The Government Knows AGI is Coming
About this episode
The episode's premise: people inside AI labs and government now believe transformative AI — systems capable of most cognitive work a human can do — will arrive in two to three years, during Trump's term. Buchanan, Biden's AI adviser, explains why the U.S. government treats beating China as a national-security imperative: AI advantages in cyber offense and defense, intelligence analysis (e.g., satellite imagery), and hacking. He downplays the DeepSeek shock — export controls are working (DeepSeek's CEO called chips their biggest constraint) and should be tightened. He rebuts Marc Andreessen's claim that Biden sought regulatory capture, defends the voluntary AI Safety Institute, and rejects Vance's safety-vs-acceleration dichotomy (railroad safety standards unleashed railroads). On labor, he concedes disruption will be severe and uneven, defends the dignity of work over UBI, and admits the government lacks real plans — while rejecting the theory that DOGE's dismantling of agencies is somehow AI preparation.