Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Ownership of AI-Generated Code Hotly Disputed

spectrum.ieee.org

Post reply on HN
Print view View original

171–180 of 205 posts

Re: Ownership of AI-Generated Code Hotly Disputed

#171

Post by forrestthewoods »

falcolas wrote: post #108
Filligree wrote: post #83

Earlier quoted context omitted.

The AI learns to generate watermark-like things, because those exist on a large fraction of its inputs. It doesn't mean the rest of the output existed in the training set. It's perfectly capable of generating a completely novel picture, then slapping a watermark on it.

My point is that the watermark is deterministic. It is taken whole-cloth from an input, and reproduced as-is on an output. Thus, it is able to be attributed. Stable diffusion is more like a paintshop artist who grabs bits and pieces of other art and melds them together, and less like a painter who creates from their imagination.

> It is taken whole-cloth from an input, and reproduced as-is on an output.

I respectfully disagree with this claim. It’s not as-is. It’s remarkably similar. That’s a big difference.

> Stable diffusion is more like a paintshop artist who grabs bits and pieces of other art and melds them together, and less like a painter who creates from their imagination.

I also disagree with this. The uncomfortable truth, imho, is that what Stable Diffusion does is FAR closer to what human artists do than we’d like to admit.

Human artists are perfectly capable of reproducing copyright infringing images. They just generally choose not to for legal/moral reasons.

Re: Ownership of AI-Generated Code Hotly Disputed

#172

Post by andybak »

wongarsu wrote: post #43
andybak wrote: post #23

Earlier quoted context omitted.

Algorithms aren't copyrightable as far as my understanding goes and this is where it gets murky. The expression is copyrightable, the algorithm isn't (except when it is...) My personal feeling is that we've gone too far down the road of assigning IP to code and we should be rolling it back. I'd hate for the current AI boom to trigger an extension of current IP law.

The abstract algorithm isn't covered by copyright, but can be patented (lots of asterisks here about software parents). But the exact lines of code you write are covered by copyright. If you tell Github Copilot to write a fast inverse square sum algorithm it's allowed to reproduce the idea of the famous quake3 algorithm, but if it produces the same lines and comments as those in the quake 3 source code then that's a…

> But the exact lines of code you write are covered by copyright.

I believe it's "only if they are expressive" i.e. code that is pretty much what anyone would have written in that situation isn't copyrightable.

I think this is the part where we should be much stricter in applying the principle. Similar to "novel and non-obvious" in patent law, it's become diluted in favour of IP overreach.

Re: Ownership of AI-Generated Code Hotly Disputed

#173

Post by AlotOfReading »

visarga wrote: post #158
AlotOfReading wrote: post #45

Earlier quoted context omitted.

Leaving aside what I think IP law ought to look like, LLMs literally train on (and reproduce!) the fixed, copyrighted expression. They aren't producing a new work based on abstract knowledge of algorithms that happens to share a particular expression. It's not at all obvious to me why the training set copyright wouldn't nominally flow through the model, even if that seems impossible to actually implement in practice.

Why is everyone fixating on 1% of the outputs. 99% are sufficiently distinct. It replicates code as an unintended side effect, but there are mitigations.

If you take all the words from Harry Potter and shuffle the words together in a random order, the output will be distinct but the work as a whole is probably still derivative. If you do that for 500 book series, you still haven't generated a new work, you've simply created a legal morass of derivative copyrights.

ChatGPT is just a really impressive word-blender. All of those "sufficiently distinct" snippets are in a meaningful sense recombinations of other copyrighted snippets it was trained on. When you're in a very sparse part of the space or have a single work that's present many times in training like fast inverse square root, the model may even replicate recognizable chunks of the training set. It could be that this is sufficiently transformative that the output is not derivative, but as I said it's not obvious to me.

Re: Ownership of AI-Generated Code Hotly Disputed

#174

Post by jkaplowitz »

monsieurbanana wrote: post #32
jkaplowitz wrote: post #21

Earlier quoted context omitted.

It is an accurate statement of the law in the US, but "purely functional expression" is not used in the sense of a mathematical function. It basically means expressions that solely reflect the way you need to do the thing to accomplish the function you are trying to achieve, without any significant creative element. For example, if I asked you to write Java code to print the sum of two integer variables named a and b…

It seems like a very blurry line for any non-trivial piece of code.

It is a blurry line, like many legal tests. Human judges and/or juries apply this test in case of litigation, as advised by whatever experts the parties are able to to offer.

Re: Ownership of AI-Generated Code Hotly Disputed

#175

Post by leereeves »

belorn wrote: post #142
Permit wrote: post #119

Earlier quoted context omitted.

> Sorry but terms of service don't give you a blanket license to re-purpose someone else's copyrighted work at your pleasure. That’s correct. It’s also not what happened here. I don’t violate your copyright when I scrape your web page. I don’t violate your copyright when I train a model on a web page that I scraped. I violate your copyright only if I use that tool to produce code that violates your copyright. This is…

If the training data discussion ends on that note than there is exist a large upside. Youtube and other video hosting services has a massive collection of the worlds images, music and sound. Anyone who who want to build an AI will have access to practically unlimited amount of training data, meaning the barrier to entry will be quite low.

I doubt YouTube will let anyone download more than a trivial fraction of the videos they host, but it has been used for some specialized projects like Minecraft AI (70,000 hours of video).

https://openai.com/blog/vpt/

Re: Ownership of AI-Generated Code Hotly Disputed

#176

Post by wahnfrieden »

somrand0 wrote: post #86
wahnfrieden wrote: post #36

Earlier quoted context omitted.

or abolition but to do that without any reasonable social support nets is absurd

but what does it mean to destroy the concept of ownership (i.e. the notion of property) if we take that > The only way to truly own something, is to either share it or destroy it. could I argue that then, we would be taking ownership of the concept of ownership?? ahahha... I think this is kind of funny. And I'd admit that it's not very helpful to the goal of revising the very concept of ownership. my own parent comme…

private property successfully didn't exist in various societies, including larger scale societies and ones that had discovered agriculture, and not only under communist regimes. it does mean to share, without ownership.

Re: Ownership of AI-Generated Code Hotly Disputed

#177

Post by danShumway »

visarga wrote: post #159
randombits0 wrote: post #9

Code cannot be owned. A creative expression may be copyrighted. Purely functional expressions may not be copyrighted. The output of a trained AI is insufficiently creative to be copyrighted. Only humans can hold a copyright. Now with all that, there really isn’t anything here to get worked up over.

> The output of a trained AI is insufficiently creative to be copyrighted. Only humans can hold a copyright. That's overgeneralisation. A language model alone, yes, is just derivative. But a language model trained on solving problems with reinforcement learning can surpass humans. For example AlphaGo and AlphaTensor are models that learned from running simulations.

It's not a question of whether the AI can be creative in a philosophical sense, current copyright law (in the US) only protects human creativity -- AI and non-human animals can't hold IP at all, nor can they delegate it to someone else.

Under current copyright law, if you built a model that surpassed humans and trained itself on simulations and it wasn't making derivative works but instead producing purely original output, the resulting output would not be copyrightable.

In fact, under current copyright law, if you built an AI that was sapient, the creative work it made would not be copyrightable -- unless you literally got the law to grant it legal personhood or something.

Moving AI in a direction of being more fundamentally creative makes this harder, not easier. The legal defense of copyright in cases involving AI is that the creativity is coming from the human using the AI, not the AI itself. As the AI gets better at generating usable output with less directing and micromanaging, that argument gets weaker, not stronger. The question isn't whether or not the AI is creative or produces novel output, the question is whether a human exhibited enough creativity making the work for it to meet the standard of gaining copyright protections -- which, traditionally, the US copyright office has usually held isn't true for AI-generated content.

Re: Ownership of AI-Generated Code Hotly Disputed

#178

Post by janalsncm »

dahart wrote: post #54
wrs wrote: post #16

“…modify[ing] its AI model so that it traces attribution and gives credit to the original authors of the code, adding the associated copyright notices and license terms in the process…Biderman says is technologically feasible.” Is it really feasible? What does “traces attribution” even mean here? It’s not emitting “code”, it’s emitting individual tokens that each were found throughout the input corpus. The “code” is…

Why wouldn’t it be feasible? (Maybe this depends on what you mean by ‘feasible’.) There’s no technical reason you can’t back-track the weights and make a list of which tokens from which training data were sampled. The list might be long, it could be impractical, but that has little bearing on whether it’s technically possible, right? The problem here happens when the same source is sampled for many tokens in a row be…

The model isn’t sampling from training data during inference. It’s sampling from a latent space created during training, which used all of the training data to create it.

Fundamentally, the presence of a next token in some of the samples in the corpus is just as informative as the absence of that token from other samples. You can’t cite one without the other.

If you wanted to give attribution you’d need to list the entire training corpus.

Re: Ownership of AI-Generated Code Hotly Disputed

#179

Post by oehpr »

TchoBeer wrote: post #95
oehpr wrote: post #73

Earlier quoted context omitted.

>A closer parallel would be the industrialisation of painting duplication But they directly reproduce the source material. AI art they clearly does not. The luddites seem a better parallel when it comes to scale. Where a machine comes along capable of producing in much higher quantities and in much greater efficiencies. Or perhaps photography? Also a fear of scale. For a long time photographers were not considered ar…

AI art they clearly does not. It does not 1-1 recreate source material, but if we didn't have human artists anymore and replaced them all with the current stock of AI we would have no more developing art movements. It is wholly uncreative in a way that humans are not, and it does not understand what humans would find interesting, only what humans have already made. This makes it a pretty good tool for some amateurish…

>it is wholly uncreative in a way that humans are not

The work I've seen manipulating, weilding, and modifying this new tool, has been very creative In painting, reprompting, texture and uv map generation, dreambooth and dedicated fine tuning. Integrating these workflows into human directed procedures to produce in hours what would take days, sometimes months, sometimes not even within the scope of human capability.

If you think the only thing here is people jibbering at a black box and posting 512*512 images, I can see why dumping the entire thing would seem reasonable.

Re: Ownership of AI-Generated Code Hotly Disputed

#180

Post by oehpr »

inkcapmushroom wrote: post #139
oehpr wrote: post #73

Earlier quoted context omitted.

>A closer parallel would be the industrialisation of painting duplication But they directly reproduce the source material. AI art they clearly does not. The luddites seem a better parallel when it comes to scale. Where a machine comes along capable of producing in much higher quantities and in much greater efficiencies. Or perhaps photography? Also a fear of scale. For a long time photographers were not considered ar…

>AI art they clearly does not. https://twitter.com/kortizart/status/1588915427018559490 I would think these are close enough that any human that produced that output could be claimed to have plagiarized.

ugh, how tedious.

Famous images prevelant in the dataset that are trained against many times and the directly requested are produced as is, because that's litterally what was requested and the model was trained against them.

There's only one way to draw The Mona Lisa. These AI's are not used for this purpose, nor is any of this core to their functioning. And I have to assume you know that.

Post reply on HN