@br00t4c well there are many real reasons but yeah.
Notices by Daedalean (daedalean@mastodon.social)
-
Embed this notice
Daedalean (daedalean@mastodon.social)'s status on Friday, 31-Jan-2025 02:27:04 JST Daedalean -
Embed this notice
Daedalean (daedalean@mastodon.social)'s status on Thursday, 30-Jan-2025 07:02:36 JST Daedalean @br00t4c it’s not like anyone who wouldn’t got put up for any trump WH post.
-
Embed this notice
Daedalean (daedalean@mastodon.social)'s status on Wednesday, 29-Jan-2025 05:39:02 JST Daedalean @iinac @aziz @jeffowski @blogdiva absolutely. The issue you and I are having is that we disagree that there’s a way to do what your proposing that makes financial sense for both the provider/producer of the data and the business training the model.
-
Embed this notice
Daedalean (daedalean@mastodon.social)'s status on Wednesday, 29-Jan-2025 05:39:01 JST Daedalean @iinac @aziz @jeffowski @blogdiva Let’s just say we apply US copyright law to the art ai companies have already used. (We should btw). Any dollar amounts paid out that would have any material benefit to the artists would immediately end those companies. And America didn’t lose it’s advantage, american AI companies did (assuming deepseek works as advertised)
-
Embed this notice
Daedalean (daedalean@mastodon.social)'s status on Wednesday, 29-Jan-2025 05:17:21 JST Daedalean @jeffowski @aziz You’re taking a model you understand and attempting to apply it to a very different thing you don’t appear to understand. AI companies cannot pay artists enough for the amount of art (data) needed to make it worth it to the artist and have a profitable business. Hell, it’s not even clear they can be profitable currently when they’re just stealing it.
-
Embed this notice
Daedalean (daedalean@mastodon.social)'s status on Wednesday, 29-Jan-2025 05:17:21 JST Daedalean @jeffowski @aziz none of those are ai. They are basically resellers. AI isn’t taking a book, a movie, a song (or many, many books, etc) and providing mass distribution of that work to more people than an artist could possibly reach on their own. I’m not proposing banning anything like that.
-
Embed this notice
Daedalean (daedalean@mastodon.social)'s status on Wednesday, 29-Jan-2025 05:06:30 JST Daedalean @aziz @jeffowski @iinac @blogdiva never said too complex. Said not doable, not feasible. Said why. As far as ‘prohibition doesn’t work’? Copyright law prohibits use of other people’s IP without permission. Guess copyright law doesn’t work then?
-
Embed this notice
Daedalean (daedalean@mastodon.social)'s status on Wednesday, 29-Jan-2025 04:58:43 JST Daedalean @jeffowski @aziz @iinac @blogdiva which again artists opt into using because it provides some value to them. Are you honestly claiming that you had custom, individual licensing agreements with each artist whose work was in iTunes?
-
Embed this notice
Daedalean (daedalean@mastodon.social)'s status on Wednesday, 29-Jan-2025 04:50:42 JST Daedalean @jeffowski @aziz @iinac @blogdiva it invalidates the business case to do it at all. Go take your iTunes licensing platform for say visual art. Use deepseek v3 (opensource!) for your model. Figure out how much you will pay each artist for adding a piece of art. See if anyone jumps at the opportunity. 🤷♂️
-
Embed this notice
Daedalean (daedalean@mastodon.social)'s status on Wednesday, 29-Jan-2025 04:49:16 JST Daedalean @jeffowski @aziz @iinac @blogdiva dod you go out and find music to put into iTunes then negotiate with each individual artist? Or did you create a platform that had a standard license and artists opted into it? Big difference.
-
Embed this notice
Daedalean (daedalean@mastodon.social)'s status on Wednesday, 29-Jan-2025 04:43:25 JST Daedalean @jeffowski @aziz @iinac @blogdiva prohibition of these activities is my alternative. Passing laws to protect your rights to what you produce is not nihilism by any stretch of the imagination.
-
Embed this notice
Daedalean (daedalean@mastodon.social)'s status on Wednesday, 29-Jan-2025 04:39:04 JST Daedalean @jeffowski @aziz @iinac @blogdiva the computing power required to train on the data is reducing. The amount of data required is not. Deepseek uses 671 Billion parameters.
-
Embed this notice
Daedalean (daedalean@mastodon.social)'s status on Wednesday, 29-Jan-2025 04:27:06 JST Daedalean @jeffowski @aziz @iinac @blogdiva At the scales of data required to train the models a licensing agreement and financial transaction for every piece of data isn’t possible. Not enough people or time to do it. Which is a solid case for prohibiting it altogether by my lights.
-
Embed this notice
Daedalean (daedalean@mastodon.social)'s status on Sunday, 12-Jan-2025 16:24:51 JST Daedalean @dansup sign it? Sure, if they think it’s good pr. Adhere to it? 😂😂😂
-
Embed this notice
Daedalean (daedalean@mastodon.social)'s status on Tuesday, 24-Dec-2024 01:48:39 JST Daedalean @georgetakei especially since my glasses kinda take care of the first bit.