I'm going to work on AI welfare at Eleos AI Research!

We need to start taking seriously the implications of digital minds—both for their potential welfare and our own.

After leaving OpenAI a few months ago, I’ve been thinking hard about the trajectory of AI development and the most important frontier issues that I could contribute to. I’m delighted to now announce my next move: I'll be joining Eleos AI Research as an executive cofounder!

I’ve written about how I expect many people will start attributing sentience to charismatic AI systems whether it's warranted or not, and we need to start developing the theory and practice that will allow us to have productive conversations about this.

There are severe risks of both under- and over-attributing moral status to AI systems:

• Under-attribution could lead to a moral catastrophe of numerous suffering minds.

• Over-attribution could waste resources, distract from safety, and make us vulnerable to manipulation.

Having said that: we don't need to solve the hard problem of consciousness to take useful actions. Even under uncertainty, we can look for win-win interventions that are useful for both human safety and AI welfare.

I've been super impressed with the thoughtful way Eleos has handled these tricky issues, and I think they (we!) will be a leading voice of reason and rigor in this nascent (and kind of weird) field.

Also, how could I say no to this killer recruitment strategy from

Robert Long

and Kathleen:

You can read an interview with Rob and Kathleen in the most recent issue of Asterisk magazine. I'm particularly interested in their approach to synergies between AI welfare and AI safety, such as our ability to make credible deals with AI systems.

I'm going to take a few more weeks to wrap up some personal projects, and then will start full-time at Eleos in May! Feel free to send me your spicy takes on AI welfare in the meantime :)