🚀 Yann LeCun recently created AMI (Advanced Machine Intelligence) Labs. In an interview with MIT Technology Review, he revealed how we’re focused too much on LLMs while ignoring (at our advancement peril), other methods of attacking the problem that GenAI is trying to solve.

💡LeCun states that while LLMs are useful for text and code, they lack a true understanding of the physical world. They cannot reason, plan, or predict consequences because they are limited to the discrete world of language—To reach human-level AI, we need a “conceptual breakthrough”. Our currently application stack in the business world doesn’t cut it.

🌍Instead of just predicting the next word, these systems learn by observing video and sensor data—much like a baby learns about gravity—to build an abstract representation of how the real-world operates, (i.e. digital twins).

🔓LeCun is an advocate for open-source AI, criticizing the “closed” approach of labs like OpenAI and Anthropic. He believes AI should be an open platform to ensure a high diversity of “AI assistants” reflecting different languages and value systems, rather than being controlled by a US-China binary system.

🏭The goal isn’t just better chatbots. AMI Labs is targeting agentic systems that can predict the consequences of their actions: Level 5 autonomous driving and truly useful domestic robots, and holistic modeling for complex industrial processes like jet engines and chemical factories.