|
Top-level
 Your interpretation does not place much value on original creation. It seems that you do not mind AI using artist's work as long as AI reproduces it and makes it available to the public. I wonder how you feel when you create something beautiful and then see its value diminished because it is being replicated hundreds of times in your community, even with other people putting their name on it. 8 comments
 @walter4096 @paulschoe it's not breaching IP if they already own the training data. The questions about stolen surplus are legitimate, if humans are subject to copy laws, then why shouldn't machines? this is where its ambiguous IMO, we all carry traces of what we've all seen in our heads. e.g. I can draw x-wing fighters, darth vader etc from memory. but I can't sell those. some rulings have been made like this I think there's a compromise to be found that lets us all benefit from this new capability (ability to copy *generative rules* vs finished works) there's more at stake than entertainment. the world is in trouble, we have big problems we're not solving fast enough with our own minds. These image generators are evidence of visual intelligence, which would be available as part of other problem solving processes check this out https://twitter.com/svlevine/status/1714307592875647291 SD 'image edits' used to help robot motion planning given a goal (imagine intermediate steps) @paulschoe @walter4096 @i0null I'm not an artist, and your point is very valid. I think the previous post was meaning that AI could be used as a creative tool. In the original cartoon, I imagine an artist behind the AI cat, working the AI cat like a puppet (not, by any means suggesting, of course, that AI is a gateway to plagiarism) Please take these comments in the spirit intended. AI art generators have limits. I'm doing solo gamedev, I'm interested in using these generators as an assist. But they're nowhere near replacing a real artists holistic design skills (Why would they be at 0.1p/image?). artists need to focus on [1] what AI art can't do (a lot) , and [2] seeing where they can streamline (eg do a personal finetune, do img2img). whats coming IMO is a tool where an artist will be able to go from storyboard -> film, i.e. someone who does comics today is doing animation& films tommorrow. it wont be big budget quality, but it will be better than hand-drawn anim still images are indeed cheapened but demand for artwork more broadly is limitless with films,games,VR Now how this will play out more broadly is an open question. Remember, LLMs are doing the same for intellectual tasks, like coding, medicine etc.. we're all in the same boat use these tools, you'll see they have big limits. the human brain is a much bigger net, which will win? things will get wild.. bring it on, IMO |
Gregory's Server
@paulschoe @i0null
the cat is out of the bag
https://twitter.com/_akhaliq/status/1716831652545208407
this model only needs12m examples from scratch.
Disney can *easily* train inhouse without breaching IP.
You need to be thinking 'how can I use this' not 'how can i stop it' (you cant)