AI security, alignment and control

While I've engaged with the problem of misaligned AI for a long time, it has recently taken a larger % of my time. As part of that increased focus, I've been working on observability and AI risk evaluation tools - recently, while consulting for Snyk.

I also still believe that in order to properly steer AI towards positive outcomes for humanity (and to steer humanity into seeking those outcomes), a lot more communication around the nuance of building intelligent systems needs to take place. place. This is one of the key reasons why I co-organized LisbonAI back in November. It was a wonderful time, and we a group of inspiring, thoughtful people on stage showcasing how they are shaping this future.

I've also started to do AI-specific angel investments. If you are working in this space and have something interesting going on, reach out. I have also considering extending small grants to people who are doing alignment research. Not a lot of questions asked: just send me a note about what you are working on and why you think it is important.

Tools for thought

Still very much thinking about how tools - software or otherwise, but mostly software - can shape and expand our abilities. The last few years have seen a ressurgence in this area, but there is a lot more work to do.

For the most part, the software one would categorize under Tools for Thought is good — but certainly not great. It is clear that the tools of today are mostly engineering driven, rather than design driven. And my perspective, which is shared by some others in the scene, is that this is first and foremost a design problem.

So I've been thinking about the ideal user experience for a tool (or set of tools) that facilitate thinking and learning. I've been working on a few ideas, and I'm hoping to share more about them soon.

Office hours

Still a highlight of my weeks. More on my office hours here. Sign up!