Hacker News Reader

Haha 1) this is crazy 2) I think it'd be more fun to train the data set against Christmas trees. The Eiffel tower's lights don't much resemble other buildings since the lighting is decorative and not functional. Much closer to a Christmas tree ;-)

This kind of logic when applied to law is debunked in the widely circulated essay "What Colour are your bits?": http://ansuz.sooke.bc.ca/entry/23.

Short answer: the law sees right through trickery like this, focusing on your intent instead.

Interesting - thanks for the link, but I'm not sure my question is actually "debunked" by that article.

I outlined two possible scenarios: either

- the lights are not exactly correct, and so this (manipulated) photo isn't covered by the copyright on the lights on the Eiffel Tower or

- the lights are correct (or close enough), and so the lights on the Eiffel Tower are in the obvious places (given that a ML algorithm could predict them) - in that case, is there enough creativity in the arrangement for copyright to have been granted.

I think you're right that, in the second case (which the article addresses) copyright would back the original creator - the "but ML predicted them so there was no creativity" argument is probably pretty weak. (It's also different from the idea laid out in the article you linked to - in that case, the recreator of 4'33" was very prescriptive about what they were creating.)

However, in the first case, I don't see how the law "sees through trickery" - I don't think there is trickery going on. My intent is to create a plausible-looking image of the Eiffel Tower at night (e.g. to illustrate an article on it). It doesn't matter whether the lights are in the right place - I'm not using it to discuss the lights.

As someone else commented (https://news.ycombinator.com/item?id=16198190), the photo of the Statue of Liberty in the article is not actually the Statue of Liberty on Liberty Island in Manhattan - it's actually in Tokyo. However, as a casual reader, I see the photo and recognize it (incorrectly, as it turns out), which is all the author needs to make his point.

> It doesn't matter whether the lights are in the right place - I'm not using it to discuss the lights.

It doesn't matter that you managed to find a way to create a photo that looks like the Eiffel Tower at night by not actually taking a picture. Your goal was to make a picture that looks like the copyrighted thing. Therefore, you effectively copied it. That it's a poor copy doesn't stop it being a copy.

Put this another way. Let's say I ran a random number generator until I had an exact bit-for-bit copy of a JPEG that is a copyrighted image. Did I copy it? I think the essay establishes that I did, because that was my goal. The process of stopping only when I had an exact bit-for-bit copy will have made it a copy in a judge's eyes.

What if I stopped when, rather than being a bit-for-bit copy, it was "good enough"? I don't see that this would make it any different.

What if I used something more intelligent than a random number generator to increase the chances of me getting the desired result sooner? I don't see how that would make it any different.

Conclusion: how, technically, you come up with the result doesn't matter. If your goal was to come up with something that looks like it, and you stop when you have something that resembles it (doesn't matter if the match is good or bad), you will be deemed to have copied it.

Thanks - interesting!

So, let's imagine that the Statue of Liberty in Manhattan was copyright, such that you couldn't take (or at least benefit commercially from photos) of it...

The author of the original article used a photo of a very similar statue, but one that is based in Tokyo. I don't know whether they're identical, but let's assume not, and that there is no similar copyright.

(I appreciate this is very hypothetical, but so was your "let's say I ran a random number generator" argument.)

The author of the original article used a photo of the Tokyo statue with the intention that readers would interpret it as the Manhattan statue.

Would the author be deemed to have infringed the copyright on the Manhattan statue?

By your "if your goal was to come up with something that looks like it, and you stop when you have something that resembles it (doesn't matter if the match is good or bad), you will be deemed to have copied it" definition, it seems they would.

> Would the author be deemed to have infringed the copyright on the Manhattan statue?

Possibly. You hit the nail on the head when you say "with the intention that readers would interpret it as the Manhattan statue". It's all about the court's interpretation of the intent of the parties involved.

I don't think anyone can predict what a court would decide. Perhaps the court would say that because any statue would have done, and what the reader interpreted it as was immaterial to the content of the article, it wasn't an infringement. I wouldn't be at all surprised if this is what would happen.

All I'm saying is that if you try to "work around" copyright using "parallel reconstruction", the courts are entirely able to see through this if they so wish. The courts are never going to say "you got us there". I'm not making any claim as to whether they actually will or not in a particular hypothetical case, since that depends on the circumstances of the court's interpretation of the bigger picture (intent, appropriate interpretation of the law, etc), which is far more subjective.

This is an abuse of machine learning. Amy algorithm can be tweaked to achieve the intended result, even if you limit the dataset to buildings before the Eiffel tower had lights. It always comes back to a human.

I think that's exactly the point he's making. There's only so many sensible ways one could decoratively light the Eiffel tower, and OP is suggesting that machine learning could generate it artificially.

If that's possible, is there really anything that is non-obvious and copyright-able? If it can't be done, the manipulated photo cannot be held accountable for copyright violations (but it certainly would look like it ought to). The fact that it can come down to something any human can optimize for only makes the law surrounding the copyright of a public display on a building even more absurd.

Well, of course a human could have come up with the light scheme. Anything copyrighted by definition has to have human involvement (cf. https://en.wikipedia.org/wiki/Monkey_selfie_copyright_disput...) The whole point of IP protections is that you were the FIRST human to come up with it. And now we are in a whole different conversation: does it matter that you were the first?