[ars]

> manually selecting which outputs to post

That is not the law. The actual work must demonstrate creativity, and a mechanically generated works of this sort do not.

"The U.S. Copyright Office has taken the position that "in order to be entitled to copyright registration, a work must be the product of human authorship. Works produced by mechanical processes or random selection without any contribution by a human author are not registrable."

[stcredzero]

The actual work must demonstrate creativity, and a mechanically generated works of this sort do not.

The law here lags behind the technology. By procedurally generating something like a landscape, a programmer can creatively define a whole set of possible landscapes. It's fallacious to say that this can't involve artistic awareness and creativity. Here, the law is trying to apply 19th century models to 21st century media.

[ars]

The program used to generated the landscapes is copyrightable, the landscapes themself are not.

> It's fallacious to say that this can't involve artistic awareness and creativity. Here, the law is trying to apply 19th century models to 21st century media.

A problem with copyrighting mechanically generated works is there is no end to them. What awareness and creativity is shown by letting a computer make 1 million slightly different images?

All the creativity is in the programming, not in the result.

[stcredzero]

The program used to generated the landscapes is copyrightable, the landscapes themself are not.

But this is problematic. Procedurally generated landscapes can still have a definite "feel" or "look." These attributes can be protected as "trade dress." But there is no way to preclude other attributes we haven't thought of yet.

A problem with copyrighting mechanically generated works is there is no end to them. What awareness and creativity is shown by letting a computer make 1 million slightly different images?

Yes, you're right, there are algorithms that actually don't show much creativity. But this is actually already covered in the "Threshold of creativity" laws -- if the declaration that the output of an algorithms cannot be protected is removed. If you apply most of the same "Threshold of creativity" laws to a notion of meta-creativity, you still arrive at a usable and coherent law.

All the creativity is in the programming, not in the result.

Clearly, the truth is that creativity can manifest in the result.

Also understand that modern technology makes the current formulation of the law problematic. What if I simply used cheap data storage to slurp the output of a program so that I could use the output of someone's copyrighted code without permission? I could imagine doing this to a procedurally generated MMO world. This would seem to make the practical effect of the law quite divergent from its intended effect. Introduce a notion of algorithmic meta-creativity, and the law would protect against that kind of infringement nicely.

That said, in many cases, it would be advisable for the copyright holder to not reserve copyrights for such media. But in that case, I think the providing the option is the wiser choice.

[ars]

> These attributes can be protected as "trade dress."

That's trademark, not copyright.

If you carefully tuned your program to create a single result you found best that might be copyrightable.

On the other hand if you wrote a generic program that randomly adjusts the variables to make lots of nice looking landscapes then the landscapes are not copyrightable, even if you did a great job on the program so all the results were beautiful.

> Clearly, the truth is that creativity can manifest in the result.

That doesn't make it copyrightable. I could plant trees in a beautiful pattern, but I can not claim copyright on photographs taken of the result.

> What if I simply used cheap data storage to slurp the output of a program so that I could use the output of someone's copyrighted code without permission?

Not following. Does "cheap data storage" make a difference? How are you "slurping" the output? What sort of program is it?

> I could imagine doing this to a procedurally generated MMO world. This would seem to make the practical effect of the law quite divergent from its intended effect.

And what's the problem? The procedurally generated MMO world is not copyrightable. That's the intended effect - to only copyright what a person actually does.

You might be able to claim a dress trademark on the MMO world if it was distinct, but not copyright.

[stcredzero]

That's trademark, not copyright.

I never said it was copyright. Please readjust your mental model of who you're speaking to accordingly.

If you carefully tuned your program to create a single result you found best that might be copyrightable.

How is this different than carefully tuning your program to create a range of results with specific attributes? Example: All of the generated landscapes are aesthetically pleasing, but have enough open area next to obstacles to enable ambushes and also manage to look creepy at night... Such a result might indeed take

That doesn't make it copyrightable.

I'm saying that it should. IP laws date from a time where the kind of automation that makes the above possible was unthinkable. The assumption that an act of creation would result in a particular set of data is no longer warranted. Neither is the assumption that the automatic production of art is either not possible or would result in uninteresting and stereotypical output.

Not following. Does "cheap data storage" make a difference? How are you "slurping" the output? What sort of program is it?

Specifics aren't needed, but "cheap data storage" makes it possible to store the entire content of certain procedurally generated algorithms from earlier computer systems as files of ordinary size on today's typical machines. (Example: The entire Elite universe.)

And what's the problem? The procedurally generated MMO world is not copyrightable. That's the intended effect - to only copyright what a person actually does.

I'm saying that this very notion dates from a time when procedurally generated output was unimaginable to most people, and that to only copyright what a person actually does is trying to apply 19th century notions to 21st century technology. It's just like the early 20th century arguments for airlines having to pay farmers to overfly their land.

[ars]

You still hit the problem that it's possible to make a program that can output every possible permutation of every possible song. Saved as midi files you could actually realistically do that for short songs.

How can you extend copyright to include such a thing? Presumably you don't want to - but it on you to somehow distinguish between them.

Also, pre-computer automation was not as impossible as you might think, some examples: A double pendulum with a pen, a spirograph, a sand bucket on a rope, a spinning top with paint.

And see this reply from someone else: https://news.ycombinator.com/item?id=8032980 - generated artwork has existed before computers, and it is not copyrightable.

I have not completely decided if I would agree with your proposal or not, you would need to be much more specific about the limits of where copyright would not apply. Because as broad as you make it sound I would not agree, but if more limited then maybe.

[dllthomas]

Manual selection is some "contribution by a human author."

[ars]

There would be a compilation copyright to the whole work, but no copyright to the individual mazes.

[dllthomas]

Photography is a creative work, mechanically produced by manual selection of particular views of something you did not necessarily create. This doesn't seem a tremendously far cry - in a sense, he is taking photographs of maze-space.

[ars]

The result of a surveillance camera is not copyrightable. Only if a person controlled the camera. Even if the person carefully selected a site but then just left it there to record whatever would come, it's not clear that would be copyrightable.

If you fed in specific tuning parameters to the procedure and fine tuned them until you liked the output that might have a claim. But if you just let it work randomly and picked the ones you liked best that would not.

It's not enough to simply like the result, you have to actually do something to create that specific result.

[dllthomas]

How is "if a person controlled the camera" terribly semantically different from "picked the ones you liked best".

"Even if the person carefully selected a site but then just left it there to record whatever would come, it's not clear that would be copyrightable."

But in this case he selected a site (wrote the code), recorded a pile of footage, and then went through and selected particular shots based on his own personal criteria.

[ars]

He has a compilation copyright, but not copyright on the individual images.

If you leave a camera in public for a few days, then come back and edit the result into a movie you have copyright for the movie, but not for the individual stills that make it up.

To have copyright on the stills (rather than the compilation) you have to have some creativity into making them, not merely selecting them.