With more than 40 years in the IT sector, he is one of the pioneers of modern artificial intelligence. He was also head of AI at Meta and is considered by many to be one of the great voices, given his experience, when it comes to this technology.
Precisely leaving Meta has been his great escape route to be able to say what he really thinks about the AI sector, its movements and, above all, its main protagonists.
The message now is clear. He believes that the Silicon Valley giants are investing billions in projects that do not take AI where they promise, and system errors accumulate like cars in an accident.
«There is a herd effect where everyone in Silicon Valley has to work on the same thing,» he comments in an interview for The New York Times. «This doesn’t leave much room for other approaches that could be much more promising in the long term.»
According to LeCun, the language models in use today, like those that power ChatGPT, have too many limits. They are not the key to achieving human intelligence, much less superintelligence.
«LLMs are not a path to superintelligence, or even human intelligence. I’ve said that from the beginning,» he says. “The entire industry has been taken over by LLM,” he says.
Artificial intelligence is not as smart as you think or they want us to believe
For years at Meta, LeCun has advocated for open research. I believed that sharing progress freely was the best way to move forward and minimize risks. «If everyone is open, the field progresses faster,» he insists.
But now many companies are leaving openness aside to maintain a competitive advantage, and that, according to him, It can cost them dearly compared to faster and more efficient companies, especially in China.
That is why he believes that China is taking steps that could leave Silicon Valley behind. «Good ideas come from there, but in the US there is a superiority complex that makes it difficult to see talent abroad,» he says.
«I’m tired of the term artificial general intelligence»
According to him, Talking about artificial general intelligence is not only a mistake, but a very dangerous one.
Consider that thinking about something like artificial consciousnesses that could at some point replace people is very far from adjusting to the reality that exists today, or that could happen in the future.
What’s more, at a recent Nvidia event, He has acknowledged that he is «fed up with the term artificial general intelligence»which doesn’t stop ringing at all hours.
«AI is destined to become a powerful tool, not an autonomous threat,» LeCun specified.
According to his criteria, it is clear that AI will help workers in many tasks, especially those that are more repetitive and autonomous, but that at no point will it replace humans, as Sam Altman or Elon Musk suggest. Among other things, because he won’t be able to.
Yann LeCun is clear that artificial intelligence is a tool, but nowhere near the revolution that is usually imagined. In fact, he is one of those experts who are clear that AI will not have human capabilities such as understanding or the possibility of long-term planning.
His perspective, much more pragmatic, is situated in an intermediate position between those who consider that AI will change the world, and those (of which there are also some) who describe its advances as a real bullshit. Could it be true, once again, that virtue is found in the middle?