Yann LeCun, chief AI scientist at Meta, says that a “new paradigm of AI architectures” will emerge in the next three to five years, far beyond the capabilities of existing AI systems.
LeCun also predicted that the next few years could be the “decade of robotics,” in which advances in artificial intelligence and robotics come together to unlock a new class of intelligent applications.
Speaking in Davos on Thursday in a session titled “Technology Debate,” LeCun said that the “flavor of AI” we have right now – namely generative AI and large language models (LLMs) – does not it’s right up there. so much. It’s useful, sure, but it falls short on many fronts.
“I think the shelf life of the current paradigm (LLM) is quite short, probably three to five years,” LeCun said. “I think within five years no one in their right mind would be using them anymore, at least not as a central component of an AI system. I think (….) we will see the emergence of a new paradigm for AI architectures, which may not have the limitations of current AI systems.”
These “limitations” inhibit the machines from truly intelligent behavior, LeCun says. This is due to four main reasons: a lack of understanding of the physical world; a lack of persistent memory; a lack of reasoning; and a lack of complex planning skills.
“LLMs really aren’t capable of doing any of that,” LeCun said. “So there will be another AI revolution in the next few years. Maybe we’ll have to change the name, because it probably won’t be generative in the sense we understand it today.”
“World Models”
This echoes sentiments LeCun has espoused in the past. At the heart of all this are what will come to be known as “world models” that promise to help machines understand the dynamics of the real world. This includes having memory, common sense, intuition and reasoning ability, characteristics that go far beyond those of current systems, which primarily involve pattern recognition.
LeCun previously said this could still take about 10 years, but today’s estimate brings things closer to the horizon. Though how far it will get in that time frame isn’t exactly clear.
“LLMs are good at manipulating language, but not at thinking,” LeCun said. “So that’s what we’re working on: getting systems to build mental models of the world. If the plan we are working on is successful, in the timeframe we hope for, within three to five years we will have systems that are a completely different paradigm. They might have some level of common sense. They might be able to learn how the world works by observing it and perhaps interacting with it.”
“The decade of robotics”
As impressive as generative AI, capable of passing the bar exam or discovering new drugs, is, LeCun believes robotics could be a central component of the next wave of AI applications in such real-world scenarios.
Meta itself is doing research work in the realm of robotics, but so is the AI darling of the moment, ChatGPT creator OpenAI. Earlier this month, new job listings emerged detailing a new OpenAI robotics team focused on “general,” “adaptive,” and “versatile” robots capable of human-like intelligence in world contexts real.
“We don’t have robots that can do what a cat can do — understanding the physical world of a cat is far superior to anything we can do with artificial intelligence,” he said. “Maybe the next decade will be the decade of robotics, maybe we will have artificial intelligence systems smart enough to understand how the real world works.”
TechCrunch has a newsletter focused on AI! Sign up here to receive it in your inbox every Wednesday.
