Superphysicists!

A few months ago I went to visit Rich Sutton in Edmonton. We have known each other for about a decade. We have never worked together on anything. But he is my go-to person for learning about AI things. In addition to being one of the most clear-thinking people I’ve ever met, he is also a great person. When I started my AI journey long ago I was not part of that community. But he always made me feel welcome. He has always been very patient with explaining technical things, and why he thinks about AI in the way he does.

We discussed what was going on in the world of AI. He told me one of his past students, David Silver, was thinking about starting a new company. This company’s mission would be to fully embrace learning from experience. This would lead to both a fundamental understanding of what intelligence was that would apply to all agents, from viruses to people to robots to software systems. It would also be possibly the only principled path to superintelligence. That sounded pretty cool!

I’d never met Dave. But we had a couple of introductory calls to discuss, and decided that it would be cool to get together. I flew to the UK and we hung out for a week. He described his vision and I was hooked. So I moved to London. Together we founded Ineffable, began hiring a team of the best people in the world, and raised enough capital to make a major dent in the problems we’d need to solve.

One of the things that became quite clear in the early days was how good the people Dave had assembled were. I don’t just mean technically good. I mean extraordinarily creative, with good ‘taste’ (an underappreciated aspect of A+ engineers and scientists), massively productive, and with the experience of building and using large-scale systems both for research and production systems. For the first time in my professional life I was the least competent person in the room. One of the symptoms of this: I started being classified as ‘non-technical’… which was quite humbling, given the PhD in theoretical physics, more than a decade of working in RL, being the founder and CTO of the world’s first quantum computing company, etc. But everything is relative, and in this group, I was never going to be able to keep up.

I did however have an idea about one area where I might be able to contribute — the intersection of the approach to superintelligence Ineffable is building and fundamental physics. The way an Ineffable-style superintelligence would learn to do (and define) physics isn’t necessarily the same as how we do it. It seems like a very rich area to think about. To this end, I’ve launched another ‘art project’, which I am calling the Superphysicist Project. I will be using this blog to document the things I am working on and thinking about.