All public GitHub code was used in training Copilot • Hacker News Board

Re: All public GitHub code was used in training Copilot

#711

Post by qayxc » Sat, Jul 10, 2021, 12:00 AM UTC

skinkestek wrote: post #668

qayxc wrote: post #661

Earlier quoted context omitted.

> ... will you defend it just because I claim it is a tech preview? That's a straw man argument and you know it. Code snippets are in no way shape or form comparable to entire software products and CoPilot neither installs anything nor is its intention to knowingly violate licences or copyright law. Disingenuous straw manning like this doesn't help the discussion and only serves to distract from actual issues.

> That's a straw man argument and you know it. It is absolutely not in my opinion and that particular idea did not cross my mind at all so the idea that I knew it is patently double false. But let me try to be constructive here and be even more precise: Would it be OK if I launched a tech preview of my AI poem writer companion that would copy lines but also complete stanzas from famous poets, rock bands and singer-so…

> Would it be OK if I launched a tech preview of my AI poem writer companion that would copy lines but also complete stanzas from famous poets, rock bands and singer-songwriters?

Yes it would be if it only happened ~0.1% of the time and if quoting verbatim wasn't the intended function of the system but merely a side-effect. In fact, that's what artists sometimes do deliberately.

It's what happens with other GANs as well and all that needs to happen is to educate users about the possibility of this. As long as you don't take ownership of the output produced by your AI (and neither do Microsoft), it's at the discretion of the user what they use the generated content for and in which context.

It has been demonstrated that training data can be extracted from any large NLP model [0] so this wouldn't come as a surprise either.

[0] https://arxiv.org/abs/2012.07805

https://towardsdatascience.com/openai-gpt-leaking-your-data-...

Re: All public GitHub code was used in training Copilot

#712

Post by remram » Sat, Jul 10, 2021, 12:38 AM UTC

zohch wrote: post #649

remram wrote: post #559

Earlier quoted context omitted.

If I upload somebody else's GPL code to GitHub, I also can't grant to GitHub the (implicit) legal rights to use that code in Copilot, because they are not mine to give. I could previously mirror GPL code, because the GPL granted me the rights I need to grant GitHub as part of their ToS; but if they change their ToS, or if the meaning is changed by them adding vastly different features to their Service, this becomes a…

Can you explain what limitation in GPL would prevent someone from using it as training data? Also, if you are not allowed to upload GPL to GitHub, seems like the right answer is don't.

GPL does not prevent someone from using it as part of something else, so long as that other thing abides to the terms too. In particular, GPL and many open-source licenses require attribution. The fact that Copilot spits out code from other places without attribution clashes with that limitation.

Whether you're allowed to upload GPL code to GitHub or not depends on whatever their Service is at the moment, since the terms say you grant them all the rights "necessary to provide the Service".

Re: All public GitHub code was used in training Copilot

#713

Post by serf » Sat, Jul 10, 2021, 1:58 AM UTC

mikewarot wrote: post #681

colordrops wrote: post #643

Earlier quoted context omitted.

You don't encrypt your data before uploading to backblaze?

Oh heck no, I never encrypt data. I run windows. It can't ever be secure, anyone who wanted to hack me could. Scrambling the data really makes things worse as any accident requiring recovery of my data is also probably going to lose the encryption key. The only time I ever lost any significant chunk of data (a persons lifetime set of photos!) was because Windows encrypted data at rest, and thus it couldn't be recover…

> any accident requiring recovery of my data is also probably going to lose the encryption key.

... why?

i'd hate encrypting too if I threw away all best-practices regarding it -- losing a key with the failed system is a "problem exists between chair and keyboard" type of issue.

Encryption protects your data from yourself, from your adversaries, from serendipitous grey-moral types, and from the prying eyes of over-zealous data-collection conglomerates.

You seem experienced in the field, so I won't presume what your best practices are -- but to be enthusiastic against encryption is a form of cheer-leading that I think I cannot ethically support; the longer I live and the more pervasive companies get to be with their data collection policies then the more powerful and required tools like encryption seem to become.

Re: All public GitHub code was used in training Copilot

#714

Post by SonnyTark » Sat, Jul 10, 2021, 3:13 AM UTC

I just don't get this, it's ai not a search engine, unless we deliberately bait it, it won't spit out verbatim code snippets. I'd also use all public code in github if I wanted to train a similar tool. Furthermore, tabnine has been doing the same for years and not a single dramatic statement about it.

This simply feels like anti-Microsot people flocking to what they see as some exposed Microsoft flesh for social media biting.

Re: All public GitHub code was used in training Copilot

#715

Post by jcheng » Sat, Jul 10, 2021, 7:03 AM UTC

ThrowawayR2 wrote: post #690

jcheng wrote: post #525

Earlier quoted context omitted.

Only if you agree that copyleft licenses are also stupid; without copyright, there's no way to prevent companies from making closed-source forks of code you wrote and intended to stay open.

The whole point of copyleft was as a stepping stone to get to RMS's four freedoms ( https://www.gnu.org/philosophy/free-sw.en.html ) which effectively eliminates copyright for software.

Freedom 1: “Access to the source code is a precondition”

With no copyright/copyleft, how do you enforce the rule that derived works must provide access to the source code? I’ve never heard that copyleft was a stepping stone—rather, it’s the stick that fully realizes the four freedoms.

Re: All public GitHub code was used in training Copilot

#716

Post by hankman86 » Sat, Jul 10, 2021, 7:08 AM UTC

fleddr wrote: post #551

To me, the particular use case and whether it is fair use or not, is of minor interest. A far more pressing matter is at hand: AI centralization and monopolization. Take Google as an example, running Google Photos for free for several years. And now that this has sucked in a trillion photos, the AI job is done, and they likely have the best image recognition AI in existence. Which is of course still peanuts compared…

I don’t think that your premise concerning “planetary-scale AI” (and the ability to pull it off) holds up. If Google and Microsoft are so dominant and had such an insurmountable head start, why are we seeing such an enormous number of AI startups? In fact, there are countless startups busy figuring out how to make AI work for software development. I’d even argue that copilot was not that expensive to build. I very much doubt that GitHub (or Microsoft for that matter) had a huge team working on this or has spent such a vast amount on hardware resources that they’d outcompete the rest of the market by virtue of their cash reserves. Any decently funded startup should be able to finance such as effort. Especially since in this case, the training data is cheap (and legal) to access for anyone.

Where Microsoft does have an “unfair advantage” is in their marketing and sales firepower. Replicating their B2B and B2C sales channels is indeed very expensive. GitHub will be able to monetise Copilot by some upselling campaign. Then again, startups regularly manage to break into markets that are supposedly locked down by the likes of Microsoft.

Re: All public GitHub code was used in training Copilot

#717

Post by hankman86 » Sat, Jul 10, 2021, 7:11 AM UTC

spyckie2 wrote: post #619

fleddr wrote: post #551

To me, the particular use case and whether it is fair use or not, is of minor interest. A far more pressing matter is at hand: AI centralization and monopolization. Take Google as an example, running Google Photos for free for several years. And now that this has sucked in a trillion photos, the AI job is done, and they likely have the best image recognition AI in existence. Which is of course still peanuts compared…

The thought that data monopolization will be a moat against competitors is actually argued against by VC firms specializing in AI companies, who claim that after a certain amount of data (which is accessible to most people) the additional data isn't going to improve the model much. https://a16z.com/2019/05/09/data-network-effects-moats/ https://a16z.com/2020/02/16/the-new-business-of-ai-and-how-i...

And in case of Copilot, the training data isn’t a moat anyhow. Last I looked, everyone could freely access GitHub public repositories.

Re: All public GitHub code was used in training Copilot

#718

Post by hankman86 » Sat, Jul 10, 2021, 7:13 AM UTC

visarga wrote: post #678

fleddr wrote: post #551

To me, the particular use case and whether it is fair use or not, is of minor interest. A far more pressing matter is at hand: AI centralization and monopolization. Take Google as an example, running Google Photos for free for several years. And now that this has sucked in a trillion photos, the AI job is done, and they likely have the best image recognition AI in existence. Which is of course still peanuts compared…

They are not hoarding the latest results, except for a few cases where the general public is a year behind their secret sauce. Take a look at the huge zoo of planetary-scale models that are published by the big companies and universities (HuggingFace, https://modelzoo.co/ , ...) The problem with the huge models like GPT-3 is that they are too expensive even to run by regular people, not train.

Regular people yes, but no problem for decently funded startups.

Re: All public GitHub code was used in training Copilot

#719

Post by hankman86 » Sat, Jul 10, 2021, 7:19 AM UTC

cbolton wrote: post #691

fleddr wrote: post #551

To me, the particular use case and whether it is fair use or not, is of minor interest. A far more pressing matter is at hand: AI centralization and monopolization. Take Google as an example, running Google Photos for free for several years. And now that this has sucked in a trillion photos, the AI job is done, and they likely have the best image recognition AI in existence. Which is of course still peanuts compared…

Well of course only the huge companies can develop products that require enormous resources. But I'm not too worried here because everyone gets access to larger datasets every year, and it gets cheaper to process every year, so whatever Microsoft or Google is capable of doing now, smaller companies will be capable of doing in a few years.

It’s also a huge call for innovation. When a student learns to code, (s)he doesn’t need to analyse millions of Git repositories to get good at it. Throughout their entire career most developers will probably only see comparatively little code. Perhaps the equivalent of the Linux kernel, if that. And yet, we’re able to learn from the little we see and get reasonably good at coding. It even stands to reason how much better one gets by reading more code (most of which is pretty crappy anyway).

Re: All public GitHub code was used in training Copilot

#720

Post by hankman86 » Sat, Jul 10, 2021, 7:20 AM UTC

sdevonoes wrote: post #669

fleddr wrote: post #551

To me, the particular use case and whether it is fair use or not, is of minor interest. A far more pressing matter is at hand: AI centralization and monopolization. Take Google as an example, running Google Photos for free for several years. And now that this has sucked in a trillion photos, the AI job is done, and they likely have the best image recognition AI in existence. Which is of course still peanuts compared…

We can decide to stop using some (or most of) Google services. It's hard, but it's not that they are pointing us with a gun in order to use their services, right? Sure, for the cases when one cannot escape Google, use it; but for the rest of scenarios? It's all about tradeoffs: Can I live without YouTube? Can I live with DuckDuckGo (Google Search is "better" but I don't mind)? etc.

Google search would be hard to replace. In fact, if Google search was turned off over night, the world would probably see a major economic downturn, caused by a sudden drop in productivity.