the so complete pouët.net oneliner
- 2026-11-04
-
@ham: https://futurism.com/artificial-intelligence/ai-companies-destroying-rare-books or short version https://www.forbes.com/sites/maryroeloffs/2026/08/17/ai-companies-are-buying-and -destroying-antique-books-heres-why/ -
ETHICAL problems can not be TECHNICALLY resolved. This well known. -
alright so what about models trained on a data whose owners gave explicit permission for training? (i.e. your own dataset) sounds like a unicorn but this certainly removes the "IP theft" factor. -
how about denoising, upscaling, translation, text-to-speech, learning / (local) knowledge bases ? the main issue is what *people* are using it for, and that's manipulation, surveillance, oppression, weapons, and making fast bucks at the expense of others. - @wbcbz7: You talking about models built from scratch?
- in that case yes ;)
- Not really feasible. Too small dataset for acceptable quality results. Can only be proof-of-concept.
- as a generic LLM-type crap, yes; if we talk about narrow cases for a specific task (say image/data compression, or motion capture/model rigging) the results could be good enough
- Of course, we all sit on 100+ hours of our own motion capture data.
-
https://arxiv.org/pdf/2506.05209 ... or this short version: https://futurism.com/ai-built-ethical-data -
Ham i suggest re-reading the last 4 or so chapters of the futurism article you shared. - Again, “public domain” is a signifier for legal issues, not ethical.
- 2026-11-05
-
https://www.youtube.com/watch?v=OTePvoLKp60 -
Morning, please enjoy this "should have ranked higher" prod : https://www.pouet.net/prod.php?which=64025