4 ms·

Static computation is not reasoning (these models are not building up an argument from premises, they are merely finding statistically likely completions). Com

Static computation is not reasoning (these models are not building up an argument from premises, they are merely finding statistically likely completions). Computational thinking/reasoning would be breaking down a problem into an algorithmic steps. The model is doing neither. I wouldn't confuse the fact that it can break it into steps if you ask it, because again that is just regurgitation. It's not going through that process without your prompt. That is not part of its process to arrive at an answer.

ramoz 2y ago
The point is emergent capabilities in LLMs go beyond statistical extrapolation, as they demonstrate reasoning by combining learned patterns. When asked, “If Alice has 3 apples and gives 2 to Bob, how many does she have left?”, the model doesn’t just retrieve a memorized answer—it infers the logical steps (subtracting 2 from 3) to generate the correct result, showcasing reasoning built on the interplay of its scale and architecture rather than explicit data recall.
thelastbender12 2y ago
I kinda agree with you but I can also see why it isn't that far from "reasoning" in the sense humans do it. To wit, if I am doing a high school geometry proof, I come up with a sequence of steps. If the proof is correct, each step follows logically from the one before it. However, when I go from step 2 to step 3, there are multiple options for step-3 I could have chose. Is it so different from a "most-likely-prediction" an LLM makes? I suppose the difference is humans can filter out logically-incorrect steps, or prune chains-of-steps that won't lead to the actual theorem quicker. But an LLM predictor coupled with a verifier doesn't feel that different from it.
int_19h 2y ago
Models trained to do chain-of-thought do it without any specific prompting, and a specific system prompt that enables this behavior is often part of the setup, because that is what the model was trained on. I don't see how that is "regurgitation", either, if it performs the reasoning steps first, and only then arrives at the answer.

ramoz 2y ago

The point is emergent capabilities in LLMs go beyond statistical extrapolation, as they demonstrate reasoning by combining learned patterns. When asked, “If Alice has 3 apples and gives 2 to Bob, how many does she have left?”, the model doesn’t just retrieve a memorized answer—it infers the logical steps (subtracting 2 from 3) to generate the correct result, showcasing reasoning built on the interplay of its scale and architecture rather than explicit data recall.

thelastbender12 2y ago

I kinda agree with you but I can also see why it isn't that far from "reasoning" in the sense humans do it. To wit, if I am doing a high school geometry proof, I come up with a sequence of steps. If the proof is correct, each step follows logically from the one before it. However, when I go from step 2 to step 3, there are multiple options for step-3 I could have chose. Is it so different from a "most-likely-prediction" an LLM makes? I suppose the difference is humans can filter out logically-incorrect steps, or prune chains-of-steps that won't lead to the actual theorem quicker. But an LLM predictor coupled with a verifier doesn't feel that different from it.

int_19h 2y ago

Models trained to do chain-of-thought do it without any specific prompting, and a specific system prompt that enables this behavior is often part of the setup, because that is what the model was trained on. I don't see how that is "regurgitation", either, if it performs the reasoning steps first, and only then arrives at the answer.