Home > Business
Agence France-Presse
PARIS — Generative AI, the technology behind ChatGPT, has already hit a dead end, and instead hopes for a new artificial intelligence that resembles human rationality, the chief scientist at Facebook owner Meta said. .
“AI and machine learning today are really bad. Humans have common sense, but machines have no common sense,” Yann Lecan told reporters at a meta-announcement event in Paris.
LeCun spoke as Meta announced its latest AI project called the image-based Joint Embedding Predictive Architecture (JEPA).
The project aims to go beyond generative AI like ChatGPT to give machines the ability to conceptualize abstract ideas, rather than simply spewing out what exists online.
“Generative models are a thing of the past. We are abandoning them in favor of collaborative embedding prediction architectures,” LeCun said, promoting the meta-project he is leading.
“My prediction is that in a few years generative large-scale language models will no longer be used and there will be better ones to replace them,” he added.
LeCun is considered a leading thinker on AI, and since launching ChatGPT and the image-based Dall-E last year, he has continued to criticize the hype around the generative AI models that power them.
LeCun believes that the anxiety and excitement surrounding generative AI has significantly inflated its practical capabilities.
Meta CEO Mark Zuckerberg said in a Facebook post that the JEPA tool is open source, meaning researchers are free to tinker with it.
The aim, he said, is to develop AI that “more accurately reflects how people understand the world.”
“We need models to perceive and make predictions about the world, and this study is another step in that direction,” Zuckerberg added.
Compared to its rivals, Meta takes a more personalized approach to ChatGPT-style AI for social media platforms Facebook and Instagram.
Meta introduced generative AI into their product, but it was not as popular as Microsoft or Google.
In parallel, we also released an open-source AI model that requires less computing power than the technology behind ChatGPT.
Related video
