Hacker News Reader

Comment by TeMPOraL | Hacker News Reader

My take for almost a year now is: LLMs are like your inner voice.

The part where we can "analyse what we just said (or were about to say) (or what we just typed and didn't press send yet!) and conceptualise it as a whole to evaluate its veracity" involves going back-and-forth in yourself. A loop. LLMs are the part inside the loop, and so if you want to get better results, you have to feed the output back to LLM.

This is, arguably, part of what the "conversation mode" does, by always including the growing message log in each request - which is why e.g. GPT-4 is good at correcting a mistake once you say, without pointing it out, that there exist a mistake.

View on HN

> which is why e.g. GPT-4 is good at correcting a mistake once you say, without pointing it out, that there exist a mistake.

It is also good at adding a mistake, so this is completely unrelated. It just rolls to get another random.

What you should add is that ChatGPT does better when it writes out every step, so what you said previously is still true just that your example doesn't properly support it.

The "chain of thought" thing where you tell it to "think step by step" is a slightly different thing. That's a single forward pass, where it can only reason based on what it output earlier. But once you reply to it, and the interface feeds the whole conversation back to the LLM, it now can look at the entirety of its previous answer.

When it does it step by step it can look at every step when it writes down its final answer after the steps. There is no extra magic that happens just because you write something in between.

Edit: To expand on that, a humans inner thought would have caught the mistake at that point and corrected it by writing more instead of ending the text. The LLM basically never does that unless you tell it to fix the answer.

Every mistake you don't catch as being wrong in your inner thoughts simply become a part of your subjective experience/existence until challenged again later, if not forever.

Reply on news.ycombinator.com