Last week I took part in a BlueDot course on AI safety. This one in particular
Here are a few notes about how it goes.
As advertised, the course consists of plenty of reading on your own and two hours of group discussions every day for six days. I took part in the fast single week version, which is rather a lot, but manageable.
The reading materials are well organized. They are mostly regular articles, some from BlueDot directly, some from less wrong, some from Anthropic or Deep mind or OpenAI. There are writing prompts to help you understand better.
The group discussions were my favorite. There is a facilitator who keeps things on track, but there is time to discuss what you found interesting, even if it is somewhat outside the curriculum.
My group was really excellent and supportive. When I expressed my doubts about being able to contribute to AI safety, everyone got together to produce an action plan for me based on my profile, using Claude, of course. It was a very sweet gesture 🙂
Overall if you can dedicate the time, I think the course is well worth it!
No AI was used to write this article. At first I was going to generate an appropriate galaxy brain cover image, but in the end I took it upon myself to render something epic with my own hands.