• 0 Posts
  • 16 Comments
Joined 1 year ago
cake
Cake day: June 30th, 2023

help-circle




  • That is not how the animation industry works, at all. Most creative art positions require demo reels and portfolios. But sure would be nice if animators were that valued a recruiter would hunt usenet for their old pirate samples.

    Also worth nothing, especially in animation, your cuts in an episode are often going to be various 10 second shots spread between multiple episodes across a season. That’s why demo reels make sense.



  • And yet, if you are being instructed to reproduce the Mona Lisa in explicit detail, billions of times as a computer is, 1 layer of abstraction in a digital training is not at all the same as your example with human artists.

    Oh, and Bot 1 isn’t instructing Bot 2, it is instructing Bots 2-1,000,000, all at the same time. And each round it kills 999,998 of them, and births another 999,998 of them based on the best of them and goes again.

    This is not like you example. It is more like brute forcing a password.


  • Tell me you don’t know how AI works, while sounding arrogant.

    “literally has no memory” is a fun semantic argument against the actual reality that these models have been caught attempting to reproduce the signatures of artists they trained on, and in some cases reproducing the likeness of copyrighted photos from key phrases. Like the “Afghan Girl” fiasco that say them ban the phrase immediately.

    This was with Midjourney https://youtu.be/kqPKNksl9hk?si=i9rFLHyg8ImRuTjb

    Your argument is basically that if I type in “Mona Lisa” into a generative engine and get a copy of the Mona Lisa with slightly wavier hair, and some trees in the background are different, that it is not derivative. Which is a laugh.



  • More accurately: Should we ban someone from launching a new TV show called Smashing Bad that stars Wally Witten, a high school chemistry teacher who turns to crime to make fentanyl and…

    Yes I think we should. That is derivative work. The AI projects are not a tool, they are automated derivation made on models based on stolen art.

    These tools need to be trained on data sets that were not stolen, first off. And the results need to be a tool, like generative fill in photoshop, that is based on the rest of the image. Not a replacement for the image itself.







  • I’m a lifelong windows power user, and above average even in my industry for knowledge on technical expertise.

    Nothing I know translates to Linux. Not the file structures, the commands, the permissions, the file systems.

    You truly have to commit to learning an entirely parallel form of computing environment to become comfortable in Linux. And being frank, it is the most customizable and unique user experience out there, but it is also infinitely less user friendly. And for every time a 2 line terminal command fixes a problem and saves time compared with windows, there are dozens of instances where time is wasted for hours learning that command, its exact syntax and usage, and if it is the one you need for your circumstance.

    Another user here recently said that it was when they were going through and compiling their own drivers to make their Webcam work and having to follow guides to make system specific tweaks that they just quit and went back to Windows for ease of use.

    Linux is the OS of power users. Not even power users like me, but extreme power users who either have the time or training to learn that parallel system. All of which is easy if this is your job, but in many ways you are learning a second language of sorts.