Some thoughts as someone running an AI safety university group and a pre-Kairos university group advising initiative, familiar with Kairos’s current programs (I do not work at Kairos):
Kairos’s team seems diligently focused on finding and training good people to work on AI safety for impact, rather than just “getting more people to output mech interp papers”. This includes helping people make better judgements about how their work actually impacts the world. I.e., answering questions like “Should you work on safety at the labs?” They also strongly screen applicants to their programs to filter for people who actually care about making AI go well. This is difficult to do perfectly, and the amount/quality of filtering probably differs for the type of program and its target audience, but I think they do a pretty great job.
Being imperfect is fine. AI safety is more neglected than capabilities, which means, for example, inadvertently training one capabilities researcher in exchange for training one safety researcher is likely still worth it (I think their ratio is much better than this).
Kairos’s focus isn’t limited to technical AI safety; their support helps people do policy and political work too. Among a few other things they support/run, university groups cover a range of different sub-fields and areas in AI safety, and the Generator Residency has trained generalists and led to the Longitude DC Intensive. Though they’re limited in some ways as a 501(c)3, making people more knowledgeable and well-connected to the AI safety space makes them more effective at any safety work they do.
I’m a college student who runs an AI safety university group (at UC Santa Barbara) and largely sees similar things among students I interact with. Is there an AI safety student group at this university? Like the ones at Georgia Tech or Wisconsin-Madison? Are they moderately active / do they run fellowships? Have you spoken with them? Mostly just curious (and would recommend that you do talk to them if you haven’t)