That’s basically what AI is best at: incorporating lots and lots of data points (every single pixel on an image, many trillions of pixels on previous images used to train the model) to determine the best answer to a question with clear definitions of what is better or worse on a single-dimensional quality axis that most people would agree on. In this case, whether a de-noised photo looks good and accurately represents the image that could have been captured of the exact same scene using more light or a better sensor or whatever.

It’s generative AI where things fall apart. The analogy I always use is that these math techniques are great for inferring what is between pixels in a low res image, but starts going off the rails when it starts guessing what’s beyond the edges of the image. These AI algorithms are just math, and math with this many data points needs to start considering the limitations described by chaos theory (in the mathematical sense).

Once there’s a public expectation for profit from the two biggest AI labs

I don’t think that it’s necessarily the order things will go.

Anthropic and OpenAI have bills to pay, including obligations to pay a ton of money to partners/vendors in the coming years. They need cash to do that. And the cash can come from 3 main places: revenue from customers, loans from lenders (usually in the form of issuing bonds), and issuing new stock/equity to raise funds.

The IPOs are basically their last chance to raise funds by issuing new equity. Even if each of them raise $100 billion, that’s probably not enough to hit self sustaining profitability. As lenders start to get more skeptical, in a macroeconomic environment with high interest rates and high energy prices, they start to run into a very real possibility that they hit illiquidity issues, even if their core business model actually works by then (and I don’t think it will). There’s just not enough runway for them, even in the most optimistic long term projections.

I want a convertible, a laptop that folds all the way back and gives a tablet/stylus interface, but has stiff enough hinges to hold the screen like a laptop at any angle with a keyboard, and that closes and covers the screen like a laptop. They played around with some form factors like that 10 years ago but they never really took off. Plus they tended to be Windows only, with poor Linux support on the hardware.

Whatever the reason, it’s basically become true.

Annualized Revenue rate is a made up number

It is fundamentally a projection about the future using recent results, and transparent in its methodology. It’s shit methodology, but it’s not hiding anything. So I don’t know that I’d classify it as fraud.

Playing games with “revenue” is another problem, but because that’s misleading about the past, and making an active decision on whether to classify money that was never gonna end up in their pockets as revenue share to partners or cost of revenue. Companies sometimes engage in shenanigans where they give discounts but classify the discounts as marketing/sales expenses rather than foregone revenue. That kind of stuff strikes me as more squarely in the category of accounting shenanigans.

Financial Times smells blood in the water. They’ve been reporting on leaked OpenAI documents for a bit the last few weeks.

If the major financial outlets start reporting on the negative, there may be a sentiment shift that does actually destroy the companies propped up by hype rather than fundamentals.

The number they used for annualized revenue for these rapidly growing companies is taking the most recent month’s revenue and multiplying it by 12. It’s not a guarantee about the future, but it is misleading.

The extreme version is taking a good week’s revenue and multiplying it by 52, or even worse, multiplying a day by 365. Those are outright dishonest. But not accounting fraud, because it’s not accounting at all.

The tablet one eventually came true. Desktops and laptops aren’t anywhere near as popular for personal use by non-technical people.

In the U.S., copyright protections don’t extend to functionality, though. So copying even copyrighted code in a manner to replicate functionality is fair use when there’s no other way to accomplish the same thing.

For example, if you have a piece of equipment that checks the software on a cartridge for a string of text that says “Produced by or under license from Sega Enterprises Ltd.” before running that software, then it’s fair use to copy that exact text so that your software can run on that equipment. And it’s fair use to reverse engineer and decompile licensed cartridges to see what the bare minimum necessary to make it work.

One way to prove that you didn’t copy the software any more than is strictly necessary for functionality is to fully document the functionality, and then have a skilled programmer take the documentation and write new software from scratch, without ever having seen the original software whose function is being copied. That’s a “cleanroom implementation.” Compaq and other IBM clones built their own BIOS software to implement the exact same functionality of copyrighted IBM code, and created an entire industry of IBM compatible PCs that weren’t actually licensed from IBM. Similarly, Google moved Android off of Sun-licensed Java using a cleanroom implementation (when Oracle bought Sun and Google wanted to get away from Larry Ellison’s abusive pricing practices).

Ok, so if it’s permissible to reverse engineer the code to create documentation of how it works, and then have someone else take that documentation and implement the functionality using new code, how do AI/LLMs fit into this? Can it be said that it’s truly a “cleanroom” when the reverse engineering and decompilation functions are done by the same software that converts the decompiled code into documentation in human language, and then is the same software that converts the documentation into newly implemented code? Doesn’t quite hit the same way, and I’m not sure the courts would see it the same way.

All of this is a gray area, and people shouldn’t confidently predict what the courts will decide in specific nuanced examples. A lot will depend on the specific details, so there isn’t going to be much room for sweeping generalizations.

Conway’s Game of Life is Turing Complete and that fact has been fucking with my mind for like 25 years.

Yeah, nothing quite like visiting a channel for the first time in a few weeks, while your client spends minutes trying to load the data to actually catch up with what you missed.

It is believed that pi is “absolutely normal” in the strict mathematical definition: that in any base numbering system, any finite sequence of digits appears as frequently as any other sequence of that length. If that’s true, then yes, a particular binary string, including one that corresponds to a particular digital file (like a particular mp4 file), can be found in pi.

That said, it hasn’t actually been proven that pi is normal. If it’s not normal, then the number can be infinitely non-repeating and still never hit that particular sequence of digits.

Hey some websites took the effort to progressively encode the images so that they’d slowly blur into place. Of course, you’d never fully know whether an image was worth loading until the very end.

Yeah, you’re probably right. Still, if big tech’s ad behavior (ads for stuff you’ve already bought, ads for things you’d never buy but might have looked up information about for non-consumer reasons) is any indication, I’m not actually sure they’re using really sophisticated techniques here that can’t be thrown off by one-off bored sessions.

When I’m bored, I get price quotes from Lyft and Uber to the airport. I’ll click around and look at the different tiers of service. Then I choose not to actually request a fare, and close the app.

I fly enough to where both apps probably have me as a regular airport customer, but that I’m price sensitive enough where I just won’t use their service sometimes.

I do the same with Doordash and other food apps, where I just put together potential orders to get a sense of how much something would cost if I ordered it, and then just never order it, letting that cart expire. I’m hoping this poisons their data or adds something to the price sensitivity metrics.

I also look at airfares, hotel prices, etc., a lot. It’s not for the purpose of manipulating their algorithms, as I just like to get a sense for how prices move over time. But if there’s a side effect that these variable pricing algorithms take into account people who window shop and walk away, maybe I’m helping push prices down.

If they’re gonna create a panopticon, we should fuck with their data.

AMP was two different things.

1. Google defined a standard that was an approved subset of html/css/js designed to load quickly on mobile screens, with none of the extra cruft that slows down sites as experienced by phone users.
2. Google also agreed to host AMP-compliant pages on their own ultra-fast servers. Many smaller sites, especially local news stations and local newspapers, agreed to offload the extra work by allowing Google to host their content. Those hosted pages had a whole system of referring back to the canonical URL on the actual outlet’s own website (and the desktop version of the page), and had little things referring to the AMP project.

I was 100% on board with #1, but #2 was controversial and created a lot of the conflicts of interest that tore the whole thing apart. Google spun off the AMP project as an open governance model, and then it was taken over by the OpenJS foundation.

In the end, the performance gap between mobile devices and desktops narrowed, and at this point most people’s phones and mobile browsers can load pages just as quickly as a laptop, so the purpose of the project itself is less important to the actual web as it exists today.

No paychecks mean no customers.

That’s never been a concern with the corporate executives. The macro level stuff is someone else’s problem.

We saw it in the 90’s when they stopped hiring managers from within, instead relying on bringing in outside consultants, and then directly hiring the ones they liked, so that there was no longer a path between the typical employee’s seat and their boss’s seat.

We saw it in the 2000’s when companies realized that they didn’t want to train entry level employees anymore, and would rather wait for someone else to train them, and then hire midlevels. Except too many employers did that and then the pipeline of future midlevels dried up in a lot of industries that were then ill positioned to survive the tumult of the 2008-2010 recession. Then something similar played out in the 2010s in certain other industries.

It’s always been buck passing. They don’t care.

He doesn’t even have that. He just pretends, by misleadingly saying he has a degree from Wharton. He does have a business degree from undergrad, from the University of Pennsylvania, whose business school is called Wharton, but most people say Wharton degree to refer to an MBA from that business school. He takes advantage of the confusion by referring back to his Wharton degree, and it sometimes works.

Google’s AMP was supposed to stop this, but they never could get around the conflict of interest between “our search engine experience is improved by prioritizing fast loading results” and “our ad business needs websites to carry a ton of ads,” ended up carving in enough exceptions and an anti-AMP backlash that they pretty much walked away from the whole thing.

I’m also waiting to see how Linux support for the A-series chips develops, now that the MacBook Neo has A18 Pros running MacOS. There are people working on tools to allow Asahi’s m1n1 bootloader to work on the Neos. Unclear how much of that effort will ultimately be upstreamed to Asahi itself, but I’m hopeful.