Yann LeCun

Profile

Chief AI Scientist at meta and Professor at New York University. A foundational figure in Deep Learning and Convolutional Neural Networks.

Research Philosophy

• World Models: Advocates for the development of architectures that can predict and model the physical world, enabling true adaptive AI. See detailed analysis in Yann LeCun’s Argument: World Models for True, Adaptive AI Beyond LLMs.
  • Argues that current systems lack the ability to build internal models of reality necessary for general intelligence.
• Anti-LLM Paradigm: Maintains that “Language is not Intelligence”; argues that large-language-models (LLMs) are fundamentally limited because they lack the reasoning and planning capabilities required for AGI.
  • Critiques the reliance on next-token prediction as insufficient for understanding causal relationships in the world.
• Self-Supervised Learning: Focuses on enabling models to learn directly from sensory input without massive human-labeled datasets.

Key Research & Architectures

• JEPA (Joint-Embedding Predictive Architecture): A non-generative approach to predictive modeling.
  • Proposed as a strategic path beyond large-language-models, aiming to separate representation learning from reconstruction.
  • Designed to work in conjunction with World Models to enable efficient planning and decision-making.

References

• Yann LeCun’s Argument: World Models for True, Adaptive AI Beyond LLMs (Video: “World Models: Enabling the next AI revolution” at ETH Zurich)