Since AI digests big data, the application
to legal arsenal is of interest. There are
numerous success stories we won't talk about,
but instead focus on the nature of AI,
its relationship with empiricism, and
fundamental limitations of judicial
decision-making.
Starting with the simplest feature of AI,
it's a computer system analyzing data. Although
it can do sophisticated rational deduction through
computation, data is no replacement for evidence.
The measurement of electron spin isn't electron itself.
Lots of properties about the electron can not
be deduced from spin measurement alone, that is,
insufficient data.
The insufficiency of data imposes significant
limitation on AI. It might draw conclusion based
on past experiences and probability, but certainty
from direct evidence can not be reached via
computational means. Flaws are unavoidable.
Thus, while it might be desirable to ask AI
for the possibility of wrongdoing, a sound
judicial system should be built upon evidence,
rather than insufficient data.
The most valuable application of AI in
the judicial system, from our point of view,
is to transform a legal system vulnerable
to human sentiments into a logical system
so that errors may be corrected. Laws
should be written as a axiomatic system
with suitable logic.
To illustrate our point with a example,
we examine catch-22 laws. Such laws are
designed to frame people, no matter what
people do. It exploits inconsistency in
legal statements. AI can do extensive
deduction so that such laws may be identified
and fixed before any absurdity arises.
Better, a axiomatic legal system allows
individuals to use AI to evaluate legal
consequences, and to inform action. It's
unreasonable to train everyone as a lawyer,
or to consult lawyers for every action.
AI can help humans navigate the legal
landscape. Let's see if the law school is
smart enough to carry out these innovations.
Music in the Workplace
April 21, 2023
The digital music revolution started
by Apple is largely personal. As a result,
music in the workplace faces some issues.
Some choose radio. Some listen to YouTube. Those
who use Apple Music might set up a business
credit card on iDevices and give employees
complete control over them, a elevation
of access.
It's also desirable that a single
subscription works company-wide.
Complications like compatibility with
corporate broadcast system may require
new apps or device designs. All these
considerations can be integrated to
a general-purpose business device that
takes care of meetings, telephone,
networks, together with music.
Clairvoyance
April 17, 2023
The difference between a wizard and a
fortune-teller is clairvoyance. Steve Jobs
is a wizard. Taiwan politicians who bet
on the metaverse are fortune-tellers.
To be fair, a good politician doesn't
need to be a tech revolutionary, but then
he/she should not set tech agenda, either.
It's already clear that Taiwan politicians
are bozos.
UI: Simplicity and Certainty
April 12, 2023
Foolish media often overlook complexity.
They thought complexity is undesirable and
used Leonardo da Vinci, Einstein, and Steve Jobs
to indoctrinate the public. Fortunately,
with the
publication
of his letters
by the Steve Jobs Archive,
we know for certain that Steve Jobs understood
the
power
of complex technologies, but
wanted the user interface to be simple so that
common people may use it.
Simplicity isn't for simplicity's sake, but
to make powerful complex technologies accessible.
Sometimes, simplicity in UI can not be
achieved, but a more basic standard should
be preserved, that of certainty. To put it
simply, don't use small buttons for touch
screens.
Uncertain UI results in confusion and mistakes.
That's why Adobe use palettes to organize
functionalities in Photoshop, rather than
a bunch of windows. Good organization reduces
uncertainty in UI.
Certainty is why the Save As box in
Aqua
is much better than Windows. In Aqua,
the attached box tells clearly which window
is being saved, while in Windows, the
pop-up box leaves much for guess.
Simplicity is virtue, but certainty
is basic justice.
VR Headsets
April 8, 2023
There is a lot of speculation about VR
headsets, including
claims
that Apple might
be its last savior. While the outcome
is uncertain, pretty much everybody knows
Apple is half-hearted in VR development,
prioritizing AR. That's why we won't
emphasize Apple's attempt here.
It's useful to clarify VR's market.
Meta probably expected VR as a mass-market
product, but at present, its uses have been
largely for gaming. Like gaming PCs, it's
a niche.
For VR to break into mass market, more
features are needed. There is also a need
to elucidate VR's nature. For example, it's
very much like TVs, a stationary experience,
rather than iPhones, a mobile experience.
Content suitable for VR isn't hard to find,
but the weight of the headset must be so light
so that extended use won't cause harm.
To be honest, there are few indications that
major corporations are doing VR right.
The major UI feature isn't simply
point-and-click. It's that when a VR headset
is put on, the user is isolated from the outside
world. There are few gadgets like this, not
iPhones, not iPads, and not TVs. Actually,
I'm pretty hesitant to try VR just because of
the isolation.
Although the problems mentioned here
are obvious, I doubt any VR maker, including
Apple, really cares. Therefore, I'm somewhat
pessimistic about the success of near-term VR
headsets. Maybe some pros will take it seriously
in the future, but it seems reasonable not
to be too excited for now.
Apple Watch Dead
April 6, 2023
My Apple Watch quickly ran out of its battery
and refused to boot again today. It appears to
be dead. It seems the right time to reflect on
the lovely gadget.
From the beginning, Apple wisely emphasized
notifications and fitness features on the Apple
Watch. The main issue is the somewhat unsatisfactory
battery life that compromised daily usage.
I found my watch lasted a few hours per single charge,
down from a whole day.
Meditation is a great feature, too. But the
feature requires better execution. Since I
meditate during walks, the Meditation app never
worked for me. It should be enhanced with
better app interoperability, especially considering
people who record voice memos in the middle.
As always, Apple software is buggy, not limited
to watchOS. It's usable, but really unpleasant.
The link between Apple Watch and iPhone can be broken,
for example. A complete reset follows.
AI isn't a strong suit for Apple, but we suspect
it can help a lot. Scheduling reminders is one thing,
but getting live information can really help if
iPhone isn't immediately available, like wearing
gloves. Isn't it nice if Apple Watch can tell you
when a restaurant nearby opens?
Due to its nature, Apple Watch is much more
restrictive than the iPhone. Its development requires
deep understanding of human interaction. We
believe there is much more to accomplish,
but for brilliant craftsman.
Terry Gou's Presidential Bid, Part 1
April 5, 2023
Terry Gou outlined several points in his
core agenda for President. He wanted to resolve
conflicts between USA and PRC in order to keep
peace in the Taiwan Strait. He also mentioned
the importance of technology in a prosperous
economy. Overall, his proposal is very reasonalbe,
so we would like to help.
Policies of strategic importance, like
industrial policy for the semiconductor industry,
are indication that a politician has long-term
vision for the island, but welfare of common people
is what attracts voters. Here we talk about
the latter.
Housing is a place most people in Taiwan
want immediate improvement. The availability
of affordable housing is urgent. Currently,
Taiwan politicians tackle the issue with
market crackdown (DPP) and public houses for
rent (TPP). None of these approaches are
appealing. It seems that a policy similar
to Singapore housing program is called for.
There is also the problem of low wages.
DPP tried to solve the problem with rising
minimum wages, but that doesn't make economy
much better. The situation is very complex.
On the one hand, Taiwan industries should be
steered toward high value-added sectors.
On the other, labor bargaining power needs
to be enhanced to capture the added value for
commons. The easy path of financial liberalization
does more harm than good.
No politician can win based on the merit
of good policy agenda alone. But we feel Taiwan
politics is already kind of shit that a good
policy agenda may be very refreshing. If there
is opportunity, we will talk more. At present,
let's see how the events unfold.
Gadgets for Research
March 30, 2023
Apple gadgets have been instrumental in
learning and research. More and more students
are taking notes and doing homework on iPads.
It's also vastly more convenient to pack all
textbooks inside a tablet than printed books.
However, Apple's innovation in learning
and research seems quite stagnant, perhaps due
to sole focus on education, exemplified by
the Classroom app.
As everybody knows, learning/research
is much deeper and wider than education.
The iPad brings conveniences for sure,
but we believe the fundamental feature
should go beyond note-taking and digitization.
While reading philosophy on the iPad in a Cafe
is very enjoyable, that doesn't make
a philosopher.
Conventional people would suggest the use
of GenAI. It's a noble goal to make a
Vulcan out of AI, but we are nowhere near
it, and thus GenAI in its current form is
insufficient.
Here we offer a small improvement based
on mobile-cloud paradigm.
It can serve as a superb organization tool
for learning and research. Currently the
scattered web doesn't offer organization
comparable with the academia, so to learn
deep stuff, one still rely on human connections.
The obvious fact that there is no digital library
of algorithms, or mathematics, is indication
that the full potential of the mobile-cloud
paradigm hasn't been realized yet. One should
be able to get reasonably deep with the guidance
offered on the Internet, rather than relying
on his/her own connections. It's democratization
of wisdom.
While Generation Z love printed books, we
believe full-scale digital apparatus can benefit
humanity much more.
USA-PRC Technological Competition
March 12, 2023
Both USA and PRC realized the importance
of owning pieces of critical technology to
safeguard national autonomy. Computer industry
is the current focus of political establishment
on both sides. From semiconductor to AI,
governments seek to direct national resources
to aid development.
However, subsidies and monetary metric
are often not enough. As seen before, governments
could easily waste money on dubious ventures.
Since the matter is quite politically sensitive,
we refrain from providing details, but point to
the success of Apple chip design team as a clue.
It's crucial to study which conditions made Apple
a chip design powerhouse.
Technological leadership is hard. While
money can be printed on demand, technology can
not be developed in thin air. Both USA and PRC
dropped any pretenses of free market approach.
It remains to be seen which alternative is
successful.
Leonardo da Vinci's Discovery
February 20, 2023
As we suspected, gravitation was
discovered
by Leonardo da Vinci. The only thing preventing
him from a sound theory was mathematics. The
experimental data permit
unique solution (PDF)
through a differential equation. History of
modern physics is rewritten.
USA-PRC Competition
February 10, 2023
Imperial economists see USA-PRC compeition
as a new Cold War. It's certainly a possibility.
However, it could also be like Apple versus Google,
where everybody has something to gain, rather than
a Cold War.
Hope wisdom will prevail.
Limitations of Knowledge Engines
February 8, 2023
We
talked
about some limitations of knowledge
engines before. There, a simplified model of knowledge
engines is used, that of a interface to knowledge database.
Here we provide a generalized version, applicable
to any reasonable knowledge engine.
Let's begin with a example. Recently, ChatGPT is
very fashionable. We can find all sorts of information
with Google Search. However, at present, the prime example of
knowledge engine, Wolfram|Alpha, still returns absurd
results for the term chatgpt. It's a town, not a
piece of AI.
The reason why Wolfram|Alpha is so wrong can be
explained by our previous blog post, but more generally,
the deep reason is that knowledge engines exclude
things outside of current knowledge, like uncertain future.
Uncertain future isn't current knowledge, but can be
terribly important.
Therefore, we need information systems that help
us explore possibilities, in addition to knowledge.
We believe knowledge engines are useful, but not that
useful. It needs to be extended to incorporate
useful information beyond current knowledge.
Teaching Calculus
February 4, 2023
Fashionable textbooks teach calculus
with scientific applications. While the
appeal to problem-solving is valuable,
we find that students often can not
grasp the logical structure.
A alternative is to use logical/axiomatic
approach to teach mathematical deduction,
and separate mathematics from scientific
applications. The advantage is conceptual
clarity.
A example we used to demonstrate
the power of calculus can serve as a
test of comprehension. It's proving
that the exponential function is a
transcendental function.
A simple proof requires mathematical
induction, differentiation, and division
by the exponential function. The beauty
of the deduction is that a algebraic
property can be proven analytically
with basic calculus.
Perhaps the logical/axiomatic approach
will allow students to appreciate the
deduction. The fashionable textbook
approach doesn't quite fit. The example
might tell the difference in understanding.
Tim Cook's Real Test
February 3, 2023
Unsurprisingly, USA government is planning
to regulate Apple. So far, Apple's response
has been advertisement and lobbying. The real
test for Tim Cook is for Apple to provide better
alternatives to regulatory retaliation and win.
Most people bet on the USA government, so it's
a real test for Tim Cook.
Digital Automation
January 26, 2023
In common people's experience, digital
automation is not very impressive. The
applications are cleaning houses, printing
documents, completing financial transactions,
etc. While the advancement of AI adds
self-driving cars, article writing, and
facial authentication to the portfolio,
the general public still treat it as a
novelty, rather than something critical.
Yet, facing a aging population, digital
automation may be the best hope we have
to maintain living standards without compromising
on retirement age and welfare.
It's well-known that the elderly pose
fiscal stress for the government. Social Security,
healthcare, and other benefits are under scrutiny
as the elderly population grows. While it's
barbaric to slash spending and cut benefits,
something has to give if the working population
ratio shrinks and productivity remains the same.
Unimaginative governments indeed slash spending
and cut benefits. It's stupid.
A much better way out is to invest in technologies
that increase productivity, so that living standards
can be maintained while the working population ratio
shrinks. Digital automation is a excellent candidate
for such industrial transformation. Although it's
not as shiny as Apple, the adoption is critical
if we want to keep human decency.
Policy makers need to carefully evaluate
the economics of such transformation. Will public-private
partnerships work better? Will globalisation help?
Will new institutions be created to facilitate
digital automation? History will judge whether
we care about the elderly.
Theory and Practice
January 22, 2023
Physics is fortunate to have a number of
really smart theorists that offer insights
for experiments. It's not exaggeration that
the 20th century is a century of physics.
Computer industry isn't that fortunate.
Although there are brilliant theorists like Turing
and Shannon, computer scientists made no efforts
to piece everything together. It's no surprise
that the iPhone, and mobile gadgets more generally,
came as a result of inspiration
and experimentation, rather than a theoretical
prediction. We, at Glacier Studio, speculated
about the form and function of mobile gadgets
a few years before the launch of the iPhone,
and only recently came up with a theory
unifying mobile gadgets, cloud, and AI.
Wouldn't it be nice if the iPhone can
be placed within a sound theory so that its
future and significance can be properly handled?
For example, many people believe that the iPhone
phenomenon is about to end, while we believe that
mobile gadgets constitute eternal innovations,
just like emails. A sound theory would explain
why smartphones aren't going away.
So far, our judgement is based on the
observation that there are no better ideas
than mobile gadgets, cloud, and AI working together.
We believe it's sufficient, but a theory would
make it much more satisfactory, so that Jack
Ma's thesis that implants will replace smartphones
can be put to rest.
Just like physics, the computer industry
can be much better formed with similar
theory-experiment methodology. All it takes
is for brilliant people to join the conversation.
Complexity beyond Computation
January 14, 2023
Several months ago, we wrote a article
about complexity beyond computation. Here
is the
PDF.
Apple Business Connect
January 13, 2023
Apple
launched
Apple Business Connect to facilitate
business presence
across apps, especially Maps. Taking a step
further than Google, Apple allows more direct
and informative interactions between businesses
and their customers.
As a long-time Google Maps user, it always
feels a bit annoying that restaurants can not
post timely updates to inform customers about
open hour changes. All we can see was a message
indicating that holidays may affect business hours.
It's second to useless. Hopefully, now with Apple
Maps, the scenario can be gone.
Apple Business Connect is not just a Maps
feature. We think it's a philosophical objection
against the metaverse. Metaverse is not defined
by virtual reality, but largely a business platform
for virtual goods. While games are attractive,
virtual real estate borders on frauds. Apple
Business Connect, on the other hand, serves as
a virtual interface for largely real goods and
services. Our guess is that it will perform better
than the metaverse.
People like doing real business virtually, but
not virtual reality business. The reason is simple.
Virtual reality goods are only a small fraction of the
GDP. Perhaps Meta should reconsider its own economics.
From business perspective, Apple Business Connect
might one day become as important as Google Ads.
More and more people find goods and services on Maps,
rather than search, especially for restaurants. If Maps
can go far beyond Yellow Pages, it will be a
welcome transformation.
Surveillance and Manipulation
January 6, 2023
Apple CEO Tim Cook expressed
concerns
over corporate and government surveillance activities.
He worried that the feeling of being watched may
drastically change individual behavior and therefore
harm the society in a profound way. While this view
is certainly true, it underestimates the damages of
unscrupulous surveillance that enable further
manipulation.
The situation is not new. PRC government
long used surveillance
and manipulation against its people. What's new is
that USA seems to agree with the practice without
significant resistance, which is monstrous.
If corporations or governments can arbitrarily
block communication based on surveillance, the result
is not merely changed behavior, but the onset of 1984.
Individual's fate is no longer in his own hands, but
the authority determines it. Freedom is no more.
Dystopia ensues.
A major question is whether the Silicon Valley
is sane enough to stop it.
Art and World Exploration
January 5, 2023
World exploration has been dominated by science.
Quantum mechanics explained why atomic spectra can
be discrete. General relativity explained why gravity
has influence over time measurement. But in a sense,
no satisfactory science has been found to explain the
whole complexity of the world. The value of art is
to capture the world when science fails. It can be
our mysterious existence, or daring imagination.
The onset of modern gadgets provided unique
opportunities for artists and common people to
greatly expand their spiritual dimension. Random
inspirations can be instantly recorded on mobile
gadgets and preserved in the cloud. The organization
is superb. Sharing is convenient. There is no
better time to be a artist.
Yet, paradoxically, people often don't take
advantage of mobile gadgets to enrich their spiritual
life, but engage in banalities such as political
conspiracy theories and extremism. Perhaps the
search for belonging has lost its innocence.
Can humanity be saved?
Cinema is a great example. Instead of popularizing
Bresson and Tarkovsky, content distribution platforms
systematically cultivated rubbish. Perhaps the
cinema industry is craving for novelty, but in
many ways, Bresson and Tarkovsky are much more innovative
than Marvel, just like the Beatles. It's a tragedy.
The cynical manipulation of people's life
has ruined many. From media distortion to outright
censorship, art and world exploration are actually
in decline in a age of unprecedented opportunities for
spirituality. Evil rules.
Life is short. Modern cynicism intends to
fill it with junk. Redemption is one of the most
important questions of our time.
Mobile-Cloud Paradigm
December 30, 2022
In the Digital Revolution blog, we explained
how
the presence of cloud computing changed the mobile
landscape. The sync functionality not only provides
conviniences for users, but also results in
market transformations that favor cross-platform
synchronized apps. Here we explore some of the
features and ramifications of the mobile-cloud paradigm.
Maps: Cloud side of the Maps app is a marvellous
aggregation of map data. Naive people may think they
are merely a list of locations, but the reality is that
to keep Maps globally consistent and locally efficient
is a technical achievement. Mobile not only gives
users anytime anywhere access to Maps, but also
makes feedback and marketing extremely useful. Just
see how many discounts and gifts users may receive
in a restaurant with a rating improvement campaign.
Social Networks: Cloud side of social networks,
like that of Maps, aggregates information and use it
to provide exploration, discovery, and connection.
While forms are many, the online community really
keeps families and friends closely together. Mobile
allows users to interface between the outside world
and the online identity. Inspiration is everywhere,
and easily shared.
Filter Bubble: This is a significant
ramification
of the mobile-cloud paradigm, conducted with forced
personalization. It introduced localization of global
information. Without access to global structure,
localization limits the scope of search and various
other activities. Not only so, the mobile
surveillance
capability further endangers privacy. People's
online activities become profiles that may be exploited.
The power of mobile-cloud paradigm is immense,
as well as
challenges.
Although the academia concentrates its efforts on AI,
we believe that the mobile-cloud paradigm is a comparably
important research area that may unite AI with interfaces
and infrastructure. Time will tell.
The Glacier Studio Blog Launch
December 25, 2022
Glacier Studio had two freezed blogs,
On the World
and
Digital Revolution.
These blogs are highly specialized in their content.
As time goes on, we feel necessary to have a blog about
general, as well as specialized, topics. Thus, this blog
is established. Merry Christmas!