Helping OSS communities take back power, is AI Alignment

I have been heads-down focusing on two things recently. The first is putting together a proposal that reflects how I imagine AI alignment for open source communities can happen through collective action, without requiring sign-off from frontier labs or really any tech company. I am looking for grants to cover this work, so if you are interested (or have ideas), please reach out. I will keep slowly working on it nonetheless.

The second thing I am doing is preparing a talk for the Prompt conference in Victoria next month. While I have lived and worked in the area for a very long time, it’s been a minute since I felt part of the local tech community, and I am excited to talk about ‘Power Back to Communities’, where I will share, in more depth, some of my research, efforts, and proposals for the future.

All of my work is really about answering the ‘Aligned with WHOM?’ question in open communities, and then designing policy, tooling, and standards that can help answer ‘with the creators of the context’.

On that same note, I was able to attend a fantastic talk by author of 'the Alignment Problem' - Brian Christian at the University of Victoria this past week.

It was really comforting to see seniors, students, and everyone in between in the audience coming into this language of alignment. It’s not a new term, but it seems to be slowly seeping into society’s subconscious as people start to name ‘that thing’, that contrary thing AI seems to do, from autocorrecting our creative words to self-driving car crashes and reward hacking. It all just seemed to be clicking for more people. I find that exciting: we are getting to the human problems now. Hey, maybe people will fund human work again as well 😄

I will write another blog post soon on some of my work evaluating policies for open communities through CHAOSS metrics that we are ready to ship any day now.