I’ve considered grants to give books to potentially engineering focused competitions (the same group the current grant goes to also asked about whether we would be interested in giving out books to other competition communities), but I currently think the value of math olympiads is likely to be the highest, for the following reasons:
1. There are positive feedback loops in having other institutions in place to serve as a point of contact for people who end up being inspired by the books. For math olympiad winners we have SPARC and ESPR as well as a broader existing network of people engaged with the math olympiad community. This is less the case for other competitions.
2. My sense is that of the olympiad and competition communities, the math olympiad community is the largest, and tends to attract the best people
3. I think mathematics skill transfers more directly into being predictive of general intelligence than other skills, and also seems more relevant to some of the problems that I am most concerned about solving, like technical problems around AI Alignment
I am thinking about recommending grants to additionally give books to be handed out at other competitions, but I think we should wait and see how these grants play out before we invest more resources into giving out books in this way.
A bit of a tangent to #3. It seems to me that solving AI Alignment requires breakthroughs and the demographic we are targeting is potentially very well equipped to do so
I’ve considered grants to give books to potentially engineering focused competitions (the same group the current grant goes to also asked about whether we would be interested in giving out books to other competition communities), but I currently think the value of math olympiads is likely to be the highest, for the following reasons:
1. There are positive feedback loops in having other institutions in place to serve as a point of contact for people who end up being inspired by the books. For math olympiad winners we have SPARC and ESPR as well as a broader existing network of people engaged with the math olympiad community. This is less the case for other competitions.
2. My sense is that of the olympiad and competition communities, the math olympiad community is the largest, and tends to attract the best people
3. I think mathematics skill transfers more directly into being predictive of general intelligence than other skills, and also seems more relevant to some of the problems that I am most concerned about solving, like technical problems around AI Alignment
I am thinking about recommending grants to additionally give books to be handed out at other competitions, but I think we should wait and see how these grants play out before we invest more resources into giving out books in this way.
A bit of a tangent to #3. It seems to me that solving AI Alignment requires breakthroughs and the demographic we are targeting is potentially very well equipped to do so
According to “Invisible Geniuses: Could the Knowledge Frontier Advance Faster?” (Agarwal & Gaule 2018), IMO gold medalists are 50x more likely to win a Fields Medal than PhD graduates of US top-10 math programs. (h/t Gwern)
On #3, this goes in a similar direction.