: “We are clear that human political control of our nuclear weapons will be maintained at all times, and we strongly encourage other nuclear states to make a similar commitment.”
Professor Michael Osborne, Dyson professor in machine learning at the University of Oxford, does not believe AI should be “anywhere near nuclear weapons”.: “The realistic concern is that AI could be used in some sort of early warning system, the detection of incoming attacks. Given the many failure modes of existing AI, it’s not too hard to imagine that false warnings could be provided that might lead to really harmful escalation.
“You can imagine a kind of automated defensive system on the border between North and South Korea, and a bird flaps its wings in a way that AI misidentifies. You may not have heard of Vasili Arkhipov but, for many, he is the most important figure in modern history. That day US forces dropped non-lethal depth charges in the waters around Arkhipov’s submarine in an attempt to encourage the boat to surface. However, the crew of B-59 had been in a communications blackout and thought they in the middle of an outbreak of war.
Fortunately for the world not just two, but three senior officers had to agree to launch the most destructive of weapons. Arkhipov refused to sanction the launch of the torpedo, calmed Savitsky down, and it was never fired.
Accidental armageddon would be a bummer
Skynet