Meta’s Chief Scientist Debunks AI Hype, Declares ChatGPT-Style AI Outdated
On Tuesday, the chief scientist of Meta, the company that owns Facebook, declared that generative AI, which powers ChatGPT, had reached a dead end and pledged to develop new AI that mimics human reasoning.
“Today, artificial intelligence and machine learning are really fucked up. People have common sense, machines don’t,” Yann LeCun told reporters at Meta’s launch event in Paris.
LeCun spoke as Meta announced its latest AI project, called Image-Based Joint Embedding Predictive Architecture, or JEPA.
The project aims to go beyond generative AI like ChatGPT and give machines the ability to conceptualize abstract ideas and not just feed off what’s online.
“Generative models are a thing of the past, we’re abandoning them in favor of collaborative embedding of predictive architecture,” LeCun said, promoting the Meta project he leads.
“My prediction is that in a few years generative large language models will no longer be used, we’re better off replacing them,” he added.
LeCun is considered a major AI thinker and has criticized the hype surrounding generative AI models using ChatGPT or image-based Dall-E since their launch last year.
LeCun believes that the fears and excitement surrounding generative AI are fueling its true capabilities.
Meta CEO Mark Zuckerberg said in a Facebook post that the JEPA tool is open source, meaning it would be available to researchers.
He said the goal was to develop artificial intelligence that “more accurately reflects how people understand the world.”
“We need models that visualize the world and make predictions. This research is another step in that direction,” Zuckerberg added.
Compared to its competitors, Meta has adopted a more discreet approach to ChatGPT-style AI on its social media platforms on Facebook and Instagram.
Meta added generative artificial intelligence to its products, but without the same publicity as Microsoft or Google.
At the same time, it has also released open-source AI models that require less computing power than the technology using ChatGPT.
