Yann LeCun Got $1 Billion For World Model AI. These Robots Learned 1,000 Real-World Tasks In 24 Hours

The LLM era of AI has been defined by scale: more parameters, more tokens, and more compute. While GPT-3 had 175 billion parameters, and billions of tokens, modern frontier models might have hundreds of billions more parameters, and trillions of tokens of training data. Former Meta AI lead Yann LeCun just raised $1 billion to test his long-standing belief that more tokens is not the answer for physical AI, and that world models are the next frontier of learning, especially for robots. But there’s already a real-world training methodology for physical AI that teaches robots like humans, and it’s seeing some incredible success.