Slashdot Mirror
Inventor Says Google Is Patenting His Public Domain Work (arstechnica.com)
Rob Riggs writes: Jarek Duda, the inventor of a compression technique called asymmetric numeral systems (ANS), dedicated the invention to the public domain. Since 2014, Facebook, Apple, and Google have all created software based on his breakthrough. Google is now trying to patent a video encoding scheme using the compression technique. The inventor is fighting Google in the European courts and has won a preliminary ruling. The fight's not over and Google is also seeking a patent with the USPTO. A Google spokesperson says Duda came up with a theoretical concept that isn't directly patentable, "while Google's lawyers are seeking to patent a specific application of that theory that reflects additional work by Google's engineers," reports Ars Technica. "But Duda says he suggested the exact technique Google is trying to patent in a 2014 email exchange with Google engineers."
All inventions are built on the complex work of others. But you cannot preclude someone from inventing new things using your public domain technology.
Did you RTFS? He is claiming that Google's extensions are also documented prior art.
If he invented the machine screw, Google is claiming a patent for a machine screw used to hold together a bookcase.
There is no transformative act, simply a straightforward application in an expected field.
Patent examiners only look at prior patents for prior art. They don't have the time to look else where.
If you share an invention that isn't patented, someone will find it and apply for a patent. It it really is novel and nothing like it has been patented before it will be granted. They will then use it to sue everyone you shared the information with. Starting with the little guys who can't afford a decent lawyer.
It's extremely time consuming and expensive to invalidate a patent.
Which is no longer novel or non obvious. Google should not be granted a patent. Even on the derivative use.
The problem is: How do you know some other company won't patent it, and the USPTO or courts won't allow it? If so, they can sue Google for using it.
If I were Google, I would file the patent. If it gets rejected, then there is a paper trail showing that the invention is not patentable.
Prior art can only invalidate a patent if said prior art was itself patented.
That’s patently (pun intended) false. Prior art, patented or not, can be used to invalidate a patent. You can’t patent an existing invention, regardless of if you’re the first to file. If nothing else, that should be patently obvious (pun oh-so-intended) on account of the filing’s failure to pass the “non-obvious” test. If someone else has already invented it, the idea is obvious at that point, particularly so if the inventor verifiably disclosed it to you prior to your filing.
Ok... I just read up on ANS. It was quite an enlightening read and it's humorous because I remember thinking of a similar pattern when working with arithmetic coding 10 years back.
Here's the thing. The simplicity of ANS is elegant in its nature.
It's also extremely obvious in hindsight, though I'll be absolutely shocked if anyone in the courtroom except for Duda will understand that.
I would however say that because ANS is effectively so amazingly simple in its nature, if Duda's argument is to keep Google from patenting their extensions as opposed to trying to file for a patent himself later, I believe it would be much easier to suggest that the base math as well as the extension are non patentable under the pretense that they are not as much algorithms as opposed to mathematical discoveries or formulas.
There are some catches to this.
The probability distribution S may justify the math as an algorithm rather than as a formula.
But I would say otherwise that no part of this compression should be able to be patented under the same grounds that a Taylor series couldn't be patented. And while I haven't become an expert on the math yet for ANS, it reminds me of a generalized infinite sum problem.
So... call me crazy, but I'd imagine that pretty much anyone with a masters degree in some form of math should be able to act as an expert witness to simply say that this is a discovery and not an invention and therefore is not patentable.
The only difference between first to invent and first to file is when two people try to patent the same thing. In first to invent, the patent office has to examine all the documentation behind it and figure out who rightfully invented it first. In first to file, it's simple - whoever gets their application to the patent office first wins.
The "1 year disclosure" is a separate issue - in the US, you could disclose your idea to the public up to a year before filing your patent. That's it. Obviously, under first to file, this is no longer workable (since someone could see your idea and patent it first).
Be aware the US was the only country in the world with first to invent and 1 year disclosure. Every other country in the world was first to file, and no disclosure.
The implications are bigger - first, no disclosure means the first time someone speaks out in public about the idea, the idea is no longer patentable. By spilling the beans prior to filing the patent, you've invalidated your right to the patent (after all, what's to keep someone else from filing a patent somewhere else and stealing it from you?).
First to file makes no attempt to figure out who invented something first - so the little guy no longer has to prove he got this idea while hanging a clock, slipping and bonking his head on the toilet. BTW, it was rumored that Bell got the patent for the telephone by beating whomever else it was by about 15 minutes to the patent office.
This way, disclosure trumps patenting worldwide. The fact this guy's algorithm is public means you cannot patent it at all. What can be patented is a novel modification to it, though I'm not sure what Google did to it to make it work with video.
Let's see if we can find a pattern here:
1. Creating a general algorithm that can be applied to many different problems - No protection since math is neither patentable nor copyrightable.
2. Apply the algorithm to a specific problem - Patentable. 25 years of protection.
3. Writing a shitty almost off-topic post on slashdot in a thread about the algorithm - Copyright. Life + 70 years!
So, the less important the creative work is to society is, the more protection it gets.