US Supreme Court declines to hear dispute over copyrights for AI-generated material

The idea that copyright requires a human is pretty well established

The question is more when a human is part of it. If we say a program generated it based on human input, what is the status with Photoshop? If "AI" is required, when does something become "AI"? Photoshop has many ways of using systems functioning similar (random, statistical methods, etc.). The same topics goes for "learning". When is something learning? Eg. early Copilot spit out license and copyright texts (and password), later it was taught to strip those, or they have been stripped by another system. Does that mean if I train something to strip license text and copyright material, based on neuronal networks I am free to distribute software, movies, etc., because the "AI" made it?

Also when it has no copyright, because it was not created by a human, does that mean companies cannot possibly own anything created by such systems?

If I read a book, thereby learn it, may I just do what up till now was copyright and trademark violations?

I think generative "AI" just shows ridiculousness of copyright law application. The rule appears to be "if you are rich, copyright law does not apply to you".

There need to courts defining where the limits are, or if there are none then effectively overturn copyright violation. We had the whole topic with "ripping in lower quality", which is essentially what generative AI appears to be doing. Lossy compression. Does that mean that lossy compression is enough to get around copyright law, again, because it is a statistical method and one could argue that at least some algorithms to a certain degree "learn" what the important bits are. Some algorithms are even pre-trained, with certain values tuned via publicly available data or some lossless algorithms like brotli are pre-trained with the internet as well.

So where is the line being drawn?

Even video editing and audio mixing is enough to get copyright.

I think the big problem here is the "sometimes" in often very non-sensible decision processes.

A big problem is that judges are no experts and when they have experts to support them then with how these hearings go important questions might not be raised and a lot of it remains guesswork.

I think it therefor is important to draw the lines to other technologies where the exact opposite decisions appear to be taken. Right now courts appear to make rulings that declare that "AIs" are almost mythical beings. But they are way less a being than a human is.

We don't seem to have any other rulings where simply using an algorithm over a copyrighted material strips it from the copyright, unless we for example declare it a form of art, but prompting an AI to output copyrighted material hardly seems like a work of art. At least not more than creating other algorithms.

The whole topic of putting in copyrighted material and then getting it out again without it being copyrighted reminds me of how some file sharing tools (more art projects) tried to get around laws for technicalities.

To over-simplify: Why can the AI be the author but the torrent program, ripping program, etc. cannot be when both are software applying algorithms?

And I am not saying it is or it isn't, but think what the implications of this are. Why is one thing fine, when having a camera "learning" the movie (in altered form even) isn't. Also when does it become a mix? And aren't the people that are confusing the algorithms employed by YouTube by putting it only a part of the video and changing speed and all that not also clearly remixing the videos?

Ah yes, the guy who was the expert witness when two sides decided to litigate what double space meant: https://azeemba.com/posts/double-space-court-case.html

My unqualified speculation is that codebases vibecoded from scratch are unlikely to qualify for copyright, leaving them in the public domain and with unenforceable licenses.

Does Eske Rex hold the copyright to art created by his swinging pendulum?

https://www.eskerex.com/?portfolio_page=drawingmachine-series

He chose the colours, designed the machine, and through gravity set it in motion.

I've been vibe-coding a raw Git repository viewer (https://repo.autonoma.ca/treetrek) from scratch. Although I have typed less than 0.1% of the implementation, I have guided the architecture, input the prompts, instructed how to refactor the code, performed code reviews, and prompt-cycled many hours of optimizations, security hardening, and plugging memory leaks. In effect, I have done the work to plan and orchestrate the software.

In this analogy, Rex's construction of the machine is the architecture, and movement of the pendulum to draw is the production of source code.

I can't guess how much human work would be required to obtain and maintain copyright.

With regards to copyright law, where would you... draw the line?