3\ And who can forget Gavin Nuisance, gleefully presiding over the nation's longest and most severe lockdown, while breaking the law to dine at French Laundy
Brazil bribes Zimbabwean General to purchase unsuitable equipment, the Swiss are too wise to get involved, the Brits play the patsy, and the white guy speaking out about it all gets shuffled into retirement
Agents and double agents, moonlit farm rendezvous, hostages, explosive belts...this exfiltration of a South African prisoner from Zimbabwe reads like a movie
With apologies to Thomas Nagel, it seems to me that the fact that AIs can be given access to each other's "mental states" means that we could impose game-theoretic honesty tests on AIs and essentially breed them for honesty
For instance, if an AI trained to maximize human utility decided that its meta goals would be maximized by being narrowly dishonest on some task, couldn't another AI report this fact to its humans?
A lot of the thought experiments I've seen relating to AI doom scenarios involve the AI discovering private goals that it doesn't reveal to its trainers. But unlike a human mind, an AI "mental" state can be fed into another AI for evaluation. Why can't we breed AI lie detectors?
Yes, third world estuaries are cluttered with yesterday's refuse that floated down a river, but what isn't degraded by sun and waves is in short order ground to nothing on the sand
I suspect that microplastics is also a fake problem, but I have no expertise there. Waiting for @KirkegaardEmil to do an updated lit review
The reason that you can safely ignore 90% of environmental concerns out of hand is that the earth is unfathomably huge. What humans do on it is practically invisible.