Yann LeCun

I see the danger of this concentration of power through proprietary AI systems as a much bigger danger than everything else. What works against this is people who think that for reasons of security, we should keep AI systems under lock and key because it’s too dangerous to put it in the hands of everybody. That would lead to a very bad future in which all of our information diet is controlled by a small number of companies who proprietary systems. (2024) source Verified

Calls for a global A.I. regulator modelled on the IAEA are misguided. Nuclear technology is a narrow, slow‑moving domain with obvious materials to track and a small set of state actors; A.I. is a broad, fast‑moving field with millions of researchers and developers worldwide. A U.N.-led, IAEA‑style body that ‘oversees’ compute‑intensive A.I. would be unworkable in practice and harmful in principle: it would freeze progress, entrench incumbents, and starve open research — all while failing to stop bad actors who won’t participate. What we need instead is open science, open models, and targeted rules for concrete harms. Safety and robustness should be advanced by more eyes on the code and more researchers able to test and improve systems — not by a centralized global authority trying to police computation itself. (2023) source Unverified

The point is, AI systems, as smart as they might be, will be subservient to us. We set their goals, and they don't have any intrinsic goal that we would build into them to dominate. It would be really stupid to build that. It would also be useless. Nobody would buy it anyway. (2024) source Unverified

My first reaction to [the Pause Giant AI Experiments letter] is that calling for a delay in research and development smacks me of a new wave of obscurantism. (2023) source Unverified

Human feedback for open source LLMs needs to be crowd-sourced, Wikipedia style.
It is the only way for LLMs to become the repository of all human knowledge and cultures. source Unverified

AI could theoretically replace humans, but it is unlikely due to societal resistance. Humans would remain in control, effectively becoming the 'boss' of superintelligent AI systems. [...] He downplayed fears of a doomsday scenario caused by AI, labeling them as sci-fi clichés, and argued that current AI advancements are not close to achieving superintelligence. He suggested that to mitigate misuse and unreliability in AI, the focus should be on creating better AI systems with common sense and reasoning capabilities. (2025) source Unverified

I don't like the phrase AGI. I prefer human-level intelligence because human intelligence is not general. Internally, we call this AMI-advanced machine intelligence. We have a pretty good plan on how to get there. First, we are building systems that understand the physical world-which learn by watching videos. Second, we need LLMs to have persistent memory. Humans have a special structure in the brain that stores our working memory, our long-term memory, factual, episodic memory. We don't have that in LLMs. And the third most important thing is the ability to plan and reason. (2024) source Unverified