AI Trained on 4Chan Becomes ‘Hate Speech Machine’

https://www.vice.com/en/article/7k8zwx/ai-trained-on-4chan-becomes-hate-speech-machine As algorithms improve, new analog ASIC chips specially made for training AI rapidly develop, etc, the internet will enter a new era- one in which the automated detection of SPAM by central authorities is significantly hampered. Any model of social media which does not lend some power over curation over to its users, use varying degrees of reputation, or other models which distribute curation are fated to fall. What does lain think? Also, "ethics". I don't see any ethical issue with training an AI on publicly-available data, nor in open-sourcing a model, seeing as stopping technological advance in the long term is inevitable- thinking otherwise is surely delusional roleplaying as the arbiters of morality, no? What does Lain think?

> The bot, which Kilcher called GPT-4chan, “the most horrible model on the internet”—a reference to GPT-3, a language model developed by Open AI that uses deep learning to produce text—was shockingly effective and replicated the tone and feel of 4chan posts. “The model was good in a terrible sense,” Klicher said in a video about the project. “It perfectly encapsulated the mix of offensiveness, nihilism, trolling, and deep distrust of any information whatsoever that permeates most posts on /pol.” (and because I think this image is funny)

>>60610 This is far from news but it's an opportunity to talk about stuff. Chans had always been the center of most mainstream internet movement. >chad meme is mentioned one a week by most male "streamers" and "influencers" >most of "normans" memes originally were chan soykaf (pepe, *jak, troll) >"memetic" soykaf from military to paranormal crap >AI (MS' Tay for example had innate hate-speech pre-learned among other AI which had used 4chan archives) There is no other place better for training AI speech than publicly available 4chan/imageboard archives. Although they went all out with the fake posters (AI/bots) training "on-the-site" recently around 2019 (conversations on big boards at 4chan felt "fake"). I think it's fine if they use those to train their AI but recently even stuff happening around imageboards eventually become the norm around mainstream sites (memes leaking out) so I'm predicting that these same people or mainstream site workers are modeling trends based on imageboard "wisdom" https://en.wikipedia.org/wiki/Wisdom_of_the_crowd#Higher-dimensional_problems_and_modeling

>>60612 forgot to clarify >I think it's fine if they use those to train their AI It's fine as long as they don't make their AI meddle/post along with human conversation. This creates a feedback loop which only degrades the entire purpose of the site and the trainer, anyone who does so (use AI to create posts on imageboards) are either stupid or seeking to destroy imageboard interconnections of human mind (most of these types are those against hate speech, I'll just remind them that destroying imageboards is not the effective solution). You cannot target a godform egregore. The egregore targets [you].

we already had this thread https://lainchan.org/%CE%A9/res/59237.html

>>60610 based as fuarrrk man...

>www.vice.com >What does Lain think? I think you posted a link to a notoriously dishonest clickbait rag and I didn't click it.