Views: 0
Over the past few years, we've become accustomed to a very specific idea of "artificial intelligence": chatbots like ChatGPT or Gemini, capable of writing, summarizing, or programming with astonishing fluency. They all share the same technical foundation: large language models, or LLMs.
These systems are trained on enormous amounts of text extracted from the internet. Then, when we ask them a question, they don't "reason" in the strict sense of the word: they predict, word by word, the most likely answer based on what they've already seen. It's a spectacular ability, but it remains a sophisticated form of statistical prediction, not a true understanding of the world.
Enter Yann LeCun, one of the pioneers of modern artificial intelligence and winner of the 2018 Turing Award. After more than a decade as chief AI scientist at Meta, LeCun announced his departure from the company at the end of 2025, following disagreements with management about the direction of research.
His argument is compelling: continuing to scale LLMs, adding more computing power and more data, will not lead us to human-level intelligence. It lacks something essential: understanding the physical world as, for example, a small child or even as a dog does.
With this idea in mind, LeCun founded AMI Labs (Advanced Machine Intelligence Labs) in Paris, led by entrepreneur Alexandre LeBrun. The market response has been spectacular: the company closed a $1.03 billion funding round, the largest seed round in European history, valuing the company at $3.5 billion. Among its investors are names like Nvidia, Jeff Bezos, and others.
You can see the company's website at: https://amilabs.xyz/
The project's core technology is called JEPA (Joint Embedding Predictive Architecture), an idea that LeCun had already been developing at Meta. Instead of generating text or images pixel by pixel, JEPA learns abstract representations of the world: it understands relationships, causes, and consequences, ignoring irrelevant details.
Simply put: while an LLM asks, "What word comes next?", JEPA asks, "What will happen next, and why?". It's the difference between memorizing symptoms and understanding causes. In other words, it's not about seeing "what" is happening, but about fully understanding "why" it's happening.
AMI Labs is not currently seeking to compete with conversational chatbots. Its initial applications target sectors such as robotics, healthcare, aeronautics, and industrial design: systems capable of predicting how a real-world situation will evolve and planning actions accordingly—something much closer to common sense than simply generating text.
It's worth noting that LeCun isn't alone in this idea: other leading figures in AI, such as Fei-Fei Li with his World Labs project, are exploring similar paths. It seems that a growing segment of the scientific community believes that the future of artificial intelligence lies not just in speaking better, but in understanding better.
It's still too early to know if JEPA will deliver on its promises, but one thing is clear: the race for truly intelligent AI, capable of understanding the world as we do, has just gained a major player. And only time will tell.
EN 
ES