Is this necessarily correlated? I would think that the potential to cause disaster is rather linked to whether AI is given authority to enact autonomously decisions, i.e. given access to critical systems.
if there's no benefit, AI's net cost is more than traditional systems. no benefits, no access.
Yeah, it's flat out a bad take. If the AI is utterly incapable of doing anything useful, it would be utterly incapable of being an existential risk.
i was thinking more if the current crop of AI are useless, they won't get enough money pumped into them to lead to runaway amplification. if there's genuine risk, then there will be rewards. if we're close to that kind of AI, then we can't afford for our enemies to find it first.