Yann LeCun has appeared to be a dissenting voice when it comes to the current excitement around LLMs, but even he seems to believe that we are well on the path to AGI.
Meta’s Chief AI Scientist Yann LeCun has said that AGI (Artificial General Intelligence) is achievable in a 10-year timeframe. “You think AGI far, far away or unlikely?” he was asked on a podcast. “No, I don’t think it’s that far away. I don’t think my opinion about how far it is is very different from what you will hear from some (Sam) Altman or Demis Hassabis. (It can happen) quite possibly within a decade,” he added.
“But it’s not going to happen next year. It’s not going to happen in two years,” LeCun continued. “It’s going to take longer. You don’t want to extrapolate the capabilities of LLM and say we’re just going to scale up LLM, train them on bigger computers and more data, and human level intelligence is going to emerge. It’s not going to work this way. We’re going to have to have those new architectures and systems that learn from the real world and can plan hierarchically,” he added.
Yann LeCun has been famously less enthused than many other researchers about the progress of LLMs over the last couple of years. While some scientists say that the current progress in the capabilities of LLMs is evidence they’re the on path to eventual superintelligence, LeCun has said that LLMs by themselves aren’t very smart. Just last month, he’d said that LLMs weren’t even as smart as a cat. “Right now, we are not even at the level of a cat. So, we’ve got some ways to go before we get to human level. Any human,” he’d posted on X.
But it appears that YeCun seems to still believe that we’ll reach superintelligence in a decade, which could indicate extraordinary progress in the next few years. He does concede that it’ll mean there’s some luck and technical advancements on our side. “5 to 10 years would be if everything goes great, all the plans that we’ve been making will succeed. We’re not going to encounter unexpected obstacles, but that is almost certainly not going to happen,” he said. But it does appear that the people building AI systems are becoming increasingly bullish on the prospect of creating AGI — OpenAI’s Sam Altman said that they now know what to do to build AGI, and Anthropic’s Dario Amodei says that AGI is possible by 2026. And with even Meta’s Yann LeCun’s giving a 5-10 timeframe, AGI might well be upon us sooner than we expect.