What is a world-model?

post by Adam Shai (adam-shai) · 2023-02-16T22:39:16.998Z · LW · GW · 1 comment

This is a question post.

Contents

  Answers
    2 Viliam
None
1 comment

I have seen the concept of world-model used to talk loosely about a neural networks or an agents understanding of the world, but I was wondering what resources (blog posts, journal articles, etc.) exist that talk more precisely about what is a world-model, what are its parts, how do you know if a given system has one or not, if a system does have one, how do you know its structure, does a system necessarily even have a world-model to begin with, even if it's a trivial one? etc.?

Thanks!

Answers

answer by Viliam · 2023-02-27T13:52:57.122Z · LW(p) · GW(p)

In absence of a better answer, I would say that a system has a world-model if it can answer various questions about the world mostly coherently (i.e. the answers to two different questions should not make completely different assumptions about something).

1 comment

Comments sorted by top scores.

comment by yc (AAA) · 2024-09-25T07:40:41.084Z · LW(p) · GW(p)

https://arxiv.org/pdf/1803.10122 I have a similar question and found this paper source. One thing I am not sure of is if this is no longer the same concept/close enough concept that people currently talk about, nor if this is the origin.

https://www.sciencedirect.com/science/article/pii/S0893608022001150 This paper seems to suggest something at least about multimodal perception with reinforcement learning/agent type of set up.