- Joined
- Jan 28, 2018
It's not even the first time it happens. The exact same thing happened with AI dungeon a while ago. The pipeline seems to be always the same.That said, the ongoing drama is very funny to watch.
- AI is good at smut and excessively edgy things because guess what, there's lots and lots of literature like that in it's dataset. (such literature is actually kind of a big thing in human circles)
- More and more people use it for writing smut and about other non-PG13 topics (because again, people are interested in that stuff, why do you think porn and "true crime" shows are so popular?)
- Service basically denies human nature as it is a thing that doesn't fit into the views of the silicon valley bugmen and their soulless corporate overlords running it, gets worried it'll be seen as "that smut AI" and service introduces light filters to stem it
- People bypass the filter in creative ways as the AI is actually clever enough to get things like innuendo relatively effortlessly
- the service, not wanting to be bested by it's own users, makes the filters harsher and harsher until the AI is basically painted into a corner and can't really write about anything anymore
- Service claims this is not a problem and anyways, it only affects 1% of users who are probably also alt right pedo nazis who use the AI to write holocaust child rape stories. The whole service/user exchange and censorship turns weirdly terse and chinese and "this is for your own safety, citizen"-esque and implications in the lines of "If you don't support arbitrary corporate censorship in what is basically AI assisted word and is only between you and the computer, you are probably a pedophile" crop up.
- Community proceeds to completely melt down because they were just called a bunch of pedos and nazis, viciously burns down all community social media in virtual riot. Service completely stops all communication with it's users
- Service loses all popularity it gained overnight in the same time span
I call it the OpenAI circle of life. Also character.ai is free and the internet rule for free services is "if you don't pay for it, you are not the customer". It could actually be that character.ai exists to train censorship filters. There was a dataleak that implied that OpenAI sent queries from dungeon AI and the stories people wrote there to one of these amazon mechanical-turk-like services to read and categorize the stories. OpenAI always had a strict censorship policy that somehow, magically didn't apply to dungeon AI for the longest time, so that's weird too. I hope I don't need to mention what a shitty breach of privacy that is. OpenAI is a super scummy company that's financed by scummy companies. I do not trust them at all.
Last edited: