They Classified Math in 1940s | MOONSHOTS

The discussion begins with the fundamental observation that AI is essentially mathematics, which means that regulating AI models is equivalent to regulating mathematical concepts themselves. This connects to a conversation with Jeff Sites and Ben Buchanan, who was responsible for AI policy and authored the Biden executive order. Buchanan acknowledged that such mathematical regulation is possible, citing the precedent of nuclear physics classification in the 1940s, some of which remains classified today. The speaker then reflects on the broader implications of this approach to scientific knowledge, noting a striking pattern in physics progress. They observe that significant advances occurred up until the Einstein and John von Neumann era, but progress appears to have slowed considerably since then. This raises important questions about whether the classification and restriction of certain mathematical and physical knowledge may have inadvertently hindered scientific advancement. The speaker expresses hope that AI might help unlock solutions to current problems, while also wondering whether previously classified knowledge could hold keys to solving contemporary scientific challenges.