Post Ar7hnO2teoTz8cxOYS by quinn@social.circl.lu
(DIR) More posts by quinn@social.circl.lu
(DIR) Post #Ar7hnIcpoHfGKLj0jo by quinn@social.circl.lu
2025-02-14T12:57:11Z
0 likes, 0 repeats
What We Talk About When We Talk About AI https://www.emptywheel.net/2025/02/14/what-we-talk-about-when-we-talk-about-ai-part-one/I've published the first part of an explainer on artificial intelligence I've had going in my mind for a while.Would love to hear what you think of it, and if you like it, share it!
(DIR) Post #Ar7hnJGtPLhuKaT0Ay by ManiacalV@mastodon.social
2025-02-14T13:39:50Z
0 likes, 0 repeats
@quinn You don't say anything about the collection process other than it's "difficult and expensive."No disrespect, but I am very against these tools because I choose consent over theft, and I respect artists, writers, and musicians too much to boldly blog about the systems trying to replace them (after stealing their work) without even mentioning them at all.I'd much rather read about them and how AI effects their lives.
(DIR) Post #Ar7hnJxQrBjcSWMyTw by quinn@social.circl.lu
2025-02-14T16:34:06Z
0 likes, 0 repeats
@ManiacalV I know everyone has their particular thing they want to call out about AI as fast and loud as possible and at anyone in earshot, but it was literally Part 1. Maybe don't crucify me yet.
(DIR) Post #Ar7hnKnBkl8R32aJ9c by SteveThompson@mastodon.social
2025-02-14T20:46:27Z
0 likes, 0 repeats
@quinn @ManiacalV Well, everyone's an expert on AI today.
(DIR) Post #Ar7hnLufaQiIWX5nSC by ManiacalV@mastodon.social
2025-02-14T21:04:25Z
0 likes, 0 repeats
@SteveThompson @quinn You don't have to be an expert to learn and educate others how modern commercial LLMs and text-to-image models were trained on hundreds of thousands of artists' and writers' works without their consent, though.And to ignore it you just have to care more about your nifty toys than the careers and lives it's hurting.
(DIR) Post #Ar7hnMUTRJLyJZqOGG by quinn@social.circl.lu
2025-02-14T21:32:59Z
1 likes, 0 repeats
@ManiacalV @SteveThompson this is going to be useless, but the issue isn't as simple as that.
(DIR) Post #Ar7hnNJAOpu2qnYsHA by ManiacalV@mastodon.social
2025-02-14T22:00:27Z
0 likes, 0 repeats
@quinn @SteveThompson Consent is actually really simple.Did they get consent to use people's writing and art to train their app? Yes or no?
(DIR) Post #Ar7hnO2teoTz8cxOYS by quinn@social.circl.lu
2025-02-14T22:11:56Z
1 likes, 0 repeats
@ManiacalV do you need to seek consent to read a webpage you go to? if you learn something from a webpage, are you stealing? What if the author personally hated you, would you be stealing by read their blog?
(DIR) Post #Ar7hnOeTP6XZ1AXP7o by quinn@social.circl.lu
2025-02-14T22:15:08Z
1 likes, 0 repeats
@ManiacalV from a legal point of view, it's very hard to make the case that training equals stealing. from a societal pov, i'm very uncomfortable with any precedent that says if i, or anyone or anything, is learning from a public resource and that learning itself can be condemned, and legally prosecuted. that is a terrible can of worms. i don't think there's any simple answers here, but weaponizing robots.txt ain't it.
(DIR) Post #Ar7hnYQJ4d9NK7zEQq by quinn@social.circl.lu
2025-02-14T12:57:54Z
0 likes, 0 repeats
Second part, next week. Also happy Valentine's to all you nerdy people. <3
(DIR) Post #Ar7hnZ4ieNTbLStVQG by ManiacalV@mastodon.social
2025-02-14T22:10:03Z
0 likes, 0 repeats
@quinn @SteveThompson You'll notice I haven't bashed the tech, because it's extremely useful. I love Alphafold and supported Folding@Home.I am critical of the way that SOME models were trained, and if you're going to discuss modern generative AI and just say that training was just "complicated and expensive," it's skipping vital details and shows bias.It feels more like techbro propaganda than journalism to me, but I actually care about the creatives being affected.
(DIR) Post #Ar7hojQrczyHkzoCw4 by jeffcliff@shitposter.world
2025-02-14T22:38:05.429491Z
0 likes, 1 repeats
@quinn @ManiacalV yeah there is a simple answerthe answer is "learning isn't theft".
(DIR) Post #Ar7i3H69FvLyDx3OsK by jeffcliff@shitposter.world
2025-02-14T22:40:43.047715Z
0 likes, 1 repeats
@ManiacalV @SteveThompson @quinn >And to ignore it you just have to care more about your nifty toys than the careers and lives it's hurting.Learning is a good thing. If someone learning is hurting your career, it's time to consider changing careers.
(DIR) Post #Ar7iIiLUEVFt4eaf9E by ManiacalV@mastodon.social
2025-02-14T22:26:42Z
0 likes, 0 repeats
@quinn That's a ridiculous example.You're not a machine learning algorithm. You are a human and that's how we learn.Using machine learning to train on people's copyrighted works so that your app can copy their style is so ethically disgusting that if you don't see it, you're just another #BotLicker #antihuman who needs a block, not a subscription.
(DIR) Post #Ar7iIj7hLFotUB9AIK by jeffcliff@shitposter.world
2025-02-14T22:43:29.728157Z
0 likes, 1 repeats
@ManiacalV @quinn Restricting others from learning is disgusting, using copyright to restrict the spread of culture is unethical, and if you don't see it you are a #permissionculture capo
(DIR) Post #Ar98eRdbpAPzaq37pY by quinn@social.circl.lu
2025-02-15T00:45:11Z
0 likes, 0 repeats
@jeffcliff @ManiacalV ok, now you're oversimplifying in the other direction. what is learning? are stats models the same as a child reading a book? i don't think many would say they are, but without differentiating what we mean by learning you end up perverting authorial intent, over-exposing vulnerable groups, and even disrupting people's creativity. this shit is genuinely hard to get right.
(DIR) Post #Ar98eSuJ6ZMxWuryUq by jeffcliff@shitposter.world
2025-02-15T15:13:27.429157Z
0 likes, 1 repeats
@quinn @ManiacalV what "stats" is is data and processes for learning information from data. You might as well consider using algebra to determine the missing component of a linear equation as morally questionable.>you end up perverting authorial intent,If "authorial intent" interferes with learning that is permissionculture gone amok. Authors do not get to decide whether readers read, once they release their book to the public. If this was about some kind of secret as-to-be-released private stuff then maybe but it's not. It's about copyright-restricted materials (which are inherently unethical to expose readers to in the first place)