World models
Models that simulate a toy version of the real world, allowing an AI to generate or reason about data and processes it hasn't directly observed.
As defined in Ai Sota #2026
Other definitions
AI systems that internally model how the physical world works, allowing them to predict the consequences of their own actions at an abstract level of representation and plan action sequences through optimization and search rather than autoregressive prediction.
Yann LeCun Says LLMs Have 2 Years Left…
Where it comes up · 6
- Yann LeCun Says LLMs Have 2 Years Left…5 claims
- The future of intelligence | Demis Hassabis (Co-founder and CEO of DeepMind)5 claims
- DeepSeek's Invisible Founder Just Talked for 4 Hours. It Leaked.2 claims
- Demis Hassabis: Why AGI is Bigger than the Industrial Revolution & Where Are The Bottlenecks in AI2 claims
- Ai Sota #20261 claim
- Context Graphs: AI's Next Big Idea1 claim
Most important claims · 10 of 16Search all claims for “World models” →
- You can drop agents into simulated worlds created by world models, allowing them to explore with curiosity as their main motivator, and this approach to learning through exploration in simulated environments has potential for creating agents that discover novel solutions.The future of intelligence | Demis Hassabis (Co-founder and CEO of DeepMind)
- Humans predict consequences of actions at an abstract level of representation rather than at pixel level, understanding physical principles and abstractions rather than precise pixel-perfect outcomes, which is how world models should operate.Yann LeCun Says LLMs Have 2 Years Left…
- Unlocking more LLM capabilities automatically accelerates all other AI fields, because the researchers and engineers in those fields (e.g. robotics) use LLMs for coding, so optimizing coding LLMs pays off across domains—and the world-class transformer training infrastructure built for language models is now elevating robotics learning.Ai Sota #2026
- Cost functions in world model systems might be inaccurate or world models might make incorrect predictions, so objective-driven systems can still make mistakes.Yann LeCun Says LLMs Have 2 Years Left…
- Truly intelligent systems can solve new tasks zero-shot after being exposed to their environment, something humans are completely capable of and many animals also demonstrate.Yann LeCun Says LLMs Have 2 Years Left…
- With world models, systems could solve problems with a small amount of training data or no training data at all, plus perhaps some RL-style fine-tuning.Yann LeCun Says LLMs Have 2 Years Left…
- Hassabis disagrees with Yann LeCun on the future post-LLMs; while LeCun believes something different is needed, Hassabis assigns roughly 50/50 probability that some missing breakthroughs (perhaps world models) are needed, but bets strongly that foundation models will continue being successful and won't be replaced.Demis Hassabis: Why AGI is Bigger than the Industrial Revolution & Where Are The Bottlenecks in AI
- Yann LeCun's previous contrarian bet on neural networks in the face of widespread skepticism validates the credibility of his current contrarian position on world models versus scaling.Yann LeCun Says LLMs Have 2 Years Left…
- Serious figures like Yann LeCun argue that language models alone cannot reach human-level intelligence; they contend that world models—systems that can predict the effects of actions—are necessary for AGI. Liang is betting the entire company against this thesis.DeepSeek's Invisible Founder Just Talked for 4 Hours. It Leaked.
- The central question of Demis's life is determining the limits of what Turing machines can compute, and everything Deep Mind does—from AlphaGo to protein folding to world models—is an attempt to push the boundaries of what computation can achieve.The future of intelligence | Demis Hassabis (Co-founder and CEO of DeepMind)