Yann LeCun, a prominent figure in artificial intelligence, argues that large language models (LLMs) will drive many real-world applications but will not achieve human-level thinking. His stance suggests a potential reevaluation of AI investments, pushing the industry to focus on practical uses rather than speculative artificial general intelligence (AGI).
The limits of LLMs
LeCun's argument centers on the core architecture of LLMs. These models predict the next token in a sequence, which he sees as fundamentally different from understanding or reasoning. Without additional cognitive mechanisms, he contends, they will never match human intelligence.
Practical applications over AGI
If LeCun is right, the implications for investment are clear. Money currently flowing into AGI research could shift toward tools that solve concrete problems today — in healthcare, customer service, education, or automation. LLMs already power chatbots, search engines, and content generators, and that's where the near-term value lies.
A growing divide
LeCun's view puts him at odds with those who believe scaling LLMs will eventually produce general intelligence. That divide is likely to shape discussions at AI conferences and in corporate strategy meetings. The question now is whether the industry will double down on LLM scaling or diversify into other approaches.
The outcome of that debate will define the next phase of AI development.
