"Google's mission is to organize the world's information and make it universally accessible and useful"
(didn't dig too far into this but..) why y'all need to patent this then?
Patents can be beneficial to facilitate constructive competition, but think humanity is best served by neural nets becoming the new electricity rather than the new Apple-esque walled-garden...
I don't know what Google's reason is for this patent, but defensive patents are really common these days, and I think just about any big company has lawyers saying "patent everything you can, since you need a big patent portfolio for defensive purposes". Point is, just because Google is filing this patent doesn't mean they intend to stop others from using this approach. A link to a patent application isn't enough context to know.
I don't know if this is the reason, but I think Google started getting a lot more paranoid about hoarding patents when Apple and Microsoft started going after Android OEMs in the early years. Google didn't really have any "counter-offensive" patent strategy then, which is why it went on a patent buying spree back then, although most of the available ones also got bought by Apple and Microsoft through Rockstar and so on.
So best case scenario, Google doesn't want to be caught with its pants down regarding patents. Worst case, it wants to "own" deep learning, so that nobody can really compete with them. Although I think that would be a little in conflict with their strategy to open source tensorflow.
To really figure out on which side Google is now playing we'll have to see how they respond to future patent reforms, and whether they join Microsoft and IBM to once again kill those reforms, or support the reforms to abolish software patents or drastically reduce their damage.
This is a patent on a very particular form of translation model that handles rare words, i.e. this paper that all the authors are on: http://arxiv.org/abs/1410.8206
What you wrote is literally the exact meaning I intended with the title, but now reading some comments I see many people are interpreting it as "Google atttempts to patent deep neural networkS for machine translation", (notice the plural) which has a vastly different meaning. I guess I assumed that this audience would know that it wouldn't be remotely possible to patent a broad technique that's already so widely published.
I think that even though the application is for a specific architecture, that this is still worth knowing about, since LSTM is a such well known technique for dealing with sequential data like sentences, and since so far virtually all progress in AI and ML has been driven by academia and has remained open.
I also think it's important to note that pretty much any interesting machine translation task will contain rare words, so even though they've framed it in terms of a specific task, it's one that's actually extremely broad. Machine translation is not so hard when you have tons of data and when you've seen every word many times in combination with its translations. The only really interesting case is when we have to use background knowledge and context to infer meaning. Since natural languages are notoriously ambiguous, this happens all the time. So this app may be broader than it first appears.
Google is patenting this so that if they get sued, they already have a patent. If they actually sue anyone over this FIRST, I'd be shocked. This has got to be a defensive patent, in case someone else was about to patent the same thing and run after them with a lawyer.
Why? LSTM as such isn't patented, so the concept is free to use by anyone without any restrictions or royalties, no matter who invented it.
Also, the concept of LSTM wouldn't be patentable directly even in the current liberal software patent interpretation, you might patent particular applications of LSTM (e.g. this patent) but not any and all arbitrary applications of LSTM.
You wait until sometime violates the patent, then threaten to sue them. If they don't settle or license, then you actually sue them. If the court decides in your favor you can get damages and they have to stop making the infringing thing. Same as any other patent really!
Why wouldn't this be enforceable? It's just as enforceable as any other software patents - we can (and should!) argue about legality of software patents as such, but if the patent is granted, then you simply take a look at any competing MT system, and if they seem to use this approach, you'll be able to sue them for patent violation.
Why is Google patenting this tech? It's cutting-edge machine translation. The rest of the industry is still talking about statistical machine translation. The sector is moving very fast. With the recent SyntaxNet release and neural machine translation, Google is approaching the dream: universal automatic translation, which, by the way, is an innovation goal of Obama's administration.
[+] [-] osterbit2|9 years ago|reply
(didn't dig too far into this but..) why y'all need to patent this then?
Patents can be beneficial to facilitate constructive competition, but think humanity is best served by neural nets becoming the new electricity rather than the new Apple-esque walled-garden...
[+] [-] alangpierce|9 years ago|reply
[+] [-] mtgx|9 years ago|reply
So best case scenario, Google doesn't want to be caught with its pants down regarding patents. Worst case, it wants to "own" deep learning, so that nobody can really compete with them. Although I think that would be a little in conflict with their strategy to open source tensorflow.
To really figure out on which side Google is now playing we'll have to see how they respond to future patent reforms, and whether they join Microsoft and IBM to once again kill those reforms, or support the reforms to abolish software patents or drastically reduce their damage.
[+] [-] jagtodeath|9 years ago|reply
[+] [-] wnevets|9 years ago|reply
so someone else doesn't and trolls with it?
[+] [-] laxatives|9 years ago|reply
[+] [-] jkrause314|9 years ago|reply
This is a patent on a very particular form of translation model that handles rare words, i.e. this paper that all the authors are on: http://arxiv.org/abs/1410.8206
[+] [-] shmageggy|9 years ago|reply
I think that even though the application is for a specific architecture, that this is still worth knowing about, since LSTM is a such well known technique for dealing with sequential data like sentences, and since so far virtually all progress in AI and ML has been driven by academia and has remained open.
I also think it's important to note that pretty much any interesting machine translation task will contain rare words, so even though they've framed it in terms of a specific task, it's one that's actually extremely broad. Machine translation is not so hard when you have tons of data and when you've seen every word many times in combination with its translations. The only really interesting case is when we have to use background knowledge and context to infer meaning. Since natural languages are notoriously ambiguous, this happens all the time. So this app may be broader than it first appears.
[+] [-] PeterisP|9 years ago|reply
[+] [-] VonGuard|9 years ago|reply
[+] [-] spacemanmatt|9 years ago|reply
[+] [-] transpy|9 years ago|reply
[+] [-] YeGoblynQueenne|9 years ago|reply
[+] [-] tomwrenn|9 years ago|reply
[+] [-] mistermaster|9 years ago|reply
[+] [-] PeterisP|9 years ago|reply
Also, the concept of LSTM wouldn't be patentable directly even in the current liberal software patent interpretation, you might patent particular applications of LSTM (e.g. this patent) but not any and all arbitrary applications of LSTM.
[+] [-] Cshelton|9 years ago|reply
[+] [-] sp332|9 years ago|reply
[+] [-] PeterisP|9 years ago|reply
[+] [-] therobot24|9 years ago|reply
[+] [-] transpy|9 years ago|reply
[+] [-] astrodust|9 years ago|reply
[+] [-] mooneater|9 years ago|reply
[+] [-] transpy|9 years ago|reply
[+] [-] readams|9 years ago|reply
[+] [-] mooneater|9 years ago|reply
[+] [-] qqqqqqqqqqqq|9 years ago|reply