(CTN News) – At a Meta launch event in Paris, Meta’s chief scientist gave a groundbreaking statement stating that generative AI, the technology that powers ChatGPT, has come to a point where it has reached a dead end.
As an alternative, he promised that new artificial intelligence models would be developed that would exhibit a kind of rationality that is similar to that of humans.
He expressed his dissatisfaction with the current state of AI and machine learning, emphasizing the fact that whereas humans possess common sense, machines lack it.
Meta’s new Artificial Intelligence project called the image-based Joint Embedding Predictive Architecture (JEPA) was presented by him.
As a result of this ambitious endeavor, the machines will be capable of conceptualizing abstract ideas instead of simply regurgitating information from the internet.
By doing so, they will surpass the capabilities of generative artificial intelligence.
Apparently, LeCun believes that generative models have now become a thing of the past since Meta plans on shifting its focus towards joint embedding predictive architectures in the future.
Throughout his presentation, LeCun was confident that generative large language models would cease to be used in the next few years, as a better replacement will be developed.
As an expert in artificial intelligence, LeCun has been critical of the hype surrounding generative AI models such as ChatGPT and the image-based Dall-E since the launch of both models last year.
Consequently, LeCun believes that we have an exaggerated perception of the capabilities of generative AI because of fears and excitement surrounding it. Mark Zuckerberg, Meta’s CEO, announced on Facebook that the JEPA tool would be open source, allowing researchers to experiment with it and continue to develop it in the future.
It is the goal of this project to develop artificial intelligence that closely aligns with how humans perceive the world. As Zuckerberg stated “We need computational models to be able to evaluate the world and make predictions about it. This research is a step in the right direction.”.
With its Facebook and Instagram platforms, Meta is adopting a more discrete approach to ChatGPT-style artificial intelligence for its social media platforms in contrast to its competitors.
In spite of the fact that Meta has incorporated generative AI into its products, it has done so with less publicity, when compared to companies like Microsoft or Google.
On the other hand, Meta has released open-source AI models which do not require as much computing power as the technology behind ChatGPT, but require less computing power overall.
A significant shift in the AI landscape has been marked by LeCun’s announcement, with Meta spearheading the development of new AI models that strive to replicate the rationality of human beings.
There can be no doubt that the JEPA project represents an innovative step in the direction of enhancing machine intelligence beyond the limitations of generative artificial intelligence.