Yann LeCun, an AI pioneer, calls fears of an existential threat from AI “pure bullshit”!

An assumed skepticism in the face of current language models

LeCun pulls no punches when it comes to evaluating the real capabilities of current language models. According to him, these systems, despite their apparent prowess, are far from achieving true intelligence. He points out that large language models (LLM) severely lack fundamental abilities that would be found even in a domestic cat, such as:

• A persistent memory;
• Reasoning ability;
• Planning;
• An understanding of the physical world.

For LeCun, the impressive performances of LLMs are only a demonstration of their ability to “manipulate language without being intelligent“. He even goes so far as to assert that these models will never lead to true artificial general intelligence (AGI).

An alternative vision of AI development

Although skeptical of current approaches, LeCun is not an opponent of the idea of ​​an AGI. He just believes that new methods will be necessary to achieve this. As an example, he highlights the work of his fundamental AI research team at Meta, which focuses on analyzing real-world videos.

According to LeCun, this approach could pave the way for AI systems to gain a deeper, more contextual understanding of the world. And this, unlike language models which simply manipulate symbols without real understanding.

A call for reason in an often alarmist debate

It should be noted that LeCun’s statements come in a context where fears linked to AI are omnipresent in the public debate. By characterizing these concerns as “pure bullshit” (to use his words), the researcher invites a reframing of the debate on the real issues and challenges of AI.

His position contrasts sharply with that of other influential figures in the field, who have expressed concerns about the potential risks of superintelligent AI. LeCun seems to suggest that these concerns are premature. They distract from the real and immediate problems facing the field!

Implications for the future of AI research

LeCun’s remarks could have significant implications for the future direction of AI research. By questioning the dominant approach based on large language models, it opens the way to alternative, potentially more promising avenues of research for the development of truly intelligent AI.