It seems like I see this sort of thing (people being evasive or euphemistic when discussing AI risk) a LOT. Even discussions by the AI2027 authors, who specifically describe a blow by blow extinction scenario that ends with AI killing 99% of humans with a bioweapon and then cleaning up the survivors with drones, frequently refer to it with sanitized language like “it could move against humans” [1] or “end up taking over” [2]. Guys you literally wrote a month by month extinction scenario with receipts! No need to protect our feelings here!
It seems like I see this sort of thing (people being evasive or euphemistic when discussing AI risk) a LOT. Even discussions by the AI2027 authors, who specifically describe a blow by blow extinction scenario that ends with AI killing 99% of humans with a bioweapon and then cleaning up the survivors with drones, frequently refer to it with sanitized language like “it could move against humans” [1] or “end up taking over” [2]. Guys you literally wrote a month by month extinction scenario with receipts! No need to protect our feelings here!
https://open.substack.com/pub/astralcodexten/p/introducing-ai-2027?utm_source=share&utm_medium=android&r=qmni)
https://www.dwarkesh.com/p/scott-daniel