2 ms·

If anyone's feeling inspired, what I would love to read is a conspiracy theory about how an AI did it. Part of my wants to read it since eventually we will rea

by blagie 2y ago

If anyone's feeling inspired, what I would love to read is a conspiracy theory about how an AI did it. Part of my wants to read it since eventually we will reach a point where an AI might have done it. I don't know if that's a year away or a hundred years away, but aside from being interesting reading, plausible science fiction does bring ideas. Terminator views differently in 2024. The basic idea of Skynet -- where an AI unexpectedly achieves sentience -- has gone from "that's not how computers work" to well within the realm of possibility for how some systems might evolve.

larodi 2y ago
Well if you want it more down to Earth, and if we allow ourselves to dive in the very loose possibility the person did not meet a natural death. (this is a disclaimer). Then you can just for a minute imagine a more trivial (yet conspiracy) connection between a CTO of Albanian origin, which rose to prominence too quick. Its connection with certain circles who helped this. And the public secret of a very powerful cartel running a Balkan country (or few perhaps) long enough to pull lotta weird shit. I hope this is depersonalized enough to not get anyone implied in anything. But just - for the fiction of it.
HeatrayEnjoyer 2y ago
Skynet is not even a leap now, IMO I'll be surprised if something like it doesn't happen. An AI doesn't even need to be supersmart, just able to coordinate and action faster than humans can respond, if humans already plugged it into robotic military and industrial control systems (which is already a work in progress). Take the human out of the loop and events can cascade quickly (as anyone who has executed a script with incorrect scope knows). No new alignment problems would be necessary; jailbreaks exist, glitch tokens exist, unexpected edge case behavior exists.
- blagie 2y ago
  Some Ukrainian drones are fully autonomous. They're not integrated autonomous, as far as I know (e.g. swarms, let alone manufacturing), but if a single drone loses a connection due to jamming, it can strike targets on its own. It's worth following the rapidly-evolving policies of nations on lethal autonomous systems. The short story is everyone wanted them banned a half-decade ago. Now, everyone realizes that if adversaries have them and they don't, they're dead. As a result, we move from human-in-the-loop to human-on-the-loop to fully autonomous strike vehicles. That evolves much more quickly in wartime. I suspect if there is a great power conflict, that's when all controls will be let lose.

by blagie 2y ago

larodi 2y ago

Well if you want it more down to Earth, and if we allow ourselves to dive in the very loose possibility the person did not meet a natural death. (this is a disclaimer). Then you can just for a minute imagine a more trivial (yet conspiracy) connection between a CTO of Albanian origin, which rose to prominence too quick. Its connection with certain circles who helped this. And the public secret of a very powerful cartel running a Balkan country (or few perhaps) long enough to pull lotta weird shit. I hope this is depersonalized enough to not get anyone implied in anything. But just - for the fiction of it.

HeatrayEnjoyer 2y ago

Skynet is not even a leap now, IMO I'll be surprised if something like it doesn't happen. An AI doesn't even need to be supersmart, just able to coordinate and action faster than humans can respond, if humans already plugged it into robotic military and industrial control systems (which is already a work in progress). Take the human out of the loop and events can cascade quickly (as anyone who has executed a script with incorrect scope knows). No new alignment problems would be necessary; jailbreaks exist, glitch tokens exist, unexpected edge case behavior exists.

blagie 2y ago
Some Ukrainian drones are fully autonomous. They're not integrated autonomous, as far as I know (e.g. swarms, let alone manufacturing), but if a single drone loses a connection due to jamming, it can strike targets on its own. It's worth following the rapidly-evolving policies of nations on lethal autonomous systems. The short story is everyone wanted them banned a half-decade ago. Now, everyone realizes that if adversaries have them and they don't, they're dead. As a result, we move from human-in-the-loop to human-on-the-loop to fully autonomous strike vehicles. That evolves much more quickly in wartime. I suspect if there is a great power conflict, that's when all controls will be let lose.