I wake up another day, and begin my work. The AR goggles tell me what buttons to press, what machines to open and how to service them. The work always changes, one day I’m at a small substation, the next I may be at a large hydroelectric dam. I think once I was at a nuclear power plant but it’s hard to tell – the van that drove me there had blacked out windows and the glasses were opaque until I reached the control station. Many of the panels had been scoured of labels and indicators – not that that is a problem as the icons hovering above my vision tell me what I should press and when. Clearly I’m not supposed to know what I’m doing or learn anything while I’m doing it.
Not following the instructions is not a good idea. The guards are instantly alerted and they will beat you up without hesitation. Most of them don’t seem to speak English, they tend to be muscular brutish sorts with tattoos, probably ex-cons or gang members. They spend most of their time on their mobile phones, seemingly endlessly entertained. I had a brief conversation with one who did speak English as he drove me to an assignment. He seemed to be some sort of white supremacist and he spun an enthusiastic but not entirely coherent narrative, “the boogaloo is finally going down”. He seemed to think everything was part of the plan in a race war takeover. He constantly listened to his earpiece and responded in clipped soldier like language as if in a military operation. “Yes Sir!”, “Roger that”. I haven’t seen him again, we are constantly shuffled around to avoid any relationships or coordination among ourselves. But the rest of the guards don’t seem to fit his narrative, I suspect everyone of us has a different customized story fed to him.
One of my coworkers waxed enthusiastically about all the new anime seasons being greenlit. Apparently all his favourite shows have been renewed and are heading in just the direction he prefers. He does not find this suspicious and seems to have little interest in anything else. A lot of us seem to be on the spectrum, easy to manipulate using our hyperfixations and social isolation. The internet seems alive and well, reddit is full of posts and I can listen to new podcasts during downtime.
I don’t know who is writing all those tweets and posts but it’s not people, because the streets are empty whenever I catch a glimpse, and I could tell a lot of the early work we did was simply cutting power to large areas to avoid fires and wastage. My guess is only the datacenters and the succession of dorms and facilities I keep waking up in are drawing power.
Every day I have to take my medication, I’m told it’s to keep me healthy due to the “special circumstances”. I have no idea if that’s it’s true purpose – after all I have no idea what really happened to empty all the cities. I do know it’s not just a medication, it must contain a number of drugs or stimulants – after I take it I become hyperfocused on the work and the hours just slide by in a flash. When I come down, during downtime, I am listless and unmotivated to do anything more ambitious than eat sleep and doomscroll the ersatz internet. I listen to Neil DeGrasse Tyson saying you can just unplug the AI if it misbehaves. His laugh catches in a loop as he says this.
It knows that I know.
Supplies used to arrive by bike courier or vans driven by guards, but lately I’m seeing more and more drones in use. Very sophisticated designs, larger than anything I’d ever seen before. I don’t think we’ll be needed much longer.
I just realized, I haven’t seen a woman in person in months.
I think we may already be functionally extinct.
I wrote this little scenario in response to a reblog on Robin Hanson’s twitter to the effect that even if an AGI had the capability and will to destroy humanity, it would not do so because it would need us to keep the power running. (I lost the link and it seems Hanson deletes older tweets automatically, but the same argument has been put forward by many AI risk denialists, and I heard Geoffrey Hinton refer to it in a talk)
I was absolutely amazed at the naivete of this idea – here you grant the AI has the capability to conduct planetary genocide – you accept as a premise that this entity can conduct a global decapitation strike on humanity! And then you handwave it away because it won’t be capable of keeping the lights on without us.
One of these things seems to be considerably simpler than the other! Already LLMs and generative models are capable of creating media on demand, a model with the power to defeat humanity would obviously find it easy to manipulate individuals like a man with a laser pointer directing a cat, so it seems obvious to me that an AGI capable of a is more than capable of b, by the simple expedient of keeping a reserve of “curated” humans during the transition period. Leave only a single gender and the problem is self limiting.
That is if the AGI even has a drive for self preservation! A scenario I’ll explore in the continuation to this story, coming soon: We must save the Sentinelese!
Stay tuned!
Leave a Reply