Algorithmic Patenting

An anonymous reader writes: Venturebeat reports on companies using software to "create" patents. They say a company called Cloem will use the software to "linguistically manipulate a seed set of a client's patent claims by, for example, substituting in synonyms or reordering steps in a process, thereby generating tens of thousands of potentially patentable inventions." The article says, "There is reason to believe that at least some of its computer-conceived inventions could be patentable and, indeed, patents have already been granted on inventions designed wholly or in part by software."

Comments

A good strategy

[radl33t]

This is a strategy for demonstrating the absurdity of the current patent regime, right?

Re:A good strategy

[Kevin Fishburne]

This is a strategy for demonstrating the absurdity of the current patent regime, right?

Unintentionally, no doubt. On the bright side the more absurdly and widely abused the system is the more ammunition for reforming it. I hope they cause a real mess spamming the USPTO with every possible patentable combination of words. Maybe they'll replace the USPTO staff with an infinite number of monkeys with typewriters to process them.

Re:A good strategy

[Sarten-X]

In true Slashdot style, I haven't read TFA, but TFS sounds like Venturebeat is stirring up a good old-fashioned angry mob.

Historically, patent lawsuits have been won or lost based on careful wording. A good synonym can mean an enormous financial difference for an inventor (or inventor-funding company).

Patents must be specific enough to describe a particular set of implementations of an idea, rather than just the general idea itself. Despite Slashdot's love of the phrase, "on a computer" does not a patent make. Rather, the patent must describe exactly how the computer functions with regard to the invention itself. Yes, sometimes that means describing the only reasonable mechanism, but it's still specific.

On the other hand, that specificity can be problematic when it comes time to actually use the exclusivity a patent provides. A car-analogy patent might have been worded to refer to driving on asphalt, but there are some roads that are paved with concrete or bricks. Hiring a specialist to find such trivial loopholes might be a very good investment for an applicant trying to write their patent. Any realistically-equivalent implementations can be added to the patent as additional claims.

Re:A good strategy

[fuzzyfuzzyfungus]

It looks like a very effective denial of service attack, to me. No law school is shitty enough to puke out patent examiners that cost less per hour than some glorified Markov chain, so barring some sort of rate limiting, with teeth, you should be able to shove a lot of utter shit through by sheer brute force.

Even better, if the patent examiners don't just crack under the strain, then they do the job of distinguishing vaguely worthwhile patents from algorithmic word salad for you!

Re:A good strategy

[Strangely Familiar]

The problem is in the number of possible combinations. Even if you take a small number of possible elements, and a smaller subset of chosen elements from that original set, you start generating serious numbers of patents. Say you were going to make a new bar code reader. You have 20 possible things you can put in that bar code reader. Rotating 8 sided mirror. Rotating 6 sided mirror. Oscillating mirror. Shape memory alloy actuators. DC motor drive. Stepper motor. Fixed mirror set. Red diode laser. Infrared diode laser. LED lighting. CMOS imager. Light sensitive diode. CCD imager. Fresnel lens. Cylindrical lens. Holographic lens. Etc. It would be very easy to find many more than twenty elements for "things to make a bar code reader". Just, for ha-has, stick with 20. Now, say your novel invention comprises a six element combination not found in other bar code readers. How many possible combinations? You ARE relying on a computer to brute force this, aren't you? Well, that's 20 x 19 x 18 x 17 x 16 x 15 possible combinations for six elements chosen from a possible field of 20. That's 27,907,200 combinations. (Please don't quibble over my mathematical simplifications). That would be more patent applications than have ever been filed in the U.S. All your patent applications would say so far is that you have six elements, not how they are configured or how they function. There are a ton of ways you could configure an 8 sided rotating mirror, especially if it is interacting with a fixed mirror set, or an oscillating mirror set.

.

So if you wanted to patent all the combinations, you quickly run out of money. The USPTO charges thousands of dollars to process each application. If you wanted to patent the algorithm, the USPTO has a special kind of rejection for you. It is called "undue experimentation". In other words, you would have to perform a ridiculous amount of experimenting with unworkable and wrong combinations before you found one that works. In yet other words, no one is teaching you anything new by saying, "try a bunch of different combinations; one of them will work". Duh. Edison knew that, but he did the hard work of actually trying all the combinations, and finding the very few that worked. So, if the computer program isn't actually intelligent, it will waste a whole lot of resources attempting to patent useless crap.

.

TFA says that one company is copyrighting all possible 400 word combinations in the English language. That is 400 elements chosen from a set of around maybe 100,000 - 250,000 elements (English words). Since the words are allowed to repeat, taking the 100,000 figure, we have 100,000 raised to the 400th power, or 10^2400 possible combinations. Say the company had a very fast computer and was able to express and thereby claim 100 billion 400 word combinations per second. It would need just 10^2389 seconds to claim each combination. In very round numbers, a year is about 10^8 seconds, so the comany would need about 10^2381 years to complete it's task. In very round numbers, the lifetime of the universe is about 10^10 years, so the company would need about 10^2371 lifetimes of the universe to complete it's task. Or a faster computer. Or a new law that says that a person doesn't actually have to express something in order to copyright it.

Simple Fix - Eliminate Patent System

[pubwvj]

It is time to eliminate the patent system. The only reward should be delivering the product to customers and making the sales.

Just in case

[Opportunist]

anyone needed another proof that the patent system is FUBAR.

Ignoring the economics

[dtmos]

Obtaining patents isn't free. One would have to look at the zillions of generated patent applications and decide which ones, if any, were worth the application and prosecution fees -- not to mention the attorney's fees. (It wouldn't be a very high percentage.) This is a pointless exercise.

My advice is to roll over and go back to sleep.

Re:Riiight

[ganjadude]

"it is said by some" or "critics say" is a tool used by all journalists. you cant say that is only a fox news thing. I know fox new is a fun and easy target, but seriously. this is something that happens on every network

Re:How about making patent reviews like PhDs?

[meerling]

Alright sir, I see you are here to defend patent XJ82934952H28354. Why isn't the inventor here?
> It is judge, I have it running on my smartphone.
Really? Let's see, name Random Global Search And Replace Thesaurus Based Script Bot For Patents. That's a rather long name for a person, as well as a rather odd one. What were your parents named Mr Patents?
> I'm sorry judge, but it doesn't have patents, it's a piece of software.
Software? So you mean it's one of those AI thingies? A sentient machine like C3PO? Able think, imagine, and create new ideas?
>Ummm... Not as such. It just changes words with other synonyms and sometimes reorders the steps in a sequence of steps.
So let me get this straight. You patented someone elses work, after making minor modifications to it with something that has no more creativity or understanding than throwing random words into a jumble and expect to invent this despite patent laws requirement for it to be something that an expert in the field would not find obvious and yet "running an app" is something that anyone can do, even my two year old daughter that can't read, and absolutely in no way reaches that simple, though often argued over, benchmark?
>I, uh, wouldn't put it that way your honor...
Well I did. Patent revoked. In fact, all "patents" submitted that are creations of that stupid script are hereby revoked, and don't try it again or I'm going to throw the book at you for wasting everyones time you stupid little troll!
.
(Ok, I'd expect everything after the word "revoked" would just be in the judges head, but the thought counts, right?) :)

Bad article title

[DickBreath]

Shouldn't it be called: Algorithmic Patent Trolling

Re:A good tragedy

[DickBreath]

This is a strategy for demonstrating the absurdity of the current patent regime, right?

I think you may have meant this is a tragedy for demonstrating the absurdity of the current patent regime. Just a typo, I'm sure.

But can you patent the software that gens patents?

[JoeyRox]

And so on? Such recursion might cause the patent office to explode.

Re:How about making patent reviews like PhDs?

[Jeremi]

I think it will be more like:

Alright sir, I see you are here to defend patent XJ82934952H28354. Why isn't the inventor here?
> I'm right here, your Honor!
Someone said you used a computer program to write this patent. Is that true?
> It sure is, your Honor! But then again, most everybody uses a computer program to write patents these days. Microsoft Word, for example.
Ah, I see. Carry on!

Re:How about making patent reviews like PhDs?

[radarskiy]

Clearly, people who are bad at extemporaneous speaking don't deserve patents.