You were working on AI long before LLMS became a mainstream approach. But since ChatGipt broke out, it has become almost synonymous with LLMAI.
Yes, and we’re going to change that. The public face of AI is, perhaps, mostly LLMs and various types of chatbots. But the latest of these are not pure LL.M. They are many things besides LLM, such as feedback systems and code that solve specific problems. So we’re going to look at the LLM a little bit, like an orchestrator in the system.
Beyond the LLM, there is a lot of AI behind the scenes that runs a huge part of our society. An in-car assistance program, quick-turn MRI images, are the algorithms that drive social media. It’s all AI.
You are sound in arguing that the LLM can only get us so far. Do you think LLM is over hyped these days? Can you summarize to our readers why you believe that LLMs are not enough?
There’s an understated feel to them, which makes them extremely useful for many people, especially if you write text, do research, or write code. LLMs manipulate language really well. But people are under the illusion, or delusion, that it’s just a matter of time until we can get them to have human-level intelligence, and that’s just wrong.
The really hard part is understanding the real world. This is the Moravec paradox (a phenomenon observed by computer scientist Hans Moravec in 1988): what is easy for us, such as perception and navigation, is difficult for computers, and vice versa. LLM is limited to the discrete world of text. They can’t really reason or plan, because they lack a model of the world. They cannot predict the consequences of their actions. This is why we don’t have a domestic robot that’s as agile as a domestic cat, or a truly autonomous car.
We have AI systems that have human-like and human-level intelligence, but they’re not going to be built on LLMS, and it’s not going to happen in the next year or two from now. It will take some time. There are huge conceptual breakthroughs that are going to happen before we have AI systems that have human-level intelligence. And that’s what I’m working on. And this company, AMI Labs, is focusing on the next generation.
And your solution is global models and the GEPA architecture (GEPA, or “joint-embedding predictive architecture”, is a learning framework that trains AI models to understand the world, created by Lacon when he was in Meta). What is elevator pitch?