Why Godfather Of AI Yann LeCun Believes AGI Nowhere Near Human Intelligence

With companies like Google, Meta and OpenAI racing to build Artificial General Intelligence (AGI), the godfather of AI, Yann LeCun believes that today’s advanced models, including ChatGPT, still lack a real

understanding of how the world works. Because of this, he says AI is still a long way from thinking like humans.During a recent episode of Scientific Controversies, LeCun was asked whether large language models (LLMs) can eventually scale up to achieve AGI or if their limitations are already becoming visible. He said that LLMs have definitely improved over the years, but their growth is uneven.According to him, the biggest jumps are happening in areas like -- mathematics, codin and chess.LeCun explained why this happens. These tasks rely mostly on symbol manipulation, where models search through different combinations of symbols to find the correct output.ALSO READ: Google To Soon Launch A Gadget That Founder Sergey Brin Dubbed A Mistake In maths, coding, or chess, the rules are clear and structured, so the model can simply explore different sequences and arrive at the right answer. Because of this, language models perform much better in such organised environments.However, he said these represent only a very small part of real human intelligence. When it comes to real-world tasks like planning, interacting with physical objects,or understanding how things behave, current AI systems is not ready. These tasks need a deeper understanding of the physical world. For example, a robot picking up an object must know what is possible, how much force to apply and what might happen next. LeCun believes today’s models do not have this kind of knowledge.He also stressed that symbolic reasoning alone cannot solve real-world problems and that AGI will need new approaches that go far beyond today’s language models."There’s been generation after generation of AI scientists since the 1950s claiming that the technique that they just discovered was going to be the ticket for human-level intelligence. You see declarations of Marvin Minsky, Newell and Simon, and Frank Rosenblatt, who invented the perceptron—the first learning machine—in 1957, saying, 'Within 10 years we’ll have machines that are as smart as humans.’ They were all wrong. This generation with LLMs is also wrong. I have seen three of those generations in my lifetime," he said.