sixfiveotwo

> I think a memory address is a number that CPU considers to be a memory address

I meant to say that, indeed, there must be some concept of CPU for a memory address to have a meaning, and for this concept of CPU to be as widely applicable as possible, surely defining it as abstract as possible is the way to go. Ergo, the idea of a C abstract machine.

Anyway, other people in this thread are discussing the matter more accurately and in more details than I could hope to do, so I'll leave it like that.

How would you define what a memory address is without first defining in which context it has a meaning?

I'm sorry if you feel that way.

I am in no way trying to judge you; rather, I'm trying to get closer to the truth in that matter, and your input is valuable, as it points out a discrepancy wrt TFA, but it is also subject to caution, since it reports the results of only one chess player (right?). Furthermore, both in the case of TFA and this youtuber, we don't have full access to their whole experiments, so we can't reproduce the results, nor can we try to understand why there is a difference.

I might very well be mistaken though, and I am open to criticisms and corrections, of course.

> Yeah like several hundred "Chess IM/GMs react to ChatGPT playing chess" videos on youtube.

If I were to take that sentence literally, I would ask for at least 199 other examples, but I imagine that it was just a figure of speech. Nevertheless, if that's only one player complaining (even several times), can we really conclude that ChatGPT cannot play? Is that enough evidence, or is there something else at work?

I suppose indeed one could, if one expected an LLM to be ready to play out of the box, and that would be a fair criticism.

Very strange, I cannot spot any specifically saying that ChatGPT cheated or played an illegal move. Can you help?

That's simple logic. Quoting you again:

> Machines are good at applying rules, so when they fail to apply rules correctly, it means they have incorrect, incomplete, or totally absent models.

If this line of reasoning applies to machines, but LLMs aren't machines, how can you derive any of these claims?

"A implies B" may be right, but you must first demonstrate A before reaching conclusion B..

> I think we are discussing whether LLMs can emulate chess playing machines

That is incorrect. We're discussing whether LLMs can play chess. Unless you think that human players also emulate chess playing machines?

I think the article briefly touch on that topic at some point:

> For one, gpt-3.5-turbo-instruct rarely suggests illegal moves, even in the late game. This requires “understanding” chess. If this doesn’t convince you, I encourage you to write a program that can take strings like 1. e4 d5 2. exd5 Qxd5 3. Nc3 and then say if the last move was legal.

However, I can't say if LLMs fall in the "statistical AI" category.

> Whereas the LLM makes "moves" that clearly indicate no ability to play chess: moving pieces to squares well outside their legal moveset, moving pieces that aren't on the board, etc.

Do you have any evidence of that? TFA doesn't talk about the nature of these errors.

> Machines are good at applying rules, so when they fail to apply rules correctly, it means they have incorrect, incomplete, or totally absent models.

That's assuming that, somehow, a LLM is a machine. Why would you think that?

It's a good point.

But this math analogy is not quite appropriate: there's abstract math and arithmetic. A good math practitioner (LLM or human) can be bad at arithmetic, yet good at abstract reasoning. The later doesn't (necessarily) requires the former.

In chess, I don't think that you can build a good strategy if it relies on illegal moves, because tactics and strategies are tied.

thank you, that looks awesome.

Sorry, I understand it was a bit intrusively direct. To bring some context, I toyed a little with neural networks a few years ago and wondered myself about this topic of training a so called quantized network (I wanted to write a small multilayer perceptron based library parameterized by the coefficient type - floating point or integer of different precision), but didn't implement it. Since you mentioned your own work in that area, it picked my interest, but I don't want to waste your time unnecessarily.

> I spend much longer trying to figure out how to get 1-trit training to work and I never could.

What did you try? What were the research directions at the time?

Indeed, you can get a lot more from dependent types than Damas-Hindley-Milner inference, yet does it mean that you should use the former everywhere?