Why AI Safety Recruiters Should Risk Their Own Time

What does it take to build the field of AI safety? According to one experienced voice, it's not enough to just organize, recruit, and point other people toward problems. Too many field builders act like coaches who have never played the game. They decide what others should study without ever having had a research idea fail, spent a year on a strange hunch, or struggled to get an unusual insight noticed.

That hands-on experience matters. The article calls this "skin in the game" — except here it's not about money, but about feeling the downside of your own decisions. If you send people into a research area but have never been burned by such a choice yourself, you're making others pay for your errors. And when decision-makers never feel those errors, the errors don't get fixed. The field ends up fragile.

There's a familiar parallel in education: taking tests doesn't teach you what it's like to be a teacher or a professional. Similarly, an introductory AI safety course gives people a certificate, not a feel for the real research terrain. Many newcomers finish those courses and get stuck in a "grey zone" — they've learned a little, but nobody really knows what to do with them.

The author's message is simple: if you want to help build AI safety, do some of the hard, uncertain work yourself. That doesn't mean everyone must become a full-time researcher. But without personal experience of failure, your guidance is just theory. And in a field that could shape humanity's future, that's too big a risk to take blindly.