grondilu

It doesn't seem so at all from what I've tried.

> so even theoretically this is hard to imagine.

Possibly a strech, but transistors are basically current amplifiers, so their optical equivalent should be... lasers. Indeed lasers are optical amplifiers. Whether or not they can be turned into logic gates as transistors can, I don't know.

> Mechanical computers are computers that operate using mechanical components rather than electronic ones.

For anyone who's excited about mechanical computers, perhaps it is worth reminding that an electron is about a thousand times lighter than a nucleon. Therefore, it's probably fair to say that mechanical computers will always be more energy consuming than electronic ones, because they fundamentally need to move atoms around to operate.

They took 20 bitcoins out of me, but I never bothered to fill the paperwork. Oh, well.

I'm quite baffled by the fact that LLMs can generate a dataset used to train other LLMs. One would think that such a feedback loop would produce utter nonsense but apparently not. This seems to work.

> But the first problem is, the number of legal transformations is actually infinite.

I am fairly certain the number of legal transformations in mathematics is not infinite. There is a finite number of axioms, and all proven statements are derived from axioms through a finite number of steps.

> Nor is it music without a musician.

The case of music is quite fascinating. YouTuber Rick Beato is following it very closely. He seems to think that quite soon people will listen to AI generated music even knowing it is AI-generated music. They will listen to it because they'll enjoy it and they will not care who or what made it. I personally think he is right. Music can be enjoyed in a way no other art can, in that sense it might become kind of a drug.

Now some people think that will never happen because AI music will always be bad, but recently he posted a video in which he tells us that his kids can immediately tell an AI-generated song from a normal one. He is baffled because he himself can't do that. But he also tells us that one of his kids thinks that he probably won't be able to tell it anymore in six months or something. In other words, a young person with such a good ear is confident that AI-generated music will soon be if not good, at least indistinguishable from human-made music. I think that's indicative of where things are going.

https://www.youtube.com/watch?v=zbo6SdyWGns

"I think in the future, instead of typing up our proofs, we would explain them to some GPT. And the GPT will try to formalize it in Lean as you go along. If everything checks out, the GPT will [essentially] say, “Here’s your paper in LaTeX; here’s your Lean proof. If you like, I can press this button and submit it to a journal for you.” It could be a wonderful assistant in the future."

That'd be nice, but eventually what will happen is that the human will submit a mathematical conjecture, the computer will internally translate into something like Lean, and try to find either a proof or a disproof. It will then translate it in natural language and tell it to the user.

Unless mathematics is fundamentally more complicated than chess or go, I don't see why that could not happen.

Growing Neural Cellular Automata https://news.ycombinator.com/item?id=22300376, February 2020

> improved soil health, improved ecosystem health, better water retention, less erosion, more carbon sequestered in the soil.

Regarding carbon sequestration, I think it is worth pointing out that Freeman Dyson, in one of his conferences, mentioned no-till farming as one land management methods that could be used to absorb the carbon emitted in the atmosphere by human activities.

"The point of this calculation is the very favorable rate of exchange between carbon in the atmosphere and carbon in the soil. To stop the carbon in the atmosphere from increasing, we only need to grow the biomass in the soil by a hundredth of an inch per year. Good topsoil contains about ten percent biomass, [Schlesinger, 1977], so a hundredth of an inch of biomass growth means about a tenth of an inch of topsoil. Changes in farming practices such as no-till farming, avoiding the use of the plow, cause biomass to grow at least as fast as this. If we plant crops without plowing the soil, more of the biomass goes into roots which stay in the soil, and less returns to the atmosphere. If we use genetic engineering to put more biomass into roots, we can probably achieve much more rapid growth of topsoil. I conclude from this calculation that the problem of carbon dioxide in the atmosphere is a problem of land management, not a problem of meteorology. No computer model of atmosphere and ocean can hope to predict the way we shall manage our land."

https://www.edge.org/conversation/freeman_dyson-heretical-th...

For what it's worth that sounds a lot like what Max Tegmark classifies as the "level 1" multiverse.

https://space.mit.edu/home/tegmark/crazy.html

Friendly reminder that "Paris syndrome" is a thing:

https://en.wikipedia.org/wiki/Paris_syndrome

Well, isn't it what the ISS is?

Please elaborate. Are you thinking of approximating the distance function with an algebraic expression, with algebra itself being the "programming language"?

The application to graphics is interesting. It seems to me that current image generation models struggle with stylized pictures ("ligne claire" in comics, geometric shapes and so on). After all this kinds of pictures should be easy to encode in vectoriel formats (like SVG), which are basically programming languages.

No, I didn't take any electrolyte, as it seems artificial to me.

This is only anecdotal but I had heard about the benefits of fasting for brain performance, so I once tried fasting during a chess tournament. I fasted during the first six days of the tournament. My results were disappointing so I gave up and broke the fast on the seventh day.

> OTOH I think there are theoretical lower bounds on the amount of energy that needs to be ejected as heat from a non-reversible computation, such that a non-reversible SC would still need to produce some heat.

Yes, it's called Landauer's principle, but it's so low it can be ignored for most intent and purposes.

My experience with learning chess openings is comparable : repetition is necessary but not sufficient. You need to "work" on each line, thinking deeply about each one and trying to find associations either between them or with things you already know.

This is what this post (or rather just the title, tbh) immediately reminded me of.

I remember learning about it in the early 2000s. It was considered a very challenging problem, with very important applications, most notably speech recognition in natural settings.

I wonder what is the current status on this. Is this considered solved nowadays?

Nice. Maybe a pie in the sky but considering how prevalent python is in machine learning, how likely is it that at some point most ML frameworks written in python become executable on a web browser through WASM?

I think it's due to the library used : chessboard.js[1] I used the version used in the archive, 0.3.0. It might be old.

1. https://chessboardjs.com/

It's not always going to remain limited to 5MB, which is probably much enough for an opening repertoire anyway.

Also it shouldn't be too hard to create an export/import button.

I'd argue it's more reliable than having your progress stored on a server that can close anytime, as memchess did.

The short answer is I tried listudy and I didn't like it.

I liked memchess because of how simple it was, and I liked not having to enter my own lines (though I understand this is important for serious study).

Memchess is what helped me defeat the strongest player in my local club, so I was quite frustrated when it shut down.

How do you add a GH page to a repo?

memchess.com was a neat website to learn chess openings with the spaced repetition method.

It was closed around 2020 for some reason.

I used code gathered from archive.org[1] and built a version that seems to work. It does not require a subscription/login, instead it stores progress through the HTML5 web storage API[2].

1. https://web.archive.org/web/20240000000000*/memchess.com

2. https://www.w3schools.com/html/html5_webstorage.asp