Wednesday, March 31, 2010

copyright vs procedural content

In 2006 the Damien Hurst was accused (via) of copying a design ("valium", top) from Robert Dixon ("true daisy", bottom). Hurst is a quite well known millionaire (valium's 500 prints sold for "four and five figure sums") and Dixon a relatively unknown computer/geometric artist. At first glance it looks like Hurst just coloured in, flipped and cropped Dixon's work, but looking a bit deeper it gets more interesting.

Dixon's work is based on the patterns of spiral Phyllotaxis, a natural phenomena that describes (among other things) the arrangement of needles on a pinecone and pattern of seeds on a sunflower head.

IMGP0618 - cu - sunflower head
by RaeA

This is a old mathematical pattern that can be expressed as an formula, and was well known enough to be commented on by da Vinci. Because of this Dixon was able to write code (similar to mine, that produced my diagrams below) to generate his pattern.

Simply, Phyllotaxis is a spiral pattern of points. Although many different spacings are possible, the fantastic thing is that nature always chooses the same spacing. My diagram below shows a few possible spacings, and one jumps out as being "correct" as it covers the space most evenly. This is the pattern that is used by nature, Dixon and Hurst.


But the question is - should he be able to claim copyright on it? and is Hurst's work derivative of it? Copyright law say yes - Dixon has some claim over this figure, as he created it (no matter it's source) and has a copyright claim. The law just asks that he has the work is "original, and exhibits a degree of labour, skill or judgement". Maybe he was the first to use circles of that diameter to represent the pattern. So if I can build a general machine that with some parameter imitates Dixon's machine - does Dixon own the output of that machine?

Okay, so I built the above machine after viewing Dixon's work, at some point it probably outputs something that looks enough like Dixon's work. This is a real headache for me as a programmer - is the output of my program derivative? does Dixon own the copyright to the program he's never seen?

One answer is that I should block particular frames that contain copyrighted images. To do this in this single instance, I would have to describe Dixon's work in the program to stop it infringing. The program would then contain copyrighted material and be illegal. To do it in the general instance, for all copyrighted spiral patterns, I would need to do an international prior-art search, which would cost vastly more than the two hours it took to create the above video.

There are excuses for copyright, for example if you can prove you didn't know about the existing image, that you didn't "copy" it, then the use is mostly covered under fair deal (or fair use in the USofA). But some inventions are covered by the more powerful patents - in this case even if you come up with the idea independently yourself, the first person to be awarded the pattern hold all rights to it.

In situations like this, Intellectual Property (IP) law feels very wrong, this is an algorithmic pattern, and if I'm allowed to copyright it, or copyright a video of all possible versions of it, there is nothing left for the rest of the world to use.

If I can write a program that recreates your patented "invention" or copyrighted "original" object do you own that program? If I can write a program that will write that program, given a certain command (abracadabra!), do you own that word? do you own/control the first program and the second program? are both programs illegal (under DMCA-like legislation they probably are). Does this mean that any program that can recreate patented work is illegal? does this include a computer compiler (probably not)?
At the moment these ideas are unlikely, and limited to toy examples. But in ten years, when anyone can ask the computer to create a program that lets you edit photos ("now I want a circular brush!"), and that program infringes on hundreds of patents, it seems like IP law will have to change.

Another reason to burst the IP bubble sooner rather than later? A previous post on copyright can be found here.


  1. very insightful - I do wonder, as we have more and more procedural character creation in games, what if the computer comes up with Madonna as a character, but does so through completely procedural means? Equally, as architects use genetic algorithms to model buildings, what if my procedural environment generator creates a tower already built by Anish Kapoor and puts it in my game or movie? If no human intervention was required, how can I have breached anyone's rights? Can you copyright the arrangement of a pile of sand? (you can, of course, famously copyright a few seconds of silence...)

    IP law is based on 17th C ideas and issues, and is laughably behind the reality of C21 commerce and technology. I cannot help but think that we need a new answer, and we needed it yesterday.

  2. Anonymous15:04

    copyright doesn't protect "ideas", only "unique expressions" of ideas.

    i have to wonder if Dixon's raw rendering of a phyllotaxis spiral even contains enough "unique expression" to give him legitimate claim to copyright.

    it's a bit like trying to claim copyright on a golden spiral - it's a mathematical construct of which there's only one true instance, the form itself would not be copyrightable.

    Hirst's version, on the other hand, potentially does contain elements of "unique expression" - his selection and placement of colors.

    The troubling bit is that Hirst's version apparently contains *exactly* the same dot arrangement as Dixon's. If Hirst's version had had one even just more spiral revolution, i'm sure this would be a non-issue.

    ...will have to let the courts decide that i guess.