[ma2rten]

I didn't have time to read this, but it is a single author paper, the author is not affiliated with a research group, it is not peer reviewed, it was published on a preprint server that I have never heard of.

LLMs can definitely perform some kinds of reasoning. For example GSM8K is a dataset of grade school math problem requiring reasoning that LLMs are typically evaluated at. We talk about one method for this in our chain of thought paper [1]

[1] https://arxiv.org/abs/2201.11903

[ramblerman]

> I didn't have time to read this

You had time to respond though, what a silly (and elitist rebuttal).

OP is arguing from a philosophical point of view that GPT-4 can not reason, i.e. is just repeating/parroting on trained logical arguments.

You argue by authority that yes, it actually does reason, which is a far (far) bolder claim than the one OP is making.

[adamsmith143]

>You had time to respond though, what a silly (and elitist rebuttal).

What a dumb take. It probably takes seconds to write a simple comment and far longer to read a paper.

>You argue by authority that yes, it actually does reason, which is a far (far) bolder claim than the one OP is making.

He linked a paper you can read yourself. How is this arguing from authority?

[layer8]

The paper makes detailed and reasoned arguments, and the GP quickly dismisses it without such.

[ramblerman]

> It probably takes seconds to write a simple comment and far longer to read a paper.

Unfortunately that is true, it doesn't mean you have to.

The paper he linked to doesn't rebut OP, they show that prompting GPT-4 to provide reasoning makes it provide better answers. That is a different statement than "GPT-4 is actually reasoning", and can do so consistently on novel problems.

[adamsmith143]

You should probably think about why you think that making the model output reasoning steps which lead it to correctly answer questions it couldn't before is not somehow equivalent to reasoning.