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

81–90 of 205 posts

Re: Ownership of AI-Generated Code Hotly Disputed

#81

Post by visarga »

falcolas wrote: post #69
rymate1234 wrote: post #60

Earlier quoted context omitted.

> It's been clearly displayed that these tools are emitting verbatim copies of existing code (and its comments) in their input. Which makes sense when you consider that the sort of code that is getting reproduced verbatim is usually library functions which developers may copy and paste verbatim comments and all into their project, especially when you prompt the AI with the header of a function that has been copied an…

So that should make it easy to attribute, yes?

I think harder, as it is spammed around in all directions. It's easier to attribute a unique piece of code that appears in a single repo.

But boilerplate functions don't deserve copyright protection as they are not creative. Can I copyright print('hello world!') if I post it in my repo? Do I deserve a citation from now on?

Re: Ownership of AI-Generated Code Hotly Disputed

#82

Post by falcolas »

wrs wrote: post #47
falcolas wrote: post #39

Earlier quoted context omitted.

> It’s not emitting “code”, it’s emitting individual tokens ... The “code” is the arrangement of those tokens, but that is determined by the weighting of the whole network This theory of operation is not borne out in reality. It's been clearly displayed that these tools are emitting verbatim copies of existing code (and its comments) in their input. It's even being seen in image generation, where NatGeo cover images…

The output may be recognizable by humans as being “the same”, but an ML classifier can look at what is “obviously” a cat and classify it as a turtle. So while your statement is accurate, it doesn’t explain how “traceability” could really work—what is the actual mechanism of “tracing”?

There's a whole field of work around AI Observability - tracing how data moves through a model. It's intended to help identify biases in the input data, and to help answer the question of "why did it make this decision".

And if you can trace how data moves through the model, you can identify "that output came from these inputs" and add the associated metadata.

Re: Ownership of AI-Generated Code Hotly Disputed

#83

Post by Filligree »

falcolas wrote: post #66
forrestthewoods wrote: post #53

Earlier quoted context omitted.

> It's even being seen in image generation, where NatGeo cover images are reproduced in their entirety or where stock photo watermarks are emitted on finished images. Can you cite sources? I’ve heard this claim repeatedly but have yet to see a good example.

https://youtu.be/kqPKNksl9hk?t=80 https://news.ycombinator.com/item?id=33061707 <- Watermarks (there's been a few linked here on HN)

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.

Re: Ownership of AI-Generated Code Hotly Disputed

#84

Post by belorn »

When Copilot was released the copyright discussion focused exclusively on code, but now we see very similar discussion around images with stable diffusion and the sister project unstable diffusion. When Copilot do reach the courts there will be some indication on how courts view author consent when it comes to training material. After that we might then see court cases for each form of media (images, video, text, sound), and also in their each unique contexts (books, online videos, porn, stock databases, and so on).

Re: Ownership of AI-Generated Code Hotly Disputed

#85

Post by sillysaurusx »

visarga wrote: post #75
sillysaurusx wrote: post #56

Earlier quoted context omitted.

Well yes, if you’re asked to memorize someone’s code, you’ll get surprisingly far too. The fact that models can do this isn’t evidence of anything. It’s a capability (or “tendency” if you overfit too much). I think it’s pretty obvious that if you train a model to reproduce existing work in its entirety, it fails the “sufficiently transformative” test and thus loses legal protection. But there’s nothing stopping you f…

Are you sure you can do that? Because it is trivial to paraphrase code and text.

Sure. I’ve done it for Bel.

  (def no (x)
    (id x nil))

  (def atom (x)
    (no (id (type x) ‘pair)))

  (def some (x f)
    (if (no x) nil
        (f (car x)) x
        (some (cdr x))))

  (def all (x)
    (if (no x) t
        (f (car x)) (all (cdr x))
        nil)))
I don’t even have to pull up bel.bel to know that those are almost perfect replicas. I typed it on my iPad.

EDIT: as far as I can tell, the only diff is that all comes before some. https://sep.yimg.com/ty/cdn/paulgraham/bel.bel?t=1595850613&

I could have kept going through most of the implementation.

Re: Ownership of AI-Generated Code Hotly Disputed

#86

Post by somrand0 »

wahnfrieden wrote: post #36
somrand0 wrote: post #2

ownership is a concept in dire need of revision

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 comment is already controversial with many replies and exactly 0 points (at the time of this response)

Re: Ownership of AI-Generated Code Hotly Disputed

#87

Post by Filligree »

varjag wrote: post #50
Filligree wrote: post #26

Earlier quoted context omitted.

It sounds like nonsense. The most plausible solution would be to provide credit to every single author whose code was used in the original training set; of course, that would run into gigabytes just for the credits.

And…? I think my last Xcode update was 20+ Gb.

Personally, I don't want minimalistic programs to start taking 20+GB! Xcode's bloat isn't a goal to strive for.

If such a giant list of acknowledgements were to be included, would it even help anyone, in any way? It sounds like pure inefficiency.

Re: Ownership of AI-Generated Code Hotly Disputed

#88

Post by nonrandomstring »

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…

The answer is of course it's possible, so long as you have a couple of gigabytes spare for the acknowledgements page. Any construct worth attributing will have roots in billions of parameters.

Re: Ownership of AI-Generated Code Hotly Disputed

#90

Post by somrand0 »

scotty79 wrote: post #6

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

considering that information is something static (which doesn't change) that describes (is about) how something else is changing

I think that indeed, it is the inherent nature of information to radiate itself, i.e. to share (to shine, to spread)

Post reply on HN