For years I had a blindspot in my thinking. I dismissed policy and advocacy work as “in another [lower-status] category” because I’m obviously a technical person![1] Even when I felt super pessimistic about my ability to do research that meaningfully contributes to addressing the threat of extinction from AI.
I’m not the type of person who would be “good at comms”. I’m an autistic math guy.[2] Luckily comms has learnable skills, and I have gotten better through practice.[3]
I’m now prioritizing advocacy, which currently seems a lot more promising than research.[4]
I got lucky because, thanks to Coxon, AI extinction became mainstream shortly after I started. Even so, I've been very surprised to have become the biggest anti-extinction expert voice in Finland as a young person.[5] This happened because:
So far, most notably, I…
My message is that human extinction is very likely if superintelligence is developed on the current trajectory, and I want Finland to push for an international prohibition on its development.
Some technical people doing comms have had a big impact: e.g. Hinton, Bengio, Yudkosky, Soares. Samuel Buteau single-handedly made the Canadian cross-party ControlAI statement happen. But you don't need to be as high-status as them.
I've had a surprising amount of impact in speaking out as someone in AI safety research.[9]
If you think you're not the type of person for comms, please actually consider it.
I think a lot of people motivated to help with the extinction threat also dismiss comms without good reasons. I was motivated to write this post because of, for example, these posts that discuss prioritization (when it's difficult to have positive impact in technical AI safety research) and use an appendix/footnote to scope out discussion of comms. If it's difficult to have positive impact in technical AI safety, speaking out / doing advocacy is often a much better way to have impact.
I graduated undergrad in math and computer science in one year because math was the easiest subject. I've had a ~technical AI safety research career for ~3 years.
Thank you Camille and others who have helped with my practice. I did media training sessions with different coaches and practiced with ChatGPT voice mode, 30 min – 3 h every day over a couple of weeks. Thank you Eliezer and Nate for the advice on communicating honestly and transparently instead of only strategically.
It seems to me that we definitely need more time and are not on track to solve alignment even with more investment. I imagine a lot of AI safety researchers also think we need more time.
Not because I’m young. They contacted me because I have credible expertise and I disagree with the academics and take extinction seriously compared to other Finnish experts available for comment. (They found out about me through my op-ed.)
I did talk to some other concerned experts, and having heard their stated reasons not to, I feel confident that speaking up is good in expectation.
In the article, I gave a brief scenario ending with superintelligence boiling the oceans. There were some tweets about how implausible this is; one with 100k+ views criticizing the newspaper for failing to “at least superficially” fact-check the claims, another calling me a “DEI hire”, and another saying I should become a housewife. A Finnish-language tweet with 100k+ views is roughly as rare as 10M+ views on an English-language tweet, based on adjusting down from Claude estimates.
I declined because I didn’t expect fair coverage.
I think non-technical people can have big impact through comms, especially talking to politicians, see How a cold email got the Finnish government to respond on superintelligence regulation. Following the tradition (see footnote 1), I'll say in a footnote that this post is aimed at people in AI safety research. Because the post is based on my experience as a technical person. :P
