LLMs can't self-correct in reasoning tasks, DeepMind study finds

Re:LLMs can't self-correct in reasoning tasks, DeepMind stud... by auggierose on Tuesday October 10, 2023 @11:55AM (#37833783)

I guess we have different perspectives on this. I don't define the intelligence of something by how well I can turn it into a tool. Obviously, the hope is that with increasing intelligence this becomes easier, but this is not necessarily so, and you might have to find the right way to harness its intelligence. A simple approach might not work. But just because you failed to harness its intelligence for your purpose, doesn't mean it isn't intelligent.

As I said before, is a baby intelligent? Of course. Could you use it for any kind of "production purpose"? I hope not. What about a 3-year old? You will have noticed that it can be difficult to get full-blown intelligent adults to do what you want them to do. This might even get more difficult with increasing intelligence.

[ Parent | on HN ]

Re:LLMs can't self-correct in reasoning tasks, DeepMind stud... by intended on Tuesday October 10, 2023 @02:46PM (#37835881)

The difference here is between code that is intelligent, and the text version of autotune.

And again, I ask you to put your money where your mouth is. If you are willing to assume I wasn't able to harness its intelligence, please prove me wrong.

There is nothing I would really want more, than to have genuinely autonomous systems.

My point at the start, and now is that by using Human terms to examine this phenomenon leads people to assume things about what is going on.

Testing and Evidence are what reason is built on. Asking someone to follow the scientific method, I would hope does not construe boorishness on my part.

[ Parent | on HN ]

Re:LLMs can't self-correct in reasoning tasks, DeepMind stud... by auggierose on Tuesday October 10, 2023 @03:20PM (#37836269)

Not sure how much more I can explain my point of view. I have used ChatGPT 4 for many tasks which require intelligence. Others have too. It worked, many many times. Summarising an unknown text for example requires intelligence. Proving a novel version of a mathematical theorem requires intelligence. Translating natural language into a new logic I just invented requires intelligence. Plenty of testing and evidence here. It also failed many times, but often this was because of inherent ambiguity in the task, which it helped to expose. That's pretty intelligent.

The scientific method only works if you accept the evidence. Some people don't believe that we landed on the moon. Well.

You are telling me you could not use it for what you would have hoped to use it for, and you are not allowing the use of the term intelligent until an LLM can do that for you. If that is your definition of intelligence, good for you.

But I would suggest the following instead: What the scientific method has proven is that, if you feed a very simple mechanism with enough data, then intelligence emerges.

[ Parent | on HN ]