The type of artificial intelligence that powers systems like OpenAI’s ChatGPT and Google’s Gemini will not be able to reach human levels of intelligence, Meta’s AI chief Yann LeCun told the Financial Times in an interview published Wednesday, giving an insight into how the tech giant plans to develop the technology moving forward just weeks after its plans to invest heavily spooked investors and erased hundreds of billions from its market value.
Meta Chief AI Scientist Yann LeCun said LLMs won’t reach human level of intelligence.
Getty Images
Key Takeaways
- Large language models, the type of artificial intelligence that drives popular generative AI tools like OpenAI’s ChatGPT, Meta’s Llama and Google’s Gemini, will never be able to reach human levels of planning and reasoning, Meta’s AI chief Yann LeCun told the Financial Times.
- The models, often just called LLMs, are trained using vast amounts of data and their ability to accurately answer prompts is limited by the nature of the data they are trained on, LeCun said, meaning they are only accurate when they have been given the right kinds of training data.
- LLMs have a “limited understanding of logic,” do not have persistent memory, don’t understand the physical world and cannot plan hierarchically, LeCun explained, adding that they “cannot reason in any reasonable definition of the term.”
- As they are only accurate when fed the right training data, LeCun, considered one of three “AI godfathers” for his foundational role in the field, said LLMs are also “intrinsically unsafe” and that researchers hoping to develop human-level AI should be looking at other models.
- LeCun said he and his roughly 500 strong team at Meta’s Fundamental AI Research lab are working to develop an entirely new generation of AI systems based on an approach called “world modeling,” where the system builds an understanding of the world around it like humans do and develops a sense of what would happen if something changes based off this.
- It could take up to 10 years to achieve human-level AI using the world modeling approach, LeCun predicted.
Key Background
Meta, the parent company of some of the world’s most recognizable tech properties like Facebook, Instagram and WhatsApp, is at the forefront of advances in generative AI. Meta’s LLMs like Llama 3 are considered among the most capable in the world, and its assistant Meta AI recently launched as a standalone app, bringing the company into direct competition with market leaders like OpenAI’s ChatGPT, Anthropic’s Claude and Google’s Gemini. AI models are expensive to train and investors have responded less enthusiastically to Meta’s decision to go all in on Silicon Valley’s latest craze.
The company shed nearly $200 billion in value in April and sparked a wider tumble among tech stocks after CEO Mark Zuckerberg told investors to prepare for slower growth while the company pumps funds into AI. He urged investors to show patience as it builds out its AI capabilities at significant expense despite there being no indication of where, when or even how the investment will pay off down the line.
He stressed Meta has successfully followed a similar build-then-monetize strategy in the past with products like Reels, Stories and News Feed and believes it will benefit from leading the pack with AI as well.
Forbes Valuation
Zuckerberg is worth an estimated $165.5 billion. His fortune, a large portion of which comes from his founding stake in Meta’s precursor, Facebook, ranks him as the world’s fourth richest person, sitting behind Bernard Arnault, Jeff Bezos and Elon Musk, with whom he has publicly feuded and agreed to fight in a cage match. Along with his wife, Priscilla Chan, Zuckerberg has pledged to give away 99% of their stake in Meta within their lifetimes.
