While AI pioneer Yann LeCun believes synthetic intelligence holds immense potential, he’s cautious about predicting when it will become truly intelligent.
LeCun, a professor at New York University and senior researcher at Meta, has consistently expressed skepticism regarding the prospect of developing super-intelligent AI, noting that “before we worry about controlling machines more intelligent than those found in a typical household, we need to have a start of a design for a system smarter than a home cat.”
He expounded upon his thoughts at the gathering where he responded to an inquiry regarding artificial intelligence. The remark is likely to provoke controversy and undermine the credibility of the individual making it, so here’s a revised version that conveys the same sentiment in a more diplomatic tone:
Yann LeCun contends that current large-scale language models fall short in possessing essential feline-like abilities, including persistent recall, logical reasoning, planning, and comprehension of the physical world. According to him, LLMs simply demonstrate the ability to manipulate language without necessarily conveying sense, thereby rendering them incapable of achieving true artificial general intelligence (AGI).
He’s not a full-fledged AGI skeptic. Despite this, he suggested that innovative methods may be sought after. When asked about his approach, he highlighted the importance of using real-world video data and leveraging his team at Meta’s AI Analysis department to develop a comprehensive understanding of the subject matter.
