1. Wiskkey, could u link for the this info?

  2. Are you looking for a technical explanation of how such detection works?

  3. You're welcome, and I hope you find the info useful :).

  4. Do you think this thing could write a choose your own adventure?

  5. Yes. For more info do a web search for: AI "black box"

  6. If you want to run ChatGPT locally, then no. Stable Diffusion is from a different organization than OpenAI.

  7. No, I understand that, but I guess what I’m meaning is if there is a downloadable model.

  8. For ChatGPT, no, but even if there were the model would probably be too large to run on a home computer.

  9. No. It isn't even possible in theory because of how the model is actually trained. There is no actual art in the model. In order for it to perfectly recreate something it would already need an exact copy. As

  10. S.D. v1 models have demonstrated instances of inexact image memorization - see the 4th last paragraph of

  11. Inexact does not equal precise. Which was the question.

  12. Thank you for the report :). Hopefully somebody knows an alternate.

  13. Hi! the wikiart_16384 link is down i guess, is there an alternative way to download it?

  14. Thank you for the info :). I don't know any offhand.

  15. I was looking for the comment but was it deleted?

  16. December 4, 2022 to December 5, 2022

  17. I mean AI that actually generates the image, and then says "I think these images were my strongest sources of 'inspiration'" or something to that effect, and then you can see what was used. Additionally, a ledger can be kept.

  18. First, see the Terms of Service for whatever product you're using.

  19. It's my understanding that style isn't protected in US copyright law. However, the fair use aspect for the "100 pictures" scenario is on the use of copyrighted images as inputs, which might not pass the 4th factor mentioned

  20. The best introduction that I have seen to AI copyright issues is

  21. Oh wow, that is actually alarmingly close. I wonder how much having more words in the prompts affects it. I thought the AI getting this close wasn't possible due to how it works, but this is pretty bad.

  22. That's a very rare case of overfitting caused by not 1 image in the dataset but at least a few dozen (hard to get an exact count) of an

  23. I agree with everything you wrote.

  24. An artist can make new. New is an interesting thing and very difficult to quantify, but you being a nexus of meaning and understanding of your life and experience are a unique machine of new making, and art is the thing the new produces best.

  25. AI image generators don't use images from the training dataset when generating a new image.

  26. Of course they do. That image was necessary for that model to be made. Because it’s not using the pure image but instead metadata collected from the entirety of the image makes no difference except to pedants.

  27. How to access and use this? I wanted to try a while ago and couldn’t figure it out

  28. Check the Terms of Service for Wombo Dream.

  29. That's a huge misunderstanding of that court case. It was about whether an AI itself could own copyright.

  30. Close. The issue wasn't copyright ownership. Rather, the copyright registration specified only an AI as the work's author, with no human authorship. The U.S. Copyright Office will not register a work with zero human authorship.

  31. There’s an enormous difference between drawing inspiration from something and taking those paintings, gluing them together, then claiming you’re an artist. It’s not even comparable.

  32. AI image generators do not reference images in the training dataset when generating an image.

Leave a Reply

Your email address will not be published. Required fields are marked *

News Reporter