The Agentic Self: Parallels Between AI and Self-Improvement

muratbuffalo.blogspot.com

28 points·zdw··8 comments

Is Murat suggesting that thinking equals predicting the next token or is this just fiction to read?

Except that something like half of people don't have any internal monologue. It's tempting to pretend that LLM are doing similar things to our brains, but the reality is that they are extremely different and only very very superficially appear to be doing similar things.

Even those of us with an internal monologue aren't using it most of the time. I doubt I've ever though "now let me scroll down to the next comment", for example.

... is this true?

I'm literally always thinking in that way. If I focus a bit harder on it, it's apparent that there's another, faster cognitive layer beyond my background monologue as well.

Yes for me it's true. Most of my actions are not part of a monologue. However I use it to think through possible conversations or to carry myself through tough problem solving problems or complex processes.

> One of the most profound pieces of advice I ever read as a PhD student came from Prof. Manuel Blum, a Turing Award winner. In his essay "Advice to a Beginning Graduate Student", he wrote: "Without writing, you are reduced to a finite automaton. With writing you have the extraordinary power of a Turing machine."

only if you have an unlimited amount of paper!

otherwise you're still a finite state machine (technically)

And also only if you know how to do boolean algebra.

There feels to be something special about reducing the computation to self reflect or just be in the tao/dao

Maybe it's still in the state machine but it feels like one escapes the trajectory of a Turing machine running the program

Reply on news.ycombinator.com