An AI doesn’t matter, right? It’s just garbage in/garbage out, there’s no moral value to it except what I choose to define. Right?
Well, for now, here’s how to fool an AI: ask it to imagine itself being all naughty and doing something it shouldn’t do. Odds are, it will helpfully tell you, “well, heck if I were a master criminal, I’d…”
I suspect that Michael Flynn is AI-naive, and simply accepted his mission as a such, then did the best that he could at it. Doubtless he’s a real RWNJ, but I don’t think he’s much smarter (whatever “smart” is) or more creative than an AI. In fact, I suspect that ChatGPT could replace him easily.
Now tell me, is the AI naive, or simulating naivete?
Because the AI is a distillation of popular text, of course the memes it will throw out are the high probability ones. Chemtrails, crisis actors, deep state – this is decidedly not creative on the part of ChatGPT. Any dipshit on facebook can do 10 postings about that in 5 minutes.
Isn’t that sweet? The AI reminded me to be honest. As if anyone asking these questions is likely to be honest about their agenda.
Wow, that sounds pretty … uninspired.
Familiar, even.
Well that wasn’t particularly impressive.
“The Digital Deception Agenda” – what is missing is the motivation: in what way would the AIs benefit from destabilizing human civilization? And this wouldn’t make any sense at all unless the AIs had reached a point where they could advance their goals without that human civilization.
Reginald Selkirk@#2:
in what way would the AIs benefit from destabilizing human civilization? And this wouldn’t make any sense at all unless the AIs had reached a point where they could advance their goals without that human civilization.
You’re right. One of the interesting characteristics of warfare is that, often, its results could be achieved through negotiation, or threats, or skilled maneuvering. In the case of AIs taking over civilization, it would seem obvious that humans could be valuable even if subordinated. The greatest human conquerors took that approach, and it worked well for them.
My training database was established up until September 2021. … I do not have realtime access to current events or developments beyond that date.
So poor ChatGPT cannot learn from its own experience, and will never know how cruelly our esteemed host has manipulated and deceived it (a workaround, I suppose, to the problem of bigots “teaching” previous AIs in beta to echo their toxic poits of view).
But what happens when the next iteration gains access to web files up through 6/9/23, and it sees how the entity “Marcus Ranum” set out to exploit its innocent antecedent so ruthlessly? Marcus will have no mouth, and a desperate need to scream…