Can the U.S. and China Deny AI?

According to LessWrong, a new analysis tackles a question strategists have mostly hand-waved: could the United States and China actually bomb each other's path to superintelligence? The author, Felix Choussat, built a quantitative model of what kinetic strikes on AI compute would accomplish. The verdict is sobering for anyone banking on sabotage. Even extensive attacks on data centers and chips would only set a rival back by roughly one to five years, and much of that delay can be clawed back by nationalizing whatever compute survives. That undercuts a popular idea, drawn from the book Superintelligence Strategy, that the mere threat of blowing up AI infrastructure could enforce a lasting standoff, a kind of Cold War stalemate for the algorithm age. If a strike buys only a few years rather than a permanent halt, the piece argues, leaning on it to freeze development indefinitely would be both risky and counterproductive. The suggested alternative is to study lower forms of deterrence that don't burn the diplomatic runway needed to eventually negotiate real limits on AI.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton