Google is trying to patent video compression use of Asymmetric Numeral Systems
encode.ru
encode.ru
I'm probably biased because I mostly read about software patent stories on HN and other similar outlets, but it seems like 99.9% of the time they're only used for patent trolling or killing the competition.
Patents are supposed to protect the inventors by giving them the time to license and market them before the competition can swoop in and create a similar product without having to pay for the R&D.
So do people around here have "success stories" regarding the use of software patents to actually reward creativity instead of just playing the "patent cold war" game?
I worked at a startup doing cognitive radios. We were an R&D shop only--our competitors were companies like Raytheon and Lockheed-Martin that had manufacturing muscle. We patented everything, and most of it was what you'd call a "software patent" (even though in the long run it would be partially implemented in an FPGA or running on a microcontroller). It was expensive to pay PhDs to spend a year doing Matlab simulations to figure out what algorithms would work.[1] Even more expensive was then spending years testing those in the field (because lots of things that work in simulation fall flat in uncontrolled real-world environments). Copyright was pointless. Writing the code was the easy part.
I'm a bit on the fence--I think the problems with patents are overstated, but the benefits often are too. But carving out "software patents" from other kinds of patents strikes me as odd. There is no clean line--it's often easy to take something that might be done by an ASIC and move it into software and vice versa.
[1] E.g. one project involved getting disjoint listen-before-talk networks to synchronize their "listening" periods without being able to communicate with each other directly.
It shouldn't be surprising or unexpected or frowned upon that they are used to kill competition - that's basically their raison d'être.
Asymmetric Numeral Systems ([1]) is entropy coding family currently replacing Huffman and arithmetic coding in data compressors, among others, of Apple, Facebook and Google, thanks to being up to 30x faster [2]. Its author has made it public to prevent pathology of arithmetic coding, which wide use was blocked by patents for many decades ([3], [4]). However, currently others are trying to patent basic applications of ANS – including Google for AV1 video compressor (initially suggested by ANS author, who has helped them for the last 3 years: [5]) in very general patent application, to prevent others from using it in image and video compression – claims and sources: [6] [1] https://en.wikipedia.org/wiki/Asymmetric_Numeral_Systems [2] benchmarks: https://sites.google.com/site/powturbo/entropy-coder [3] https://en.wikipedia.org/wiki/Arithmetic_coding#US_patents [4] Charles Bloom comment: http://cbloomrants.blogspot.com/2015/05/05-21-15-software-pa... [5] https://groups.google.com/a/webmproject.org/forum/#!topic/co... [6] Google claims and sources: https://encode.ru/threads/2648-Published-rANS-patent-by-Stor... Clarification: the "30x speedup" is for decoding of the fastest arithmetic coding (~50 MB/s) vs of ANS (~1500 MB/s) in benchmark [2] for similar compression ratio (better than Huffman, which standard implementation: zlibh from gzip has ~300MB/s). The inventor of ANS is Jarek Duda.
If it was for the public good, the patent would be filed via a non-profit that Google can't control.
There is a tiny number of patents they've pledged not to sue offensively for: https://www.google.com/patents/opnpledge/patents/
The fact that they have not pledged most of the rest indicates that someone there is maintaining their ability to offensively sue in the future.
If you've noticed something about most patent trolls, they're has-beens. When you can't compete, you sue on patents. Right now, as one of the most valuable companies on the planet with multiple products that are monopolies, obviously that does not describe Google today.
But it does describe Google at some date in the future. And when Google's profits are crashing, and they're trying to figure out how to stay in business, some business exec is going to realize that treasure trove of patents can solve all their problems...
Couldn't they just publish the techniques by some other means? E.g. some scientific journal. I mean, a publication is strong evidence of prior art.
If you patent something, and don't have a licensing fee or anything like that, I would imagine that it would deter some patent trolls from getting a false patent and suing people using that technique. Fighting in court costs a lot of money for people like me (and probably you) so anything that keeps me out of court is a good thing.
If the patent system was working correctly there would be no need such a thing, either something is patented or it isn't. If it isn't it shouldn't be possible for a third party to patent it post-facto since there's already prior art. At least, that's the spirit of it as far as I'm aware.
We don't know what's the rationale here. For all we know, Google might even be OK if the patent is not granted to them, as long as it isn't to anyone else, either. The radio silence is to be expected when lawyercats are involved, unfortunately.
[Ex-Google employee, not a lawyer, etc.]
But every single major IEEE, ETSI, IETF, etc. standard ends up with 10 different companies getting patents on exactly the same shit.
I assumed it to mean that Jarek Duda had joined Google, but it is Jarek speaking in the thread.
I follow the data compression sphere a bit. It really seems like Google is making a mistake here. From a patent perspective, since ANS is already "prior art", it seems like it doesn't deserve a patent.
Presumably Google wants to avoid receiving a misinformed lawsuit about ANS that they would then have to fight to prove that an ANS patent shouldn't be valid. Still, not putting Jarek in the loop (and like he mentions, not engaging with his university more) is wrong.
The best takeaway:
if anybody is also afraid of Google having a very general patent for using ANS (both tANS and rANS) in image and video compression and can explain why the above claims are just a natural mix of known methods:
go to: https://efs.uspto.gov/efile/portal/efs-unregistered
select "Existing application/patent" and then "Third-Party Preissuance Submission under 37 CFR 1.290"
as "Application number" enter: 15/370840
"Confirmation number": 5111
The title of the post was changed. It used to read that they were patenting for the public good.
This is not correct. In fact, as far as i know, the code was already open sourced, and the patent would end up released as part of AV1/etc anyway.
"If it was for the public good, the patent would be filed via a non-profit that Google can't control."
It'll be released that way, so who cares who files and pays the fees?
You really don't want non-profits filing patents. Really. This is pretty much the most horrible thing i can think of. It's infinitely easier to keep them from going bad (Example: MPEG-LA) if you can make their corporate bylaws, etc, prevent them from filing patents.
>This is pretty much the most horrible thing i can think of.
>It's infinitely easier to keep them from going bad (Example:
>MPEG-LA) if you can make their corporate bylaws, etc,
>prevent them from filing patents.
? Non-profits are corporations. They have by-laws. They also have several advantages over for-profits in the case of dissolution; eg, you can set up a non-profit so that the IP they hold is not simply sold to the highest bidder.
FTR, Xiph.Org is a registered 501(c)3 non-profit, and we hold patents.
Non-profits can in fact, "go bad" the same as any other corporation.
If they can only hold patents, and not file them, you at least cut off a source of that happening.
Is it possible to work around? So is anything in the legal world.
But it makes for a nice barrier.
What do you mean by "give up his rights"? Whether he can allow other people to use the patented idea?
> and invalidate the patent?
What do you mean by "invalidate"? Whether they can show the patent office prior art to their own patent and ask them to for that reason invalidate the patent?
> Or can a copyright holder put his work into the public domain?
That depends on the jurisdiction.
My understanding is that someone needs to hold a patent, be it a non-profit or commercial entity. What I wonder is if someone who could file a patent or already holds a patent for an idea could put this idea into the public domain without anybody holding a patent? Does this make sense?
Well, yeah? I mean, the whole definition of a patent is that it gives someone a monopoly over an idea. A patent without someone holding it would be as sensible a concept as a bank account without a holder. A bank account documents that money is owed to someone. If there is noone who it's owed to, then there isn't a bank account. And if there is noone who holds the monopoly ... then there is no monopoly?
> What I wonder is if someone who could file a patent or already holds a patent for an idea could put this idea into the public domain without anybody holding a patent?
Depends on what you mean by "public domain". If you own the patent, you get to decide who gets to use the patented idea. Obviously, nothing prevents you from deciding that everyone is allowed to use the idea without any further conditions. That you get to decide that is the whole point of the system.
https://news.ycombinator.com/item?id=14583691
Link provides steps for people to submit prior art to USPTO for this patent application (patent is not granted yet)
Google has thrown few grants for ANS research to Jarek and folks of encode.ru community, then patented their research behind their backs
Update: they did not, but they did approach Jarek's university with specific interest in that research work, and vaguely expressed interest in funding them.
Jarek's stance - "this is my work, not yours, this ends this discussion."
Google - "we have grounds to do so, and nothing you do will move us"
Update: I put things in a more neutral tone.
This seems to contradict directly with what user SquareWheel writes in another comment and who now seems to have deleted that comment.
Google promising free license to anyone who implements AV1 would have been ok, if they patented something that is only useful for AV1. But they patented "using ANS for video" so now you can't use ANS in any competing video codec. In my opinion, people absolutely should fight to invalidate this patent using any work that mentions using ANS for video from a year (or more) before the filing.
Proper algorithm patents are supported in Europe because they are equivalent to a novel electronic circuit, which is generally viewed as patentable subject matter.
His original publication from - https://arxiv.org/abs/1311.2540 . That work was a result of two years of research that was well documented, and was making noise in encoding community. His first notes on use ans for quick entropy coding date to 2008.
First production grade code was published under GPL at around late 2013 with explicit intention to enforce copyleftness of derived works, and the patent poison pill clause.
Jarek claims that Google has abruptly ended communications with him few month after the time of first code publication
Update: it seems Google never made any grants to the group, but were only vaguely feeding them with "may bees" for two years, while requiring more and more detailed "research scope declarations." It is better now to let Jarek give his own first hand account of what was going on in between his university and Google's research people
I don't like the tone you've taken. I'll assume you're merely uninformed.
Stallman's organization, the FSF, was created to protect the open-source nature of things that he was a primary creator of or significant contributor to such as gcc and emacs.
The Python foundation was created by the creator of Python, and so on.
To claim that there is no creative endeavor in these entities is to ignore the fundamental reason they exist.