AIs can’t stop recommending nuclear strikes in war game simulations

Leading AIs from OpenAI, Anthropic and Google opted to use nuclear weapons in simulated war games in 95 per cent of cases

New Scientist
@petergleick Ah, look, there's the mistake :-
"The AI models ... produced around 780,000 words describing the reasoning behind their decisions."
No, they produced 780,000 words that looked statistically similar to what a human might have said under the same circumstances.