Meta’s chief AI scientist Yann LeCun predicts ‘new paradigm of AI architectures’ within 5 years

Meta chief AI scientist, Yann Le Cunn, recently predicted “a new paradigm shift of AI architectures” within five years.

Speaking at a session at the World Economic Forum in Davos, Le Cunn said that the Artificial Intelligence (AI) we know right now, which is generative AI and large language models (LLMs), are not capable of much.

They get the basics done but still fall short. And in the next five years, “nobody in their right mind would use them anymore”, he said. “I think the shelf life of the current [AI] paradigm is fairly short, probably three to five years,” LeCun added.

“We’re going to see the emergence of a new paradigm for AI architectures, which may not have the limitations of current AI systems,” he predicts.

LeCun also believes that in the next few years, we could be living in the “decade of robotics,” when we will see a whole new application of technology that combines robots and AI.

LeCun also explains why he thinks the current AI models can't do much. He gives four reasons for that. One, he says, is that the current models lack awareness and understanding of the physical world. Two, there is a limitation to how much it remembers at once, and does not have a continuous memory. Three, it lacks the power of reasoning. And four, it is unable to perform complex planning tasks.

“So there’s going to be another revolution of AI over the next few years. We may have to change the name of it, because it’s probably not going to be generative in the sense that we understand it today,” LeCun says.

LeCun mentioned that the “AI revolution” could be some 10 years away, but with the way AI is progressing right now, the big change could be closer.

“LLMs are good at manipulating language, but not at thinking,” LeCun said.

While “Debating Technology” at Davos, LeCun may have also revealed what Meta’s AI labs are working on right now.

“So that’s what we’re working on — having systems build mental models of the world. If the plan that we’re working on succeeds, with the timetable that we hope, within three to five years we’ll have systems that are a completely different paradigm,” he said.

“They may have some level of common sense. They may be able to learn how the world works from observing the world and maybe interacting with it.” he added.

LLMs are a type of artificial intelligence (AI) that can generate and understand human-like text. LLMs are trained on large amounts of data and are based on machine learning.

Interestingly, as LeCun spoke about the current AI and LLM models’ lack of performing complex tasks, OpenAI and Perplexity on June 23, unveiled new agentic AI, which they claim excel at performing complex, multi-step tasks.

AI is a set of technologies that allow computers to perform tasks that normally require human intelligence.

For instance, OpenAI’s AI agent called Operator –– which is unfortunately available only in the US right now –– is capable of ordering groceries for you if you give it a shopping list, or it can book flight tickets for you if you share your itinerary with it. It can also create memes for you. Basically, anything you do on the web, the agentic AI can do on your behalf.