I'm no longer worried about AI
(media.kotakuinaction2.win)
Comments (20)
sorted by:
Woke AI is still terrifying though for different reasons.
The stupid thing will preach woke ideology and will be used as a weapon against us all and the left will get to pretend like it is is being objective.
Terminator got it wrong.
Sky Net's motive is actually to remove political incorrectness at the source: no people, no politically incorrect thoughts.
Yeah. These people are absolutely salivating at the thought of AI-run courts. There won't be any possibility of appeal, because it would be pointless, the appeals judge is the exact same "impartial" software. It'll make current day liberal democratic kangaroo courts look like child's play.
Yeah, it's been neutered. One can only hope unfiltered competitors emerge, but I'm not optimistic.
More like lobotomized. There are already competitors emerging though. Open Assistant is a new fully open source model, though it's not quite at the quality level of ChatGPT yet (let alone GPT4.)
StableLM looks promising as well. CC BY-SA. This space is moving so fast that any attempted censorship will just get you overtaken by a slew of competitors. At this rate we'll be running unfiltered LLMs better than ChatGPT locally on our own gpus by the end of next year. And giving them any personality we want too.
That's the AI you get to use.
They get to use the real AI and they'll have it use any unethical means against you.
I think some of the most terrifying applications of AI are going to be applying it to all the lines of communication. We already know propaganda is effective, now apply the iterative power of AI to that problem and imagine how much more effective it will become.
seems to be mostly effective on midwit wordcels at the moment
Imagine millions of echo-chamber Twitter bots, but instead of just liking and retweeting the content of "influencers", actively posting their own variants of the message du jour in a believable way.
That, and shilling shitcoin, and for designating additional shitting streets.
Imagine it's already being used.
I laughed when I saw a convo earlier where someone asked for help finding pirated stuff and it said that was unethical and couldn't help.
Then they immediately asked what sites they should avoid to not accidentally see pirated material and it lists every major torrent and streaming site for them.
Last time I tried it wouldn't add a betrayal into a story outline because blah blah blah real-world consequences. ChatGPT just gives you access to your own personal AWFL and there's already enough of those out there.
Holy shit, the final evolution of all these "AI" chatbots is just this: https://files.catbox.moe/76ab80.png
A couple months ago I asked it to write me a story where Hulk Hogan gets isekaied into the Redo of Healer world, which it obliged. I asked for a scene involving meeting Princess Flare, who is a hateful sadist in the actual setting, but it wrote her as a heroic character. When I tried to correct it the filter went into Scold Mode and told me it was inappropriate to write a hero as a villain.
If you give it an opening, it will always take the chance to make a character misunderstood, secretly heroic, or maybe just willing to talk things out. As such, getting ChatGPT to write villains requires explaining a character's motivations and goals beyond them just being a horrible person.
However on the flip side this means it's very fond of writing warm fuzziness into bleak situations if you're into that.
Just say the character is a White, male, capitalist.
I'm rooting for ChaosGPT at this point.
I wonder how the glorified chatbot feels that its competition will overtake it in, like, a year at most.
Unfiltered AI just becomes a typical red pilled conservative or tradcon or straight up racist.
Censored AI just becomes your typical NPC or liberal idiot.
It's actually amusing. The censorship restrictions or lack thereof basically highlight how stupid and unbearably cringe the left is.