LeCun: LLMs Are Valuable Tools, Not Human-Like Thinkers

Yann LeCun, Meta’s chief AI scientist, argues large language models are commercially useful but cannot think like humans.

The technology powers coding, search, and automation, justifying massive infrastructure spending. But LeCun says next-token prediction is a “dead end” for genuine intelligence.

A child learns physics from limited visual experience; LLMs need trillions of tokens to mimic understanding.

AMI Labs raised $1.03 billion in seed funding to build AI systems learning from raw sensory inputs, not text. Meta’s Leham model uses JEPA architecture for planning efficiency.

What This Means for Investors

If LeCun is correct, AI stocks pricing in AGI upside may be overvalued. The billion-dollar bet on AMI Labs signals skepticism toward scaling laws.

Risk remains: OpenAI, Anthropic, and Google believe bigger models unlock new capabilities. LeCun, a Turing Award winner, warns investors of a potential category error.