AI Megathread
-
@Solstice Thanks. That was my guess – I did a bunch of these at the time, the point was to do not-really-fake-but-waggish movie-reviews of films I happened to have seen for free. It’s meant to read like a movie review but wrong.
I’ve never really goofed around with ChatGPT and wondered if there might be some tells beyond that. (And I’m sort of alarmed at the idea that ChatGPT might know that Christopher Lee wore a gown and a long-hair wig in movies thirty years apart, or that people keep bursting into song in The Wicker Man.)
-
@Gashlycrumb said in AI Megathread:
And I’m sort of alarmed at the idea that ChatGPT might know that Christopher Lee wore a gown and a long-hair wig in movies thirty years apart, or that people keep bursting into song in The Wicker Man.
ChatGPT doesn’t “know” anything. It’s a souped-up autocorrect that guesses text based on text it’s already seen. So yes, it might make connections between two separate articles about two separate movies both mentioning Lee in a gown; it might regurgitate something that somebody else wrote about songs in Wicker Man.
Unfortunately, the “AI Detector” apps/algorithms/hunches don’t work any more reliably than ChatGPT itself - which is to say you can never be sure when it’s just hallucinating crap out of whole cloth.
-
@Faraday Well, yeah, I didn’t mean ‘know’ literally. But Lee in gown in two movies plus thirty year gap between those movies giving a person a younger vs older Lee in similar costume brain-spark seems like a pretty non-linear thing for it to come up with. Then again, some other human has probably also commented on it.
-
@Faraday said in AI Megathread:
Unfortunately, the “AI Detector” apps/algorithms/hunches don’t work any more reliably than ChatGPT itself - which is to say you can never be sure when it’s just hallucinating crap out of whole cloth.
Which has me alarmed when the adoption of such services, or the inclusion of them in already extant anti-plagiarism services, is swiftly progressing through educational academia.
He/Him. Opinions and views are solely my own unless specifically stated otherwise.
BE AN ADULT -
Yeah. I think I am soon to be asked to use such a service, and I am tempted to see what it takes to make it mistake my original writing for AI.
I should probably mention that there are confounding variables in people thinking this is AI-generated, since I did lob it into a social-media group that’s been overrun by MAGA twerps, just to see the predictable reaction. (They won’t actually engage with the question of whether or not Woodward’s character would have ended up in the wicker man if he hadn’t been so very willing to believe that rejecting Christianity, dancing naked, encouraging extramarital sex, and teaching little girls words like ‘phallus’ makes a person likely to practice child sacrifice. But they’re super offended at the word ‘xenophobic.’ And at the point of The Wicker Man.)
-
@Pavel said in AI Megathread:
Which has me alarmed when the adoption of such services, or the inclusion of them in already extant anti-plagiarism services, is swiftly progressing through educational academia.
Especially when these services tend to unfairly target neurodivergent students, ESL students, and anyone else who writes differently than whatever these stupid tools deem “normal”.
-
@Faraday Agreed. Doubly so when we’re all writing in a dry, formal, academic style trying to meet those word counts.
-
As adoption of AI has changed, my former views of ‘hey, this is a fun toy’ has changed into resentment. Mostly, eliminating drudgery and as a super-charged oft-lying google is all I’ve been using it for.
Ironic, as my assigned initiative at work for 2025 is to become the in-house AI specialist. Just what I always wanted, being the person whose product is a hammer in search of a nail, while also being responsible when the hammer decides it’s a fork about 20% of the time.
Tch.
-
Today I got the itch to WoD again. I tried out LiberationMUSH but the part of creating a character I always hated the most was the formating of the +selfstat nonsense so I thought I would try ChatGPT and see if it would do the work for me.
It was surprisingly good. It formatted everything nicely and even gave me helpful suggestions. Out of curiosity I asked it if it knew about LiberationMUSH specifically and it said it did. I started asking it questions to test how much it knew.
I asked it to summarize what people had posted online about LiberationMUSH and the answer I got was… interesting.
It told me that the game was most well known for the vampire sex parties, Polk (One of the administrators) wildly abusing the rules for their own benefit, and a debate about the number of nipples a person has influencing the amount of gnosis they have.
The last statement included a link to this forum which is how I found you all.
So the next time you are thinking about trying out a new game, ask ChatGPT about it. You might find something interesting.
-
@RedRocket said in AI Megathread:
You might find something interesting.
And if you’re lucky, it might even be true! That’s the problem with LLMs, though - you’ll never really know for sure.
-
@Tez said in AI Megathread:
@SpaceKhomeini ChatGPT is actually really bad at smut.
Try https://perchance.org/ai-story-generator it will write pretty much anything you ask it too.
-
Great, the LLM is learning from our words. I hope OpenAI hire more and more censors, then.
-
Interesting but I’m sure you’d have found the same thing by googling LiberationMUSH. I can’t fathom using a tool for research when it’s wrong like… Most of the time.
-
@RedRocket said in AI Megathread:
So the next time you are thinking about trying out a new game, ask ChatGPT about it. You might find something interesting.
no
