I think AI is no different than most things in life: Garbage in, Garbage out. This is true of the training data, the end users input at prompts, and any attempts to filter or moderate. (also for factual or moral data).
Time would probably be better spent letting people know the output is AI (watermarks/metadata) than trying to filter it or worry about whether its right or wrong.
Are Large Language Models Really AI?
|
Are large language models really AI?
Asura.Saevel said: » Dumb because anyone can download stable diffusion and render their own highly illegal material. Yup - Reactor a face type software tried to make a filter that senses NSFW and then someones like well... if I use Forge SD I can buypass that.. Then they make a Lora file of the person, and it's all downhill from there... You guys are playing too much in the dark AI.
 RadialArcana said: » AI is not AGI and it does not think for itself as you or I do RadialArcana said: » It's trained on tainted data from the internet that is moderated with a left wing bias. Its trained on everything its creators can get their hands on. Books, magazines, scholarly dissertations, movie scrips, radio transcriptions. If the library of congress were all online they would have used the whole darn thing. They did use all of it that is. Asura.Saevel said: » Dumb because anyone can download stable diffusion and render their own highly illegal material. Asura.Saevel said: » I would like to stay clear of anything political. Garuda.Chanti said: » Asura.Saevel said: » Dumb because anyone can download stable diffusion and render their own highly illegal material. Almost anyone can download. Not near anyone would know what to download (I had never heard of stable diffusion.) Far fewer know how to render anything. Grok and other AIs have democratized deep fake porn to where any mouth breathing 8th grade dropout can create it. Super true - I know tons of people with 4080s and 5080s who don't touch the stuff - think it's stupid - others just don't care or don't want to be bothered - then you have people with 4060 8gb using guff models to make videos and art now Ive been scratching with AI for a long time now, no expert or anything, but comfyui works along with getting someones cutting edge workflow to work, and it's getting pretty crazy what I am able to do now. Whenever I want to update my setup, I run my start-up log to Gemini or grok, and it looks online for the latest updates from GitHub, reviews forum posts to ensure its geting the best info, and then I briefly tell the AI to ensure it makes no errors so it triple-checks before telling me how to proceed RadialArcana said: It's trained on tainted data from the internet that is moderated with a left wing bias. Garuda.Chanti said: » Its trained on everything its creators can get their hands on. Books, magazines, scholarly dissertations, movie scrips, radio transcriptions. If the library of congress were all online they would have used the whole darn thing. They did use all of it that is. These two things aren't contradictory. Pretty much all media that's online has a moderation that is left of Radial's politics. I'm not going to argue about where the overton window is right now or what qualifies as left/right, but scrapable data is almost certainly further left than fully uncensored conversation.
[+]
|
All FFXIV and FFXI content and images © 2002-2026 SQUARE ENIX CO., LTD.
FINAL FANTASY is a registered trademark of Square Enix Co., Ltd.