Patent: Artificial Intelligence System | Hacker News Reader

Patent: Artificial Intelligence System

patft.uspto.gov

46 pointslumberjack48 comments

Software patents are utter garbage. We all know that. Most software patent attorneys know that. The system has evolved to keep lawyers employed. It is a way of taxing engineers to support lawyers. The plaintiffs need lawyers. The defendants need lawyers. There must be judges. The entire system gets weighted down with this huge load of unproductive individuals. Think of how many engineers become patent attorneys because law is so lucrative. Entry level patent attorneys in firms make $160K a year. Think of how little an engineer makes even four or five years out of college. Its more lucrative to patent and sue, than to build. These are the little things that are going to erode America's competitiveness.

I agree, I even know a friend who was studying computer science but moved over to law, his reasoning was how manipulating the law gives you a advantage over other people. The same is in Australia, our society values lawyers and they ruin almost everything.

Make no mistake, many lawyers would see the world burn just to make a bigger buck. Many of them truly don't care.

That's a courageous and bold statement to make. Apart from the weasel words that entirely undermine it.

If you want to say that all lawyers are evil, without any evidence, at least have the conviction to say that.

If you want to imply that some undefined minority, or possible majority, of lawyers may, or may not, potentially be paid arsonists, but might also not be arsonists, paid or otherwise, then... I dunno, maybe say something worthwhile instead.

"It is difficult to get a man to understand something, when his salary depends upon his not understanding it." - Upton Sinclair

One shouldn't draw the conclusion that lawyers are evil. Instead, it's a generalized principle: humans find ways to rationalize their participation in any system which awards them wealth and status, whether it's a little or a lot. Some patent attorneys know that patents are bullshit and profit from them anyway (whether on behalf of the trolls or the little guy), many internalize the belief that they are defending genuinely deserved intellectual property rights. There is a tendency for people to look at whether a system does well by themselves and their loved ones, and then work backwards to decide what is just.

(See also: the Banality of Evil.) http://en.wikipedia.org/wiki/Banality_of_evil

My experience is that law is like programming. The people who are making the "bigger buck" genuinely love their profession and take it seriously. The ones who only turned up for the bigger bucks quit because it bores and frustrates them.

Not to mention, lawyers are held to the standard of fiduciary duty. That's fairly heavy duty.

The "inventor" appears to be head of a small defense sub-contractor. http://www.crunchbase.com/company/cognitivecode

Note that the patent has not been issued. I certainly hope that this gibberish isn't taken to be sufficient details for someone to implement.

This patent was issued on 2/28/2012. It's the related application that hasn't been issued.

>Note that the patent has not been issued.

Thanks for pointing that out. I read the whole thing (surprisingly an easy read as far as patents go) and somehow missed the most important part.

Does this mean that it is still in process or was it a failed application?

Probably not yet reviewed

    "Software patents are utter garbage"

    " I certainly hope that this gibberish isn't taken to be sufficient details for someone to implement."

    "isn't this patent a little bit too reaching and isn't the logic a little bit too obvious to be worthy of patent?"

It's very clear from basically every single comment on this story that most of you have no clue about the patent system nor what to actually look for when sizing up a patent. I've picked just a few comments that demonstrate the general ignorance that can be seen here (this is not an insult, but people need to understand when they do not understand something and aren't qualified to comment).

I'd love for someone to reply with a citation that shows that just claim 1 is not novel, or is not inventive. To meet this criteria, the citation must either disclose ALL THE FEATURES LISTED IN THE CLAIM or it must be OBVIOUS to produce SOMETHING EXACTLY AS CLAIMED based on one or more PRIOR ART documents. It is not sufficient that all the features be in a single citation, but it must be obvious to actually come to the same invention AS CLAIMED.

So if someone could please find for me one or more citations which combined (and under Australian law it must be obvious to do so) which contains:

A method for a computer system to interpret an input from a user and generate a response, comprising:

1) receiving a user input;

2) converting the user input into an input array comprising rows and columns having a plurality of concepts;

3) determining if any of the plurality of concepts in the input array is derived from a root concept;

4) if any of the plurality of concepts is derived from a root concept, replacing each such derived concept with the corresponding root concept, identifying one or more related concepts that relate to the root concept, and generating a multi-dimensional array based on the input array that includes the one or more related concepts;

5) generating one or more additional multi-dimensional arrays, based on the input, containing any composite concepts, each derived from two or more concepts contained in the original array;

6) marking one or more concepts in the multi-dimensional arrays as essential based on application-specific criteria;

7) correlating a plurality of concepts in the multi-dimensional array to a plurality of first elements in a database by comparing a plurality of linear arrays derived from the multidimensional array to the plurality of elements in the database, wherein the first elements in the database includes a link to a second element in the database do not comprise possible responses;

8) determining a plurality of possible responses to the user input based on the correlation of the multi-dimensional array and the plurality of elements in the database

9) and generating a response to the user input.

There's 9 broad concepts here, and to my knowledge, at least 3, 4, 5 and 7 seem non-obvious to me, but my AI background is weak. Even if they are known, is it known to combine ALL those features in the way claimed? I'd be surprised if it were, the USPTO is actually quite good at searching in general.

It's so depressing to see a forum of people who are supposed to be intelligent and rational jumping to their guns at the mere mention of the word patent. Of course there are some junk patents, but those few should not be used to devalue the work done by inventors big, small or individuals around the world. If you don't even know how to read a patent, why do you feel like you're qualified to comment on their validity? We don't discredit academic papers because we've seen some of the words in their titles before, so why should patents be any different? It's dishartening to see such ignorance almost encouraged on Hacker News.

"The idea that I can be presented with a problem, set out to logically solve it with the tools at hand, and wind up with a program that could not be legally used because someone else followed the same logical steps some years ago and filed for a patent on it is horrifying."

- John Carmack

But if that were the case, the patent would almost certainly be invalid; the claimed invention must be inventive, i.e. having come up with something that is not obvious. What is described is coming up with the obvious solution to a problem, not coming up with an inventive solution to the problem. It's also annoying when people use soundbites from famous people as proof; they sound good and make sense on the surface but usually don't hold up to scrutiny (and I say this as someone who has a lot of time for John Carmack's advice and insights).

An example might be the invention of bloom filters[1]. I would argue that if you were looking for a way to check for membership in a set in the time before the bloom filter's invention, it would not have been obvious to devise a structure which can give false positives but no false negatives through the use of multiple hashing functions and a bit array. It is of course obvious to us now, knowing the answer, but at the time it would not have been. this is what the patent system is about; encouraging people to invest time, effort and money into inventive solutions to problems by giving them a way to protect their investment for a limited period of time. Somehow this all becomes evil when a large company is doing it, but you'll all be outraged when a startup comes up with something inventive and have their ideas "stolen" by big companies when they didn't seek to protect themselves.

[1] I'm not saying this is a shining example of a patented technology; I have no idea whether it has ever been patented or not. However it is a good example of an inventive solution and one which most people would not regard as obvious and be able to come up with through routine trial and error.

Carmack was talking from experience: http://en.wikipedia.org/wiki/Shadow_volume#Depth_fail

> But if that were the case, the patent would almost certainly be invalid; the claimed invention must be inventive, i.e. having come up with something that is not obvious.

You just proved that you don't know how patents work. I'm sorry, but that's the naked truth.

Clean-room implementations are no protection from patents.

The burden of proof is on you. I'd like for you to demonstrate that claim 1 is indeed novel and so forth and not been presented in every sci-fie movie on Earth. And while you're at it, please provide some git repo for us mortals to enlighten us on what we've been all missing since the rise of AI.

If concept 3 is non-obvious to you then you have no right on claiming ignorance on patent law from our part.

The burden of proof is surely not on me (and what you ask is actually impossible); I'm refuting those who say that the invention defined by the claims is clearly obvious. This is something which is not impossible, all that is needed is a citation or two which discloses all the features of the invention. If they have proof, then they can show it, and show the USPTO while they're at it. It is not possible to prove that something is novel or inventive, you can only show to a high degree of certainty. It is however comparatively trivial to show that it is not novel or inventive.

Also, the idea that AI covers human level intelligence goes against the whole field of AI over the last, what, 5, 6 decades? Computers that can play chess are considered to use AI, but they certainly can't take over the world.

The problem is at it's root. Obviously it is not an invention or we would have heard of Skynet at news. Since this is clearly not the case, any claim from the patent owner is simply an arrogance. If the patent has been issued this just proves ignorance on USPTO part.

The ultimate point is that if all you have to do to receive a patent is just an idea, then the patent system is broken. The patent should protect the implementation, not the idea.

Do you even know what the study of AI is about? I even laid it out for you and yet you still bring up nonsense like references to Skynet? Just because they use a term artificial intelligence does not mean they're talking about the Hollywood definition of what AI means, it means they're talking about a system which behaves in a way we would consider more intelligent than what we're used to currently (and most successful AI projects just become regarded as the normal way things are done). Google's search suggestions and related searches would be regarded as examples of AI. See http://www.ucs.louisiana.edu/~isb9112/dept/phil341/wisai/Wha... for a good explanation.

You seem to be getting stuck on the title of the patent application, and this has absolutely no effect whatsoever on the scope of the claimed invention and is, as far as patent validity is concerned completely irrelevant and must only be generally related to the disclosure.

"The patent should protect the implementation, not the idea"... That is exactly what this application seeks to do! It's seeking to protect their implementation of performing a search or giving suggestions (or something similar, on an iPhone and don't want to have to read it again on here). But you can't do that by providing source code, it would make the patent worthless because someone sticking some extra irrelevant step in the middle could claim it doesn't do the same thing. If you invented a new toaster that used one tenth the energy and took a tenth as long, would you want to be able to stop people using the same mechanism that allowed you to do that? Or would you want to be able to only stop people from making a toaster that is exactly the same as the one you've created, and if someone adds a cooling rack on top, you couldn't stop them from taking your idea which you've invested time and effort in. This is exactly the same for software patents. You patent inventive solutions to problems, in a way that means you can protect it in all it's forms, just just the one way you've written it in the one language you've chosen to do so.

Of course people should be able to apply for patents for mere ideas before actual creation; this is how you make money from being an inventor for anything more than simple back shed ideas: you come up with a novel and inventive way to do something, you talk to a patent attorney and apply for a patent, and then you go to investors and say "hey, I have this awesome idea that can make you money. Give me money to implement it, and my patent will insure that only you are allowed to use that idea or to licence it to those of your choosing". The patent system is all about protecting ideas, but people chuck a spaz when it's not something as tangible as a new gearbox or a new chemical or a new piece of mining equipment or whatever it is. It is my firm belief that people who come up with novel and inventive ideas through hard work and often a lot of their own money should be able to protect others from stealing their ideas. Without this, there is almost no incentive for people to form start ups with great new ideas they've had; large companies are free to just take their ideas and there's nothing that can be done about it.

The toaster example is relevant. It is a physical thing, like a particular git repo would be, too. In contrast, an idea is not relevant. It just happens that in US, ideas are patentable - which is nonsense. A nonsense that you happen to agree with. Yet you contradict yourself - on one paragraph you want to protect the implementation, and in another you wish you could protect the idea.

I'm not stuck on the title, and I know just as little about AI as only there are strong and weak forms of it. But if I happen to look at Star Trek tonight I should not be allowed to go and patent 30+ "ideas" that I saw in that movie or that I dreamt of.

I won't read any of the claims because of triple damages. I didn't even read past the first paragraph of your comment because you appear to be listing them. Maybe some of the problem is the thoughtcrime involved forces us to react against the title alone.

Muhaha, patent all you like, it won't protect you from the intelligence explosion (you know, that AI that turns itself into Skynet before the Patent Office can strike it down).

On a more serious note, I'm glad I'm living in a country where there's no such thing as infringing software patents.

I worked out one reason why software patents are so terribly troublesome versus patents for physical mechanisms.

Because we deal in abstractions.

You can always take this piece of code you wrote and notice that it can be abstracted to solve a class of problems. And then that this class of problems is in turn a subclass of a larger set of problems. And then it's increasingly vague all the way down.

Before long your patent application transforms from "Method to improve page retrieval on ARM7 CPUs" to "Method to improve caching in memory" to "Method to improve caching" to "Method to accelerate any computation".

That generally doesn't happen in the physical world, because physical inventions aren't easily generalisable. You can't take a tractor and, with a few lines of code, abstract it into a platonic ideal that can also spit out cars, buses, trains, planes, ships and the starship Enterprise.

It is not yet a patent: it has been filed but there is no patent number and no patent issue dates. You can find tons of such patent pending for reviews.

Its patent number is 8,126,832 and its issue date is Feb 28 2012.

Related application 60893237 has no patent number or issue date, is that what you are looking at?

Thanks yoda_sl, most everyone seemed to miss this obvious point

I feel really insecure posting this but I have to ask: isn't this patent a little bit too reaching and isn't the logic a little bit too obvious to be worthy of patent? Or are my feelings of insecurity justified and I am truly too ignorant to recognize the intricacies and complexities laid down in that sea of text?

It's not news that there are tons of broad patents of ridiculous nature which obviously have prior art.

But this isn't a broad patent, it's in fact very specific in its claims, and no one here has managed to provide anything that resembles prior art for this patent. You should stop making ridiculous claims without any proof or facts to back them up.

You are being ridiculous. I worked with a similar system during my CS studies. I.e. artificial intelligence analysis of natural language input. Patenting such old known ideas is pure lunacy. Anyway, I view patenting such ideas as ridiculous even if they weren't known before. Such abstractions simply should be unpatentable.

Got a citation that has all the features of even the first claim? That's all that's needed. The patent doesn't claim a system which does analysis of natural language input, it claims a specific system for doing such. Again, find a citation and prove it isn't novel or inventive. Until then, why should anyone consider it invalid? The law must work on proof, not people saying "Oh yeah, I've totally done that before, trust me". If it's so obvious, then it should be easy to find a citation.

Also, why should inventive ideas in software be any less valuable and worthy of protection as inventive ideas for physical objects? This patent protects a very narrow scope, anything that infringed it would have to perform all the steps in the first claim and it seems to me that it wouldn't be too hard to design a similar system which doesn't infringe. It might not be dissimilar enough to get you a patent, but it also wouldn't infringe.

Also, why should inventive ideas in software be any less valuable and worthy of protection as inventive ideas for physical objects?

Why don't you go further with your logic. Why inventive math theories can't be patentable? Or may be abstract information theory itself? You can make it absurd by pushing the border in that direction. I draw the line at the point which excludes software and algorithms from patentability.

Besides the theoretical aspect of what should be patentable, there is a very pragmatic one, which was already brought many times, and all software patent proponents pretend that it doesn't exist - patent thickets. In case of hardware they aren't a major risk, but in case of software they are completely unavoidable and tend to be horrifically tense. It's a good enough of a reason to render software unpatentable, since in the case of software the patents don't serve their purpose, instead of promoting innovation, they stifle it. Whether you like it or not, that's how it works. Europe had enough common sense to forbid software patentability, but US didn't.

The line is drawn where the invention stops being a part of the useful arts, and an algorithm on its own is not useful, but a system using a new algorithm is indeed useful, and that's where the distinction between algorithms and software lies.

Patent thickets are indeed an issue, but they've been an issue several times in the past and the world didn't stop turning. See http://www.ft.com/cms/s/2/69a96688-b776-11e2-841e-00144feabd... for a brief discussion. I completely agree the patent system is not perfect, but I do believe it's important to allow the small guy to come up with new inventions, be it software or not, and be able to protect themselves from having the big guy steal their ideas with nothing in return.

The patent system is a man made system, and all man made systems will be exploited by man if it is in their best interest, and that's exactly what big corporations are doing. But without the system, the little guy stands absolutely no chance against the big guy and with it they do. This is how innovation is encouraged by the patent system. You bet your arse that when a small company funded by those here gets their ideas stolen that people here will be up in arms. But if they have parents, they have a real method of recourse, whereas without, all they can do is call them out on it, but have no guarantee that they'll get anything at all. Without the patent system, we'd get rid of patent trolls, but we'd being in the much worse "Idea theifs", who steal the work done with other's time and investment to produce products for their own profit.

Welcome to the nonsense of the U.S. patent world.

You're not alone in thinking that having that concept patented is actually kind of... outrageous.

No, I think this looks vacuous as well.

I also wonder why claims 14 and 15 appear to repeat claims 12 and 13 verbatim, except for one comma. Does that change the meaning somehow?

Actually by just scanning through the text I only find a lot of AI rehashing and nothing new.

The sentence:

The method of claim 1 wherein the user is a computer.

describes polymorphism in OOP.

Way to go!

So how can we / could we have stopped this patent. Is there an easy online way to submit prior art for patents under consideration? That way all the reviewer has to do is accept the prior art and nuke the application

Do you have any prior art? This is not a patent for "Artificial intelligence" in general, it's a patent for a very specific method of doing something that appears intelligent, and this is actually what artificial intelligence is all about. See my comment above to see what is required for something to be relevant prior art. Basically, you're not going to find it. And who cares anyway, the scope of the claims is really quite narrow, it's going to take a lot of effort for someone to actually come up with a system which infringes this patent without having read the thing in the first place.

> my AI background is weak

> it's going to take a lot of effort for someone

First, you assert that you are not an expert in AI. Then you assert how much effort would be required to independently come up with the same implementation.

You have to chose which of those two assertions to make.

My AI background is weak but not non-existant. But this is beside the point. I'm not the one making claims this is a junk patent and that it's clearly obvious. The ones who are have provided no proof that it is obvious. For a patent to be invalid, there must be proof, you can't just say "oh clearly it's obvious" because you can say that about anything at all once you know the solution to the problem.

Regarding my comment on how much effort it would take, I was stating my opinion that it, at face value, doesn't seem obvious to me. But the assertion that it is obvious and a terrible patent is something that can be proven with a single link, but no one has provided one. And I believe the reason for this is because the naysayers have no idea what they're talking about and are too ignorant to know that just because the term artificial intelligence is used does not mean that they are claim all artificial intelligence systems conceivable to man, but merely the system they have invented.

By all means, prove me wrong, find a citation that shows this is either not new or is obvious, but until you do, the patent stands as a valid patent. Any citation you find I'd be more than happy to look through and see whether it is a suitable citation, but until then, the naysayers haven't a leg to stand on.

I'll make an array of the parts of your post that I disagree with. And optionally an array of the parts of your post that I agree with. I'll mark any root concepts, and then highlight where they appear in the rest of your comment. Then I'll make an array of possible responses, and check for frequent use of the root concepts.

Wrapping something in patent-mumbo-jumbo, doesn't prevent it from being bullshit. No one will find a link to prove that making arrays and root concepts and blah blah blah has prior art, because it's all crap. Top to bottom.

That's my opinion.

So you've addressed maybe half the essential features of the claim, and that clearly makes all of it obvious? What about the rest? Is it obvious to do all these things in the way claimed? Do you have proof this it is obvious to do so? The law doesn't work based on opinion, it works on proof first, and evidence when proof is lacking. You can call it bullshit all you want, but until you have something to back it up, you're just talking shit.

> The law doesn't work based on opinion, it works on proof first, and evidence when proof is lacking.

Yeah, and I have enough of a background that my opinion is evidence.

That was simple, wasn't it?

If we found an acknowledged expert in the domain, their opinion would be very strong evidence indeed.

Let me put it this way: the goal of establishing a patent system was to encourage public disclosure of methods. A time window was granted to the patent holder to make exclusive use of the material, but after that, to revert to the public domain.

This patent application ADDS NOTHING THAT IS NOT OBVIOUS to the public domain. That's my opinion, and I think I have enough of a background in the domain that my opinion is evidence.

Your argument is that, if I were MORE of an expert in the domain, I might suddenly realize that the claims are more subtle and inventive, than I can discern as a journeyman in the domain?

And yes, you keep cleverly asking for a link, but that's presuming that prior art is the only way to defeat a patent. An idea can be both NEW (or phrased to sound new) and OBVIOUS to a practitioner. In which case, your asking for a link is a strawman argument.

Forbes article with some background, http://www.forbes.com/sites/karstenstrauss/2012/07/09/riding...

The incredibly banal Kato Kaelin interviews Leslie Spring and Mimi Chen (at about 7:30 minutes) http://www.tubeguru.com/video/9311 Are these two musicians, vocalists, or founders of cutting edge AI Cognitive Code Corporation? Or ALL of the ABOVE??

I'm thinking that once you've got a patent on strong AI, you more or less have a patent on everything. It's a method and process for doing, well, everything a human ever could or would do.

There are hundreds of patents on "artificial intelligence". All of them are worthless gibberish.

Labeling something "artificial intelligence" does not make it so, and the patent office has little interest in what you label the invention as long as you describe the process in a manner that meets their criteria for patentability.

Trade Secret.

If you've a worthy AI engine/algorithmic break-through, software has the hugely greater advantage over physical inventions by securing it behind the API front-end. Nothing gained by filing a patent and everything to lose.

Reply on news.ycombinator.com