2,887 karma · joined October 23, 2019
Doesn't google routinely contribute to ffmpeg? Certainly there's a lot of commits from `@google.com` email addresses here: https://git.ffmpeg.org/gitweb/ffmpeg.git?a=search&h=HEAD&st=...
So when the xz backdoor was discovered, you think it would have been better to sit on that quietly and try to both wrest control of upstream away from the upstream maintainers and wait until all the downstream projects had reverted the changes in their copies before making that public? Personally I'm glad that went public early. Yes there is a tradeoff between speed of public disclosure and publicity for a vulnerability, but ultimately a vulnerability is a vulnerability and people are better off knowing there's a problem than hoping that only the good guys know about it. If a Debian bug starts tee-ing all my network traffic to the CCP and the NSA, I'd rather know about it before a patch is available, at least that way I can decide to shut down my Debian boxes.
That is, you'd rather a world where Google either does know about a vulnerability and refuses to tell anyone, or just doesn't look for them at all, over a world where google looks for them and lets people know they exist, but doesn't submit their own fix for it.
Why do you want that world? Why do you want corporations to reduce the already meager amounts of work and resources they put into open source software even further?
So how long should all bug reporters wait before filing public bugs against open source projects? What about closed source projects? Anyone who works in software knows to ship software is to always have way more things to do than time to do it in. By this logic, we should never make bug reports public until the software maintainers (whether OSS, Apple or Microsoft) has a fix ready. Instead of "with enough eyeballs, all bugs are shallow" the new policy going forward I guess will be "with enough blindfolds, all bugs are low priority".
Google did more than what is "strictly legal or required", and what they did was submit a good and valid bug report. But for some reason we're mad because they didn't do even more. Why? The world is objectively a better place for having this bug report, at least now people know there's something to address.
In no reasonable reading of the situation can I see how anything Google has done here has made things worse:
1) Before hand, the bug existed, but was either known by no one, or known only by people exploiting it. The maintainers weren't actively looking at or for this particular bug and so it may have continue to go undiscovered for another 20 years.
2) Then Google was the only one that knew about it (modulo exploiters) and were the only people that could take any steps to protect themselves. The maintainers still don't know so everyone else would remain unprotected until they discover it independently.
3) Now everyone knows about the issue, and are now informed to take whatever actions they deem appropriate to protect themselves. The maintainers know and can choose (or not) to patch the issue, remove the codec or any number of other steps including deciding it's too low priority in their list of todos and advising concerned people to disable/compile it out if they are worried.
#3 is objectively the better situation for everyone except people who would exploit the issue. Would it be even better if Google made a patch and submitted that too? Sure it would. But that doesn't make what they have done worthless or harmful. And more than that, there's nothing that says they can't or won't do that. Submitting a bug report and submitting a fix don't need to happen at the same time.
It's hard enough convincing corporations to spend any resources at all on contributing to upstream. Dragging them through the mud for not submitting patches in addition to any bug reports they file is in my estimation less likely to get you more patches, and more likely to just get you less resources spent on looking for bugs in the first place.
"Given enough eyeballs, every bug is shallow" right? Well, Google just contributed some eyeballs, and now a bug has been made shallow. So what's the actual problem here? If some retro game enthusiast had filed the same but report would that be "abusing" the maintainer's time? I would think not, but then we're saying that a bug report can be "abusive" simply by the virtue of who submits it. And I'm really not sure "don't assign employees to research bugs in your open source dependencies and if you do certainly don't submit bug reports on what you find because that's abusive" is the message we want to be sending to corporations that are using these projects.
Edit: And also, how is anyone supposed to know they should compile the codec our unless someone makes a bug report and makes it public in the first place?
And we learned the limits. Broadly verifiable, non-controversial items are reasonably reliable (or at least no worse than classic encyclopedias). And highly technical or controversial items may have some useful information but you should definitely follow up with the source material. And you probably shouldn’t substitute Wikipedia for seeing a doctor either.
We’ll learn the same boundaries with AI. It will be fine to use for learning in some contexts and awful for learning in others. Maybe we should spend some energy on teaching people how to identify those contexts instead of trying to put the genie back in the bottle.
The protocols themselves might not but as a warning to people looking for “matter” as an indicator they can have local only control, apparently the matter spec doesn’t require local only setup. I bought Honeywell’s new matter thermostat and in order to get the QR code and keys you need to register it to a matter controller, you first have to download their app and connect the thermostat to their cloud, so that you can get the keys from the app. So the matter capabilities are still useless
Also instructive would be the IBM BIOS clone, it is perhaps true that the "IBM Compatibles" killed the market that existed for IBM machines at that moment in time, but it's also true that it opened whole new markets, both to the clone makers and the ancillary businesses, but also arguably IBM themselves.
3d printing and Arduino are probably other examples where "counterfeits" might have shrunk the market for the originals (Prusa is notably reducing how open their designs are, and Arduino themselves are not the healthiest, modulo being owned by Qualcom now), but the market for Aruduino projects and ancillary supplies and certainly the market for 3d printers is massively healthy, and arguably both are healthier than if Arduino or Prusa (or really Reprap) were the single and sole providers of their products.
And I think art has an even stronger bulwark in that a lot of the value of a given "art" comes not from the art itself, but from the artist. It's very possible many famous artist's works were actually made by their apprentices, but until someone proves that, the art will continue to have value as an original work of the artist. But art is also a dime a dozen (or less). The internet is full of free or dirt cheap art and today you can go on fiver or mechanical turk and commission any number of artworks for probably less than your day's wages. But no one is buying tickets to your Fiver concert. No one buys $1k per plate dinners at Deviant Art gallery showings. But they will pay many thousands of dollars for a piece of artwork that might destroy itself because the person who produced that artwork is named Banksy.
To the best of my understanding of the law, making whole copies of a work for other people to then make fair use of is generally not considered fair use. There are some narrow cases where it has been; the two that come to mind are a case (whose name escapes me) that found reproducing an entire editorial for commentary and discussion on a discussion forum was fair use and a handful of cases regarding whether a print shop is liable for copyright infringements if a customer orders a copy made of a work.
Likewise the person downloading it can't be making fair use of the work unless they already have a copy of the work they have legally obtained. This was a key point of the Anthropic case, the use of legally purchased or acquired books was fair use, the use of the copies obtained only via piracy were not. I don't recall if Alsup addressed it specifically in his ruling, and to the best of my knowledge otherwise no rulings have ever been made on the legality of "I own a legal copy of this work, and I am acquiring a copy made from someone else's legal copy for fair use purposes" (this is often the theory behind why downloading ROMs if you own the game already could be legal).
> How is it unlike borrowing a book from the library and quoting an excerpt from it in your criticism?
Libraries have been something publishers have hated for a long long while. They routinely try to extort massive amounts of money out of them (see separate pricing for "library" copies of media, maximum lending count limits on "library" ebooks etc). One of the reasons is that a library as you point out sort of puts a big dent in the "fair use derives from ownership / first sale doctrine" idea. Again, to the best of my knowledge, cases involving using library copies just tend to get their own set of rules, and are also probably just fewer and farther between because it's unlikely you can convince a library to let you just scan all of their books regardless of the legality of you doing so.
As for the partnership thing, my point was mostly that any legal entity that can jointly own property would have to be the owner of a work for people within that entity to distribute copies between each other, and that once a person leaves the legal entity they could no longer keep or obtain copies of works owned by the entity.
Not wealth specifically, but income is probably as good of a controlling factor.
Surely making copyright even stronger (and even expanding it to cover style as some have argued in response to the Ghibli style stuff) will have no unintended consequences going forward into a future where more and more technology is locked down by major manufacturers with a strong incentive to use and abuse IP law to prevent competition and open alternatives... right?
> "stealing the work of every single artist, living and dead, for the sole purpose of making a profit"
is begging the question. The phrase "stealing" inherently presumes a crime has been committed when that is the very thing up for debate. The note that this is "for the sole purpose of making a profit" asks the reader to infer that one can not make a profit and also engage in fair use, and yet that is clearly not true. And that puts aside entirely that you're not referencing a "specific case" here, but generalizing to all forms of AI training.
We can start by examining the "stealing" phrasing, noting that up until fairly recently it was an axiom in a lot of "hacker" places that "copying is not theft". It feels somewhat contradictory to me to find that now that it's our work that's being copied, suddenly a lot of hackers are very keen on calling it theft.
We can note that a lot of our careers rest on the idea that copying is not prima facie a crime or theft. Certainly IBM wanted it to be so, and yet their BIOS was cloned and many of our careers were kickstarted (or at least kicked into high gear) by the consequences of that. Or how about the fact that despite what Apple would really like to be true, you can't stop people from copying your really cool idea from looking at it and making something similar. They couldn't win that battle against Microsoft, and they couldn't win it against Google or Samsung either.
We can talk about the fact that there is a lot of disagreement that the dead should have any IP rights at all. Notably again many "hacker" spaces were quite vocal about how the various "lifetime+" IP laws were actually detrimental to creativity and society. Until recently it was the Walt Disney corporation in their pursuit of making a profit that was one of the largest advocates for copyrights to be extended long after the creator was dead and gone.
We could also talk (again) about how many forms of fair use are pursued for "the sole purpose of making a profit". See the previous references to the IBM BIOS and the copying of Apple's GUIs. But there is also parody music like Weird Al, or Campbell v. Acuff-Rose Music. We can look also to the previously mentioned Connectix Virtual Game Station and other emulators or to the Oracle V. Google lawsuits. We can look at how TiVo's entire business model was to enable millions of people to create copies of copyrighted material every day, automatically. Or Kelly v. Arriba Soft (which is basically why you can do an image search today) and Suntrust Bank v. Houghton Mifflin (also relevant for the discussion re "the dead").
And again, most related to AI would be Authors Guild v. Google for the ingestion of thousands of copyrighted materials to in the pursuit of profit and creating the Google Scholar application. And application which I again note produces partial, but 100% faithful copies of that original work for the user and does so intentionally, something AI system actively try to avoid.
Which is why I started by asking you to walk the tech stack. If you think the law needs to be updated or you disagree with the current rulings on this matter, I can certainly understand that. So where do you propose drawing the line? What are the specific items that you feel distinguish training an AI from the various other fair use cases that have come before. Why is Google Scholar scanning in thousands of copyrighted materials with the express purpose of both making those materials searchable and displaying copies of that material ok, but an AI scanning in books with the express purpose of creating new material not ok? Why is Android, Windows, Gnome, KDE and other GUIs so very clearly copying the style (if not the works) of Apple ok, but Stable Diffusion producing new art works copying styles not?