[tehologist]

Copyrighting code never made sense to me. We already have patents for intellectual property. If two people use the same RFC or Whitepaper for an algorithm in the same language, they will probably name the variables similarly and their code will look very similar. Just like if two people wrote out the same hamburger recipe or instructions for hooking up a stereo would also write something similar.

The copyright on the implementation will outlive the patent and allow the implementor to legally take action on claims of copyright infringement. Even though a program is literally just a list of instructions to implement the expired patent.

[pornel]

Copyright protects not the idea, but specific implementation of it. It's there to prevent unauthorized copying of software. Not every software has to be novel enough to be patentable, but may still take effort to write a millionth-first JS framework.

If you take someone else's software without a license and rename variables, it will be a copyright violation, because you've copied (and then modified) it without permission.

But if you write your own software from scratch, even if it happens to be almost identical to someone else's code, that's fine. You've done your own work and a copyright owner can't stop you from doing that. They control their own work only.

As you can see, this is very much tied to human work and intent, since the concept has been invented long before ML existed. This is why ML "learning" and doing "work" is so controversial and appears to be a loophole in copyright.