Hacker News Reader

Comment by ukuina | Hacker News Reader

But the GPT4 that is accessible via ChatGPT and the API is *not* the base model. It is the instruct-tuned, post-RLHF variant.

And while GPT4 does not have a "world model", can we please move on from the stochastic parrot argument?

My inability to emit more than one word at a time (either spoken or written) does not mean I lack multidimensional planning and conceptualization; why does this line of thinking not extend to LLMs?

View on HN

Because there are no intermediate internal registers with values that can be altered by a prompt, it's prompt in -> pass through deterministic computations with fixed trained weights -> token out. Context-dependent planning (beyond what you get from next-token prediction, which is already a lot) requires additional mutable state.

Yeah but it gets the whole prompt every time, right? So the whatever state it could compute for the first forward pass, it could also compute on the subsequent one. Hence, does it matter that it can’t store the state, if it can recompute it?

Exactly. It's inefficient for LLMs to recompute state based on previous output, but that's what we have so far, and it works, surprisingly.

Even if it is "instruct-tuned, post-RLHF," saying this process creates personas is still anthropomorphizing your computer program. I'm not saying it's output is meaningless.

The thing is, the phrase "Odds are ... median response across all possible chat personas" is nonsensical in so many ways. How are you going to define chat personas? Once you do this, how do you order them to take a median? Then, why is this something that will happen with some probability? This is not a quantifiable or testable thing that can happen. It doesn't have a semantic meaning. Maybe it has an emotional meaning like "I'm excited about GPT-4", and I would agree about that! :D

Reply on news.ycombinator.com