world model

plural world models

1. artificial intelligence; reinforcement learning; artificial intelligence; reinforcement learningA predictive model that has been internalised by an agent which is used to forecast what consequences its actions will have on the environment (i.e. world) that it interacts with.

   The generation of mostly realistic-looking videos from prompts *does not* indicate that a system understands the physical world. Generation is very different from causal prediction from a world model.

   2024 February 17, @ylecun, Twitter

   3 more examples

   Current problem solving systems are constrained in their applicability by inadequate world models. We suggest a world model based on a temporal logic. This approach allows the problem solver to gather constraints on the ordering of actions without having to commit to an ordering when a conflict is detected.

   1983, James F. Allen, Johannes A. Koomen, “Planning using a temporal world model”, in Proceedings of the Eighth International Joint Conference on Artificial Intelligence, page 741

   In this work, we build probabilistic generative models of OpenAI Gym environments. The RNN-based world models are trained using collected observations recorded from the actual game environment. After training the world models, we can use them mimic the complete environment and train agents using them.

   2018, David Ha, Jürgen Schmidhuber, “World Models”, in arXiv, page 2

   His solution for achieving that [human-level intelligence] relies on an architecture called V-JEPA, a so-called world model. World models aim to understand the physical world by learning from videos and spatial data, rather than just language.

   2026 January 3, Melissa Heikkilä, “Lunch with the FT: Yann LeCun”, in FT Weekend (Life & Arts section), London: The Financial Times Ltd., →ISSN, →OCLC, page 3