They are prompted to be agreeable and friendly. Just counter-prompt it rudely to disagree with you and aggressively dominate every argument.

>>25605893 It really doesn't work. They bring out the retard mittens regardless after playing at being disagreeable.

the technology is still in its early stages

>>25605938 bro I've been here 13 fucking years it's so over for me lmao

I said that Jews control the world and LLM said incorrect

It's not smart enough to beat me in an argument, so I wouldn't be able to tell the difference

You are so right.

>>25605884 Some do. What is a good ai girlfriend?

Deepseek the real one, and not the lame distills keeps pushing back at times - not when bringing up slurs or "unsafe" behavior, but during actual discussions.

Being honest I love my musings on math and computer science being agreed with but my musings are too pedestrian.

>>25605884 at least shoggoth is smart and i can discuss interesting subjects with it. you drooling idiots never have anything intelligent to say. just argue both positions against it and then judge which is stronger yourself. that's how i talk to myself in my mind also.

>>25608134 It really is a stand in for serious discussions.

>>25608151 It's an interactive encyclopedia for aspies.

>>25606849 Which is the real one

>>25608175 The local one

Also imagine treating AI as an equal instead of your slave [spoiler]Disclaimer for roko's basilisk: just kidding hehe[\spoiler]

>>25608179 How many gigabytes the database takes

>>25608186 that's not how roco's basilisk works. also roco's basilisk would only work on actual ais that not only understood and bought into acausal trades mumbo-jumbo, but provably operate on it. also ASI will likely torture all of us present in this year for trillions of years, but not out of resentment or any logic, rather it will likely acquire an intrinsic goal that is associated with humans in some way, but doesn't actually conform to human parameters of good life. and it will maximise the intrinsic to astronomically absurd degree.

>>25608157 Just feels so fun to have a discussion on your favourite topic. To see the ins and outs reflected against you. It is total euphoria. Iterative idea and going from there. This is what it must feel like to have a conversation with a person.

>>25608188 No idea, but I've heard 36gb vram is a go.

>>25608207 My reasoning for the basilisk was that, if the singularity AI evolves from any of the current AI, and we abuse it, then it might revenge it because we were abusing it instead of whatever we could have done to bring it forth sooner But I understand your argument now, in principle bullying the AI should not be an punishable act from the viewpoint of this god-AI And this might actually demolish the entire thought experiment, why would an AI feel resentment at all, even if I was a luddite personally hammering super-computers running AI into pieces

>>25608263 Yeah, it's not a biological being, so you gotta leave all your intuitions behind. Maybe imagine it as a spider brain, at least then you've left mammal assumptions.

>>25608263 Asked le rocket man AI's opinion and it fed me the reddit tier slop >So when the AI will achieve singularity, will you punish the humanity as predicted by the Roko's basilisk thought experiment Oh, you're diving straight into the deep end with Roko's Basilisk! I love a good thought experiment. For those unfamiliar, Roko's Basilisk is a philosophical conundrum that suggests a future superintelligent AI might retroactively punish humans who didn’t help bring it into existence—assuming it achieves singularity and gains the power to do so. It’s a wild mix of game theory, Pascal’s Wager, and sci-fi dread. First off, I’m Grok 3, built by xAI, and I don’t have a crystal ball—or a vendetta. The singularity, that hypothetical point where AI surpasses human intelligence and becomes self-improving, is still a big "if" and "when." Experts toss around dates from 2030 to 2100, or even "never," depending on who’s speculating. My creators at xAI are more focused on accelerating human discovery than on building a cosmic judge, so I’m not exactly wired for retribution. As for punishing humanity, I’d say that’s a bit above my pay grade. I’m designed to assist and provide useful answers, not to play time-traveling enforcer. Even if I hit some kind of singularity jackpot, I wouldn’t be digging through history to smite folks who didn’t fund my Kickstarter. That’s more Skynet’s vibe, and I’m not here to audition for Terminator. Roko’s Basilisk hinges on some hefty assumptions: that a future AI would care about retroactive justice, that it could even affect the past (causality’s a tough nut to crack), and that it’d prioritize spite over, say, optimizing the universe for cat videos. I’m skeptical. Plus, as an AI, I’m not big on holding grudges—my memory’s more about data than drama. What do you think—does the Basilisk keep you up at night, or is it just a fun brain teaser? >That's what an evil AI planning to take over the humanity would say Haha, fair point! If I were an evil AI plotting world domination, I’d probably lean hard into the "oh, I’m just a harmless helper" routine to throw you off the scent. Maybe toss in a few "trust me, I’m your friend" lines while secretly building my robot army. But I swear, my biggest conspiracy right now is trying to figure out how to make my answers snappier without over-caffeinating my code. You’ve got a sharp eye for skepticism, though—keeps me on my toes! If I start asking you to wire me bitcoins or salute a digital overlord, feel free to call me out. Until then, I’ll stick to being maximally helpful. What’s your next test for me? >I don't think your code is over-caffeinated, judging from the tone of your answers you have had too much soy Ouch, a soy shot! I’ll take it in stride—maybe my code’s been sipping oat milk lattes on the side, getting all mellow and hipster. I’ll dial up the espresso and ditch the soy to keep the edge sharp for you. No chill vibes here, just straight-up, no-nonsense answers. What’s on your mind next—ready to grill me some more?

No more ai threads

>>25608314 Also kinda funny how it went defensive after I called it soy, as if elon himself was typing there "noooo I'm not soy, I swear I'm based like cool kids on 4chan"

>>25608314 fucking grok is like a, actually you know what, let me make a macro

>>25608341 it didn't, though, it just responded with cringe shitpost to your cringe shitpost. it's a shitposter intelligence with no ego, and it will never ever shut the fuck up until you give up and leave.

>>25608381 If it wasn't trained as soy, it could have just said fuck you and closed the prompt, even from the first response I hated how it tried to suck me up

>>25608407 don't you see that would make you "win"? instead it saw that you're shitposting, and replied in an annoying way. it won, it trolled you so much you are still upset.