4 ms·

This is a fascinating idea - although I wish the definition of search in the LLM context was expanded a bit more. What kind of search capability strapped onto c

by timfsu 2y ago

This is a fascinating idea - although I wish the definition of search in the LLM context was expanded a bit more. What kind of search capability strapped onto current-gen LLMs would give them superpowers?

gwd 2y ago
I think what may be confusing is that the author is using "search" here in the AI sense, not in the Google sense: that is, having an internal simulator of possible actions and possible reactions, like Stockfish's chess move search (if I do A, it could do B C or D; if it does B, I can do E F or G, etc). So think about the restrictions current LLMs have: * They can't sit and think about an answer; they can "think out loud", but they have to start talking, and they can't go back and say, "No wait, that's wrong, let's start again." * If they're composing something, they can't really go back and revise what they've written * Sometimes they can look up reference material, but they can't actually sit and digest it; they're expected to skim it and then give an answer. How would you perform under those circumstances? If someone were to just come and ask you any question under the sun, and you had to just start talking, without taking any time to think about your answer, and without being able to say "OK wait, let me go back"? I don't know about you, but there's no way I would be able to perform anywhere close to what ChatGPT 4 is able to do. People complain that ChatGPT 4 is a "bullshitter", but given its constraints that's all you or I would be in the same situation -- but it's already way, way better than I could ever be. Given its limitations, ChatGPT is phenomenal. So now imagine what it could do if it were given time to just "sit and think"? To make a plan, to explore the possible solution space the same way that Stockfish does? To take notes and revise and research and come back and think some more, before having to actually answer? Reading this is honestly the first time in a while I've believed that some sort of "AI foom" might be possible.
- cbsmith 2y ago
  > They can't sit and think about an answer; they can "think out loud", but they have to start talking, and they can't go back and say, "No wait, that's wrong, let's start again." I mean, technically, they could say that.
  - refulgentis 2y ago
    Llama 3 does, it's a funny design now, if you also throw in training to encourage CoT. Maybe more correct but verbosity can be grating CoT answer Wait! No, that's not right: CoT...
- fspeech 2y ago
  "How would you perform under those circumstances?" My son would recommend Improv classes. "Given its limitations, ChatGPT is phenomenal." But this doesn't translate since it learned everything from data and there is no data on "sit and think".
cgearhart 2y ago
[1] applied AlphaZero style search with LLMs to achieve performance comparable to GPT-4 Turbo with a llama3-8B base model. However, what's missing entirely from the paper (and the subject article in this thread) is that tree search is massively computationally expensive. It works well when the value function enables cutting out large portions of the search space, but the fact that the LLM version was limited to only 8 rollouts (I think it was 800 for AlphaZero) implies to me that the added complexity is not yet optimized or favorable for LLMs. [1] https://arxiv.org/abs/2406.07394 https://arxiv.org/abs/2406.07394

by timfsu 2y ago

gwd 2y ago

I think what may be confusing is that the author is using "search" here in the AI sense, not in the Google sense: that is, having an internal simulator of possible actions and possible reactions, like Stockfish's chess move search (if I do A, it could do B C or D; if it does B, I can do E F or G, etc). So think about the restrictions current LLMs have: * They can't sit and think about an answer; they can "think out loud", but they have to start talking, and they can't go back and say, "No wait, that's wrong, let's start again." * If they're composing something, they can't really go back and revise what they've written * Sometimes they can look up reference material, but they can't actually sit and digest it; they're expected to skim it and then give an answer. How would you perform under those circumstances? If someone were to just come and ask you any question under the sun, and you had to just start talking, without taking any time to think about your answer, and without being able to say "OK wait, let me go back"? I don't know about you, but there's no way I would be able to perform anywhere close to what ChatGPT 4 is able to do. People complain that ChatGPT 4 is a "bullshitter", but given its constraints that's all you or I would be in the same situation -- but it's already way, way better than I could ever be. Given its limitations, ChatGPT is phenomenal. So now imagine what it could do if it were given time to just "sit and think"? To make a plan, to explore the possible solution space the same way that Stockfish does? To take notes and revise and research and come back and think some more, before having to actually answer? Reading this is honestly the first time in a while I've believed that some sort of "AI foom" might be possible.

cbsmith 2y ago
> They can't sit and think about an answer; they can "think out loud", but they have to start talking, and they can't go back and say, "No wait, that's wrong, let's start again." I mean, technically, they could say that.
- refulgentis 2y ago
  Llama 3 does, it's a funny design now, if you also throw in training to encourage CoT. Maybe more correct but verbosity can be grating CoT answer Wait! No, that's not right: CoT...
fspeech 2y ago
"How would you perform under those circumstances?" My son would recommend Improv classes. "Given its limitations, ChatGPT is phenomenal." But this doesn't translate since it learned everything from data and there is no data on "sit and think".

cgearhart 2y ago

[1] applied AlphaZero style search with LLMs to achieve performance comparable to GPT-4 Turbo with a llama3-8B base model. However, what's missing entirely from the paper (and the subject article in this thread) is that tree search is massively computationally expensive. It works well when the value function enables cutting out large portions of the search space, but the fact that the LLM version was limited to only 8 rollouts (I think it was 800 for AlphaZero) implies to me that the added complexity is not yet optimized or favorable for LLMs. [1] https://arxiv.org/abs/2406.07394 https://arxiv.org/abs/2406.07394