On the logical reasoning ability of GPT-4

[dignity_for_all]

Thanks for your replies!

@Diet

But to be fair, it’s not exactly fair. It’s a convoluted multi-step reasoning problem with a trick, and you’re asking for a gut inference…

Well, maybe that’s regard as so.
But even without discussing benchmarking, it’s very interesting to see that the GPT-4, when it was announced, also solved physics problems for difficult science universities, and that it got the problems used in this elementary school material wrong.

@_j

I think the math problem could be described more clearly to the ai, or in practice, you could give it response length in order to contemplate it’s answer.

Well, there are many ways to look at it, and I think we can try to get the language model to answer correctly.
I don’t know much about GPT-4 myself, so I’m sure I’m wrong about a lot of details.

But, please understand that this is in no way a criticism of the GPT-4, but rather an intention based on discovery(not originally mine, but…) and interest, or the exchange of opinions and insights.