Posts by alexjc@creative.ai
(DIR) Post #ASHO2MTEJbuxSCdMBs by alexjc@creative.ai
2023-01-31T18:16:05Z
0 likes, 0 repeats
@lritter Copyright infringement already happens at the training stage, so whether it's in the model in some visualizable way, and whether it can be output, is irrelevant in some ways.
(DIR) Post #ASHO3JeyFbMl1ha4BM by alexjc@creative.ai
2023-02-01T12:50:32Z
0 likes, 0 repeats
@lritter You need the necessary rights or an exception to copyright to keep a file around for the purpose of training. If you don't have written permission from the rights holder, or a valid exception, it's infringement.
(DIR) Post #ASHOrlv5SGrP9VB6Zs by alexjc@creative.ai
2023-02-01T12:52:06Z
0 likes, 0 repeats
@lritter 🤷♂️ Read through all the copyright acts then!
(DIR) Post #ASHOroCsx4HmFHWLb6 by alexjc@creative.ai
2023-02-01T12:54:05Z
0 likes, 0 repeats
@lritter OK, then answer this question. Which article in the EU copyright act deals with the rights necessary to train a model on web-scraped data?
(DIR) Post #ASHOrr2MQeeL1VcU6q by alexjc@creative.ai
2023-02-01T12:56:42Z
0 likes, 0 repeats
@lritter ** Mutters something about pearls and swine... **I answered it here: https://creative.ai/@alexjc/109789566257445813
(DIR) Post #ASITag66XiM36IxZD6 by alexjc@creative.ai
2023-02-01T07:56:31Z
0 likes, 0 repeats
@Mastokarl @Raccoon The original paper calls it semantic compression. All works can be compressed semantically, but few are found that can be compressed with sufficient pixel accuracy.I bet with enough computation you could find better matches for even more images...
(DIR) Post #ASITvXvu5r4DrO8gxk by alexjc@creative.ai
2023-02-03T08:04:37Z
0 likes, 0 repeats
@Mastokarl @Raccoon You're right about that: it's not theft it's infringement. But Racoon's interpretation is generally correct... if the model is infringing you could say all its outputs are unauthorized derivative works and could get you into trouble wherever in the world you use them.
(DIR) Post #AWT0meFIgMAwf5NKqW by alexjc@creative.ai
2023-06-02T17:21:22Z
0 likes, 1 repeats
Stability and MidJourney's "moat" is being reduced, filled with obsolete GPU cards and schadenfreude for their VCs.Instead of 150,000 GPU hours it takes 9,200 — a 16x reduction.🔥Same happened for data 16x with Firefly; ethical competitors close by! 👏https://arxiv.org/abs/2306.00637
(DIR) Post #AWinsGsdRUTXvKU9GC by alexjc@creative.ai
2023-06-02T19:15:18Z
0 likes, 0 repeats
Remember @LAION_AI@twitter.com claimed it deleted the dataset files downloaded after use to comply w/ German Copyright Law?Well, that was a lie!🌭We now know how the sausage is made.They actually keep copies of the data and provide access to researchers. 100% non-compliant.🚨 BUSTED.
(DIR) Post #AWinsHgyQKk2RS2Liq by alexjc@creative.ai
2023-06-02T19:15:20Z
0 likes, 1 repeats
This bombshell comes from a new paper today, ironically called Würstchen, or Sausage, which admits how things work behind the scenes.This takes the prize for Most Impactful Acknowledgements section ever!https://arxiv.org/abs/2306.00637
(DIR) Post #AWinsMAXmKluKJu1Mu by alexjc@creative.ai
2023-06-02T19:15:21Z
0 likes, 0 repeats
Since web-scale datasets are so big, it's a pain to download them everytime you want to train a model — as required by German law.So what did LAION do instead?It keeps a copy around. Not only for itself, but for external researchers who need to use the dataset!
(DIR) Post #AWinsO3WlOLcB8dRi4 by alexjc@creative.ai
2023-06-02T19:15:25Z
0 likes, 0 repeats
This work was funded by Stability, making compliance with German © Law more problematic as non-commercial purpose is required.Previously, evidence surfaced that indie researchers had to sign contracts with Stability to provide them IP, which means it doesn't fall under 60d (2).
(DIR) Post #AWinsPmwK2GdYGtE8G by alexjc@creative.ai
2023-06-02T19:15:26Z
0 likes, 0 repeats
For context, here is the awesome VICE article about how LAION lied to photographer (@robertkneschke@twitter.com) in their reply to him, about having deleted the data.Not only did they lie, they are also infringing and non-compliant with German Law.https://www.vice.com/en/article/pkapb7/a-photographer-tried-to-get-his-photos-removed-from-an-ai-dataset-he-got-an-invoice-instead
(DIR) Post #AWinsRS68UmgiD9bvM by alexjc@creative.ai
2023-06-02T19:15:28Z
0 likes, 0 repeats
What happens now? 🤔Since co-author Marc Aubreville is in Germany, he'll be asked to testify in upcoming court case(s). He didn't do wrong, but has the choice to lie to defend LAION by LAY'ING — or otherwise cover up evidence.I wouldn't recommend that! (Not Legal Advice)
(DIR) Post #AWinsTFlRK6gIXOmye by alexjc@creative.ai
2023-06-02T19:15:32Z
0 likes, 0 repeats
LAION will scramble like crazy to get its story straight internally, and try to correct the record.If they do that, you know they are extra guilty & worried because the cover up itself makes the crime worse — as per the Watergate scandal.
(DIR) Post #AWinsV8kQNgO9M8DJo by alexjc@creative.ai
2023-06-02T19:15:34Z
0 likes, 0 repeats
A big thank you to all the activists who have been fighting the good fight over the months. 🙏A big middle finger to lawyers and scholars on the wrong side of history defending unethical and immoral organizations. 🖕😂Feels good. Downhill from here! 🌴
(DIR) Post #AWinsX871iMYJxqjbc by alexjc@creative.ai
2023-06-02T20:10:33Z
0 likes, 0 repeats
If you're on #TeamHuman, please do share this widely with people who tried to disagree with you & were stubbornly wrong all along!It's the Wile E. Coyote moment for LAION.@neilturkewitz@twitter.com @stealcase@twitter.com @kortizart@twitter.com @JonLamArt@twitter.com @adamconover@twitter.com @ZakugaMignon@twitter.comhttps://twitter.com/alexjc/status/1664712115565780995
(DIR) Post #AWinsYswV5PtlUleEq by alexjc@creative.ai
2023-06-02T21:01:11Z
0 likes, 0 repeats
Technically, there's tons of supporting evidence throughout LAION's public repositories, proof they kept datasets around.Now there's external confirmation & researchers to act as an expert witness, so it can be disclosed.Legal word games won't help!https://huggingface.co/laion
(DIR) Post #AWinsaX2NV5Cs8XBNA by alexjc@creative.ai
2023-06-04T16:17:53Z
0 likes, 0 repeats
About clearing up the record, Marc Aubreville wrote this despite saying he would only answer research questions:https://twitter.com/MAubreville/status/1664948944437166080You get such statement when PR agency & Law Firm have a baby; its first words "to the best of my knowledge" and "not that I know" variations.
(DIR) Post #AWinscDy5N1A7ZcyvY by alexjc@creative.ai
2023-06-04T16:17:59Z
0 likes, 0 repeats
That kind of statement doesn't help @MAubreville@twitter.com; it's misdirection at best.As a professor and the most senior person on the project, here is the quality of the answer that's expected if you intend to resolve the situation from your perspective — without any fault on your part.