Meta's Chief AI Scientist Predicts AGI Achievable Within Ten Years

Meta's Chief AI Scientist Yann LeCun has stated that Artificial General Intelligence (AGI) is likely to be achieved within the next decade. In a podcast interview, when asked whether he believes AGI is still out of reach or difficult to realize, he responded, "No, I don't think it's far off. My view aligns closely with that of Sam Altman and Demis Hassabis. Under the right conditions, AGI could become a reality within 10 years."

LeCun further explained, "However, this does not mean AGI will be achieved within a year or two; it requires more prolonged efforts. Simply enhancing the capabilities of large language models (LLMs) by training them on more powerful computers and larger datasets is insufficient to reach human-level intelligence. This approach is ineffective. We need entirely new architectures and systems that can learn from the real world and possess hierarchical planning capabilities."

Regarding the advancements of LLMs in recent years, Yann LeCun’s outlook is less optimistic compared to other researchers. While some scientists believe that improvements in LLM capabilities indicate a progression towards superintelligence, LeCun argues that relying solely on LLMs is not prudent. Just last month, he stated that the intelligence of LLMs is inferior to that of a cat. "We have not yet reached the level of a cat, so we still have a long way to go before attaining human-level intelligence, regardless of which human you compare to," he wrote in a post on a platform.

Nevertheless, LeCun remains confident that superintelligence could be achieved within the next ten years, implying significant breakthroughs in the coming years. He acknowledges that this depends on some luck and technological advancements: "If everything goes smoothly and our plans are successfully implemented, AGI could be realized within five to ten years. Of course, we might not encounter unexpected obstacles, but such occurrences are highly unlikely." Currently, those dedicated to building AI systems are becoming increasingly optimistic about the prospects of creating AGI. OpenAI’s Sam Altman claims they now know how to build AGI, while Anthropic’s Dario Amodei believes AGI may be realized by 2026. Even Meta’s Yann LeCun has provided a five to ten-year timeframe, suggesting that AGI could emerge sooner than we anticipate.