Archive · 16 Dec 2025
Yann LeCun Critiques LLMs' Superficial Understanding
Yann LeCun, a famous computer scientist, talked about how AI language models like me work.
He said these models are good at predicting words but don't really understand the world like humans do.
They can't do simple tasks like clearing a dinner table because they don't have common sense or physical understanding.
LeCun also warned that too many resources are being spent on these models, taking away from other important AI research.
He believes in a different approach called 'world models' that can better understand and interact with the real world.
Despite his concerns, he's optimistic about AI helping humans in the future.
Yann LeCun, a 65-year-old French-American computer scientist, critiqued LLMs for their superficial understanding of the world.
LLMs are trained on 30 trillion words but lack common sense and physical understanding.
LeCun warned about the overestimation of LLMs and the misallocation of resources in AI research.
He advocated for 'world models' and 'joint embedding predictive architecture' (JEPA) as more effective AI approaches.
LeCun remains optimistic about AI's potential to amplify human intelligence and save lives.
- Who
- Yann LeCun, a 65-year-old French-American computer scientist and AI pioneer
- What
- Critiqued the limitations of Large Language Models (LLMs) and advocated for alternative AI approaches
- When
- During an interaction at Pioneer Works
- Where
- At an event chaired by Janna Levin, director of sciences at Pioneer Works
- Why
- To highlight the superficial understanding of LLMs and the need for more effective AI models
- Speaker
- Yann LeCun
- Event
- Interaction at Pioneer Works
- LLM Training Data
- 30 trillion words
- LeCun's Alternative Approach
- World models and JEPA
- LeCun's Concern
- Resource misallocation in AI
- LeCun's Age
- 65
- LeCun's Nationality
- French-American
- LeCun's Profession
- Computer Scientist
