Big Technology Podcast · Wednesday, July 15, 2026
AI pioneer Jürgen Schmidhuber disagrees with the notion that Large Language Models (LLMs) alone can lead to Artificial General Intelligence (AGI). He argues that while LLMs can serve as world models, true learning, like that of a baby, involves self-invented experiments and interaction with the physical world. Schmidhuber contrasts this with ChatGPT's web-downloading approach, emphasizing the need for embodied AI that can interact with and operate physical machines.
“So, when the question is phrased like this, does an LLM by itself lead to AGI? The answer is a clear no.”
“Like a little baby, you know? A little late baby doesn't learn by downloading the web. That's what ChatGPT does. No, it learns by creating its own data streams through its own self-invented experiments.”
“And so, all the things that humans can do with their hands, their physical hands, they don't work well. Only bits and zeros and ones behind the screen, that's the only thing that is working well.”