do that enough times and it becomes a sizeable database. is it gonna be a derivative work then? at which size? some databases, like ones that point to and describe 'training data', are pretty much just 'saying if an image is blue'. describing what's on that image. is 'one annotation' a derivative work? is a million of them? either way, whatever's the number, it's apparently enough for those databases to put up a license on them, and regulate whether and on what terms are derivatives gonna be made of them.

No, a data point of if an image is blue is not illegal.

> it becomes a sizeable database.

No judge has said this.

Also, the case law actually supports me not you.

What we have described here actually sound very similar to the Google court case about search indexes/ect, which was ruled not illegal.

So are you just going to falsely claims that this judge is wrong and that all search indexes are illegal or something?

It's clearly not. It is clearly not illegal for all of Google to exist. Search indexes are not banned worldwide.

Also, I am not sure why you are even arguing about this when you previously said '"but what does the word "derived" mean anyway" lol. i don't know'.

If you don't know then thats fine! You already admitted that you don't know, so stop pretending like you are confident on any of this when you are admitting that you don't know on the most important part of it.

that's you, asking "are the results of a program that were derived from something gonna be a derivative work?". what is a derivative? what does derived even mean really, right? a program wouldn't be a derivative, results could be, and they're just derived from something. that is just gonna be the chain of processing there. and, well, "is it blue" database might not be a "derivative work" (for legal purposes). but it still could be a "work" that could be protected with a license nonetheless, huh. so, it's not a 'derivative work', but it's a 'work', and it was 'derived' from something. if you're more comfortable, it could be called 'a derived work'. not a derivative work - a very important distinction there.

anyway, that example with "a hypothetical blue program" - isn't gonna map so neatly to 'an image processing program, that takes in images, and spits out images'.

> and, well, "is it blue" database might not be a "derivative work" (for legal purposes).

Finally we got there! I am glad that you agree that according to the only definition that matters, in context here, is that this stuff could be completely legal!

Just like search engines are legal according to existing laws, yes a blue database could be legal, and finally, so could AI models weights.

I am glad that you have pulled back from basically everything here and admitted that this stuff could be fully legal.

Really you could have just said that from the beginning and solved all this confusion.

sure! except, well, it's still a derivative work. even within those that end up at 'well, for "legal purposes" and in some settled case, it isn't' - it still got called a derivative work and was considered as such, by people who filed a lawsuit, etc. and it might continue to get called that, unfortunately lol. those cases - might not translate into some kind of 'decision' about other kinds of derivatives. some kind of database, or training datasets (with multiple parameters, descriptions, etc.), or models, or AI output - are all gonna be different, unfortunately. not as much of a smooth sailing there, I'm afraid. and some random "search engine" case that keeps getting alluded to, isn't gonna map to those things as neatly, let alone all of them in that cute daisy-chaned manner. hell, google with their newer AI search tricks, like bard and generative answers - might only be yet to find out whether that checks out legally or not.