Hacker News Reader

Comment by j4yav | Hacker News Reader

The part that feels really obvious to me is that, if I made an AI that could generate music by looking through the entire (copyrighted) back catalog of the Beatles for example, and it would output music that I could control to be very much or even exactly like the original recordings, or I could accidentally do it, that it wouldn’t really be a way to launder the original licenses/copyright into the public domain.

Or maybe it is, but if so it essentially means the end of licensing because it would be trivial to make an AI that can take an input and produce the same output. Or maybe even cp is good enough to strip the source of its original license in that case.

Open source licenses are worth protecting or you break the cycle that helps more software be open.

View on HN

> and it would output music that I could control to be very much or even exactly like the original recordings, or I could accidentally do it, that it wouldn’t really be a way to launder the original licenses/copyright into the public domain.

The test for non-literal copyright infringement is "substantial similarity." If, after filtering out irrelevant and non-copyrightable elements, the allegedly-infringing work is substantially the same as the original work, then it infringes. If it infringes, then two common defenses are independent creation and fair use.

In your hypothetical, the AI-generated work would infringe the original because you stated it would be substantially the same as the copyrighted work. You can't claim independent creation because the algorithm was dependent on the original work and you controlled the output of the algorithm to be exactly like the original work. Fair use is pretty much a non-starter, so I'll skip that analysis.

So, no, you couldn't use an AI to launder copyrighted works into the public domain.

Unless you are Github, in which case having your AI copy code vertabim is ok?

> Unless you are Github, in which case having your AI copy code vertabim is ok?

You have to filter out any non-copyrightable elements before you do the substantial similarity analysis. For code, that means removing non-expressive elements like arithmetic or boolean expressions, looping, recursion, conditionals, etc. APIs are not copyrightable under the recent Supreme Court holding in Google v. Oracle.

How much of your code is actually left after filtration?

I’d wager you’d get rid of over 90% of most codebases by removing all that.

But perhaps there could be a way to make something that automatically converts these "non-expressive elements" into copyrightable elements?

Output would probably be maddening to figure out though.

That's nonsense though. If we filter out the non-copyrightable parts of music we'd remove all the pitches, eighth notes, quarter notes, half notes...

We could remove the non-copyrightable parts of text works too. Just take out all the basic building blocks of language like verbs, nouns, stop words...

Individual notes, like individual words, are not copyrightable. Longer sequences, however, can be if they are original.

You’re basically arguing that whole books aren’t copyrightable because after you filter out all the words which aren’t individually copyrightable there isn’t anything left to copyright. Which obviously doesn’t make any sense.

The tests exist as a way of determining if someone did the action of copying which is the important thing at the core of it. And in this case the facts aren’t really in dispute, it’s whether what GH is doing counts as copying.

If you had a really really good memory and remembered almost exactly how your former company implemented something and when faced with a similar problem unknowingly produced similar code. It’s iffy as to whether this is copying. Because it really probably isn’t — you’re allowed to learn from copyrighted works — but the courts aren’t omniscient and when presented with the code they very well may rule that copying was more likely than not.

But we are omniscient in this case so we don’t really need the tests. Is what GH does more like copying or learning? This isn’t something that can be determined purely from the output of the tool.

Code elements like conditionals, loops, arithmetic or Boolean expressions, assignment statements, etc. are not copyrightable, but not because they’re insignificant (like individual words in a novel), but because they’re useful articles. Copyright protects expression, not utility, which is the domain of patent law.

The problem of API copyrightability was not resolved by Google v. Oracle, the ruling was that no matter whether or not APIs are copyrightable, copying APIs is fair use (which is a concept in the USA but not in many other countries around the world).

The way I see it, if you would use Copilot to completely (or largely) reproduce an existing work (software), then you would be infringing copyright. This is similar to using an AI to largely replicate a piece of music.

If you are using it to mix a snippet of code (from a sufficiently large code base) into a large code base of your own, then you are just remixing. That is not infringement. In music, there are entire genres based on remixing. You could even take it a step further and ask yourself: what is not a remix?

Sisqó had to settle with Ricky Martin for quoting "Living the vida loca", then there's the debacle around Katy Perry's Dark Horse...

Point is songwriters absolutely get litigious over reproducing small portions of their IP.

complex.com/music/majority-sisqo-thong-song-publishing-owned-by-writer-livin-la-vida-loca

huffpost.com/entry/katy-perry-dark-horse-lawsuit-payment_n_5d43d825e4b0acb57fca3ff2

factmag.com/2016/06/25/sampling-hip-hop-copyright/

Yeah, I've heard about some of those cases. It's surprising how far copyright can be stretched sometimes.

What's more surprising is to see copyleft advocates positioned so strongly in favour of giving copyright that kind of reach. I think that in a different context, some of the cases you refer to would be used by these same copyleft supporters as examples of why copyright needs to be more weakly enforced, not more strongly.

If I don't need your permission to do something with your work, then the conditions under which that permission is granted lose their force.

At the very least it should be consistent. If Microsoft can sue me for something trivial that probably shouldn't be illegal but is, then I can sue them for something trivial that probably shouldn't be illegal but is.

And in any case I think at the point where the AI (yes I considered scare quotes but decided not to) copies the code verbatim including swear words it is kind of obvious what happens:

The AI part isn't about independent creation but about figuring out what to copy.

"Or maybe it is, but if so it essentially means the end of licensing because it would be trivial to make an AI that can take an input and produce the same output."

Yes, this is what is pretty interesting to me. I said in a previous comment that I have a really good OS generating AI. It asks you your favorite color and outputs a disk image you can use as an installer.

Right now it just happens to output a cracked version of Windows if you answer "blue". Who can know how that happened? It's a black box after all. Seems useful though, since Microsoft is loudly saying that if I distributed this it would have no license problems at all.

I think this is peak programmer logic. You can’t just declare something a black box or declare something an AI — the courts, people, aren’t that stupid. Lawyer words aren’t that magic.

If when you crack open Copilot it’s determined that it’s not actually learning and boils down to storing and regurgitating snippets of code no matter how few or many layers of indirection it’s still infringement.

What your AI actually does is underneath all the indirection is what’s important.

> [..] not actually learning and boils down to storing and regurgitating [...]

Some would say that pretty much _all_ learning involves "storing" and "regurgitating".

Aside: my daughter started writing out her name at kindergarten a couple of years ago. One of the staff seemed a bit dismissive of this, and claimed what was happening wasn't really "writing" it was "merely memorising the shapes of letters and then reproducing them in the right order". <rolls eyes>

I didn't bother to argue...

Yeah ... that was my point. What copilot does underneath the indirection is steal GPL code and derive "new" stuff from it in a way that launders the licenses.

We know what it's doing, it's doing very well understood statistical inference techniques to derive outputs.

Also, it's trivially easy to teach a real neural network to output a specific binary stream in response to a specific input. My OS AI could use the exact same technology as Copilot, I'd just train it very specifically.

I think the main point that the article makes is that for copyright to work you need some notion of a creative work, and so far it's generally accepted that snippets like

  i = i + 1

aren't creative enough to be covered by copyright. The interesting point is where you draw the line between what's boilerplate and what's creative, and legally it will presumably come down to showing that copilot crosses that line egregiously enough for someone to think they've got a successful chance at legal action.

Since that article was written people have shown it will generate quite long coherent sections. It will even generate someone’s private about me page: https://twitter.com/kylpeacock/status/1410749018183933952?s=...

But that about me page is the very definition of boilerplate text, so really it only gives weight to the argument that it's not producing original work.

You got downvoted, but I kind of like this argument. There are a million "about me" pages, but Copilot did a good job of picking one for "generic software engineer". If it could just have changed a word or two to a synonym, it would be great.

That page was so generic that I could fully believe it was originally written by a GPT bot.

Bullshit.

That not an existing aboutme page. You can go to davidcelis' website and verify that it's completely different.

Copilot just picked a random person and linked to their social media accounts. You can search any large quote within that about me on Google and not find a match, it is unique.

The only two examples of generating large sections of copyrighted work are the quake floating point hack and the zen of python. Both those examples are commonly known and copied and talked about, to the point that they have wikipedia pages.

Sure, but GitHub's own promotional pages (pretty much any of the gifs on https://copilot.github.com/ as well as other articles, e.g. https://docs.github.com/en/github/copilot/research-recitatio...) show it producing much more elaborate segments than that.

In fact, that's a crucial selling point for the product.

But as I understand it, copilot can generate much longer snippets, even entire functions.

I think the big question is, if copilot ends up copying significant portions a GPL work, not just tiny snippets, is the resulting work infringing, and if so, who is liable?

If a tree falls in the forest and no one is around to hear it, does it make a noise? If it is infringing and someone is liable, what proportion of the infringing cases would be found and what proportion of those found instances be brought to the courts?

I have almost no sense of how often code is infringed currently and how often anyone does anything about it. I have a feeling that we live in a world with constant infringement, basically no one cares, and no one does anything about it. And I would assume, the status quo will maintain its current course with this new tool. But again, I'm giving zero factual evidence, it's just a feeling from not seeing our hearing almost any news about open source code infringment.

If you use copilot and it generates a substantial amount of code, you don't know if that might be a replica of code from another project with an incompatible license. If you are law abiding and/or want to respect open aource licenses, then it is on you to figure out what if any license that code would fall under. Which means copilot would only be useful to developers who don't care about FOSS licenses or only use it for snippets that couldn't possibly be considered original enough to be covered by copyright.

Unlike patents, for cooyright, independently created work is not infringing. So, if you build a model that really does actually model music, this could be argued to be independent creation.

But there is also caselaw (involving George Harrison IIRC) on "unconscious copying", where having heard a piece is suggestive that it was not an independent creation, despite not being deliberately copied. So, training on a corpus that includes a specific piece is arguably a case of that.

There"s an interesting question of whether a model is just a sophisticated statistical compression of a corpus, or whether it is a thing in itself. I would say, if it finds patterns that are disproportionately simpler than the corpus, it has found "something".

But another view is of creation as involving a side-channel or one-time pad, in that music is created by a human and heard by a human, who have common information that has never been present in music before (e.g. specific aspects of common neurophysiology, auditory anatomy, exact heartbeat waveform, new sounds/rhythms in the world, new speech patterns, associations between existing melodic fragments and words/emotions/visuals/status etc). In this sense, truly new music is discovery of the Human Music Processing System, which ultimately involves the whole human and their social and physical experience.

You say AI, but this is just a database with a weird query language. The fact that it has to be trained on massive amounts of code and can only regurgitate variations of snippets of the training database back makes it quite clear that there is no intelligence in this thing.

If it were intelligent, it’d be given the Lagrange specifications and then I’d be able to say “Write me an open world video game based around gang culture and I’d like it to run on the raspberry pie zero.”

Oddly enough there's an Adam Neely video I saw some time ago that may address your points:

https://youtube.com/watch?v=sfXn_ecH5Rw

Wouldn't the parallel be closer to having an ai remix a bunch of songs together?

The Beatles are the farthest thing from public domain possible and would never be included in such an exercise.

Valid point - most hilariously poor example to back that point up in recent memory.

Can put Beatles lyrics as comments. Copilot do look to copy those as well. People can bomb search results, some will certainly bomb Copilot.

You're giving AI too much credit, it's just a tool, it does not have it's own intentions.

I.e. if you buy a piano or a guitar, you could play and record copyrighted music on it. That's not piano's or guitar's fault though, it's yours.

It sounds like you might be the one giving it too much credit. AI is a glorified markov chain, which is essentially a compression algorithm. I agree that it can be an instrument (I've done it: https://soundcloud.com/theshawwn/sets/ai-generated-videogame...) but it's almost trivial to train a model that memorizes by rote.

Suppose a model was trained solely on a single Beatles album. It could only spit out that album. That would be clear infringement, wouldn't it?

It’s funny that people say it’s a glorified Markov chain.

No. It’s not. A Markov chain has some very specific properties that are absolutely not fulfilled by GPT-3 models.

Just say “stochastic” if you want a buzzword. Stop appropriating Markov chains.

“It’s a stochastic” doesn’t flow, though I guess I could use "stochastic random walk."

What properties does a gpt-3 model have that a Markov chain doesn’t? (Other than effectiveness.)

ANNs aren't Markov chains

Well, yes. And apples aren't oranges, but they share a lot of similar traits.

"Given a prompt, provide a completion" is what a Markov chain does. GPT-3 is exactly the same, in the sense that apples and oranges both satisfy your hunger.

Is calculator.exe the equivalent to TI-84?

I would say that a TI-84 is a glorified calculator.exe. The TI-84 is usually more effective, but they’re both calculators.

Willfully misunderstanding your interlocutor seems ungracious, you apparently know stochastic is an adjective, so why did you try to use it as a noun?

GPT-3 is conditioned on the entire input sequence as well as its own output, which is strictly NON-MARKOVIAN. In fact, the point in saying something is Markovian is exactly that: the state transition probability only depends on the current state.

But the current state could include all of the last N output tokens.

Actually GPT-3 is a Markov chain. A Markov chain is a very general term. Just because the simplest ones are stupid, doesn't mean there aren't smarter ones.

A Markov chain is a model where: (1) there is a state, (2) the probabilities of the next state depend only on the current state.

That could describe anything from a wet fart to a deterministic computer to GPT-3 to the human mind to quantum mechanics (the real one, not a simulation).

If I unplug the Internet and all USB devices, my computer is a Markov chain with at least several trillion bits of state, so 2^(several trillion) possible states. And there is one next state, which has a probability of 1, and all other possible states have a probability of 0. That's a Markov chain.

GPT models choose the next word probabilistically, with the probabilities chosen by feeding the previous N words into a neural network. That sounds like a Markov chain to me!

A typewriter vs. a machnine that recites paragraphs of shakespeare are two different things.

Neither of them unbind the content from the original license, though.

When I press 1 key and it plays copyrighted music, that is the piano's fault.

Even if hypothetically there was such a strange bug in your piano and you decided to exploit it by recording copyrighted music and redistributing it, you would be accountable for it, not a piano.

This analogy train went too far, don't you think? All examples that I've seen on Twitter require quite an intentional manipulation by human for Copilot to produce something copyrighted. It does not recite Linux code by pressing 1 key.

If you have an electronic piano that requires a complex series of button pushes to produce copyrighted music - that's still a copyright violation. Copyright law has no notion that the difficulty of reproducing copyrighted content effects the fact of a violation.

> an electronic piano that requires a complex series of button pushes to produce copyrighted music

Surely a judge presented with the "complex series of button pushes," otherwise known as playing an instrument, would hold the player accountable for any infringement and not the piano?

These analogies have gone so far off the rails that I can't tell which side this thread is arguing for by now ;)

I think the whole swirling discussion is a little confused because there are potentially two "ends" where infringement could happen, and different people are talking about each. And the article covers both.

One end is GitHub's, at the input: Copilot's "database" was initialized from code that GitHub does not have copyright to. The contention at this end is that they are ignoring the licenses that would grant them the right to use that code.* The article, GitHub, and others assert that there's no copyright issue for creating a database of this kind (a machine learning model).

The other end is the the developer taking Copilot's output. The article seems to take the (absurd IMO) position that there's also no copyright implications here, because the output is not copyrightable at all.

*And personally this is the side that concerns me most.

If you have a piano that plays copyrighted music when you press a single key, isn’t that the piano maker’s fault?

Edit - googling, the history of player pianos vs copyright is interesting

https://slate.com/technology/2014/05/white-smith-music-case-...

https://www.techdirt.com/articles/20100712/18325210185.shtml

It could end up being tried as being both the piano maker’s and player’s fault.

The former for making and selling it, the latter for buying and using it.

Just like counterfeit goods.

No?

Depends, I think. Generally, I think accidental copyright infringement by an unsuspecting person who plays the song believing they bought it legally would not be held accountable. But if they knew the song was copyrighted and not legally licensed by the piano manufacturer, and they assembled a bunch of friends to listen to the songs, or resold them for money, then yeah perhaps. Most importantly in context, though, the piano will almost certainly not be fined or taken to court. ;)

Funny you should say that, as there is a direct line connecting player pianos in the 19th century to copyright law in the 21st:

https://en.m.wikipedia.org/wiki/Mechanical_license

Reply on news.ycombinator.com