Hacker News new | past | comments | ask | show | jobs | submit login
Algorithmically generated prior art (allpriorart.com)
127 points by dirkk0 on Apr 10, 2016 | hide | past | web | favorite | 35 comments

It looks like this was vapor-ware, no proof they actually generated what they claimed (I'm guessing if they did, they would still be around).

Yeah, a judge is not going to find "prior art" generated by an algorithm sufficient to invalidate a patent. As a political statement, this is fine (though hand-wavy), but silly as a practical solution.

Edit: That said, we do need to fix the patent system. The problem, right now, is human: there needs to be political will, effected by people, to persuade Congress to change the laws. An algorithm does not do that.

Although it's possible this thing is a joke, in which case, well played.

They may not find it to be valid prior art, but I think it'd be a relatively convincing argument to something being obvious to a person having ordinary skill in the art, as it proves there is no inventive skill required.

If we can digitally describe something as a sequence of bytes, and we stack those bytes end-to-end, can we not say that the bytes together form a (very large) integer, and that the integer already appears in the set of natural numbers?

Apparently the distinction isn't so clear.


See also: library of babel [1].

[1]: https://libraryofbabel.info/

What color are your bits? http://ansuz.sooke.bc.ca/entry/23

The problem is that the US system doesn't view software as a series of mathematical manipulations. Rather software is considered a thing like a tractor. The internals are therefore patentable.

The US system does prevent math from being claimed. Others have tried to explain the inherent math nature by making products with function languages of varying purity. These claims fell on deft ears.

A lot of us here are software engineers, right? Why should a mechanical engineer's work product be protectable, while a software engineer's work product not? Aren't both the result of creativity and labor?

>These claims fell on deft ears.

On deaf ears, you mean? (Or perhaps daft?)


A succinct explanation of why that's true and irrelevant: https://qntm.org/number

tl;dr: Abstract "existence", as numbers have, is not the same thing legally or practically as concrete "existence", as files have.

Neat. The landing page has some examples: http://allpriorart.com/

I don't think this is intended to actually create prior art that could be used in a patent dispute, but as an art project it's well-executed.

Looking at the examples, I notice that they are all rather long. I have often thought about the idea of algoritmically creating prior art, but simply by joining pairs of words. For example: Vibrating motor+keyboard, vibrating motor+screen, vibrating motor+mouse, vibrating motor+stylus...

To prevent patents like this one: https://en.wikipedia.org/wiki/Immersion_v._Sony

This is an older and more succinct implementation of that kind of idea: http://thesurrealist.co.uk/priorart.cgi

So something to exhaustively generate trivial ideas in order to prevent companies patenting things like "one+click+purchase"? I have a feeling it would get a response of "that doesn't count" but I don't know how they'd back that up.

Why couldn't this be used in court?

I guess because they don't really make sense.

"Provided is a system and program for automatically handling an error when retrieving a file for an application. The structure comprises a seat element which has a top surface, designed to at least partly contact a user, and a support element underlying the seat element and designed to be connected to the movable or stationary frame."

I thought that each sentence comprises the equivalent of a patent.

You're thinking about claims, here prior art is just public knowledge that doesn't have to be in any particular format

Cool concept as art, but is basically a patent version of infinite monkeys on typewriters hoping to write Shakespeare.

That said, with a bit of guideance such as learning/ feedback and/or being routed around an initial framework (perhaps TRIZ or similar https://en.wikipedia.org/wiki/TRIZ), there could be a much higher signal-to-noise ratio. I've seen other attempts at invention through AI or algorithm - as dmritard says, the key to useful invention isn't generation, its filtering the noise.

Although as others have said, I doubt this would be admissable in court. Randomly generated 'claims' remain random: one could say in an infinite, random universe, surely every invention has already been created simply by matter composing itself in the approprieate way.

> I've seen other attempts at invention through AI or algorithm - as dmritard says, the key to useful invention isn't generation, its filtering the noise.

That's exactly my thought on it after reading a couple.

Consider the case of chemistry: One could write an application that goes through known chemical compounds[1] and then generates a combination of all of them, and for each combination, adds a ".. for use as a food additive", ".. for treatment of <medical condition>" ".. for general purpose cleaner" ".. for use in a battery" etc.

Without doubt, within that crazy-huge list, there would be some absolutely correct claims, just as one of the infinite monkeys did in fact write Shakespeare: but which one?

This site really isn't any different.

[1] For example, harvest everything on https://en.wikipedia.org/wiki/Category:Chemistry-related_lis...

Prior art continues to be a basis for invalidating a patent: the prior disclosure of an invention to the public invalidates a later patent. All first to file does is change the rule for deciding who gets priority in the unusual case where there are simultaneous patent applications before the PTO. Before, there was an inquiry into who was the first to reduce the invention to practice. Now, priority goes to the first to file. But a patent still cannot be issued if the invention was previously disclosed to the public either by the inventor (outside the one-year grace period), or by someone else.

Thank you for clarifying that, I have had a wrong interpretation of it for some time.

Wrong. From your second link:

"Under the U.S. first to file system the inventor will still have a personal grace-period, which is not available to inventors outside the U.S. in many countries that follow a more traditional formulation of the first to file rule. This personal grace-period says that the inventor’s own disclosures, or the disclosures of others who have derived from the inventor, are not used as prior art as long as they occurred within 12 months of the filing date of a patent application relating to the invention. However, and this is a very big however, disclosures of third-parties who independently arrived at the invention information will be used against the inventor unless the disclosure is of the same subject matter. Said another way, there is virtually no chance that a grace-period will exist relative to third party, independently created disclosures."

Prior art indeed does invalidate the ability to patent something, unless the prior art was yours, and you patented said art within 12 months of its disclosure.

Yeah, so what's the workaround to that 12 months? When you file simply state you were working on it 12 months prior.

Not hard to game that at all. I'm sure business ethics will prevail though.

this is really cool. For a long time I wanted to make an internet noise machine that would make nsa surveillance too expensive/difficult. This feels like that same concept but for IP.

In theory, they could be one in the same.

This is very likely useless. Prior art has to be an "enabling disclosure" - which isn't the same as the 112 enablement requirement but it does requires that the public is "in possession" of the invention. This algorithm just makes short phrases or maybe a sentence or two that make sense together. It is unlikely that this is enough to put the public "in possession" of anything. There are problems to be solved in the patent system, but this is not a solution to any of them. See http://www.uspto.gov/web/offices/pac/mpep/s2121.html

edit: so apparently this guy is an 'artist' so maybe this isn't meant to be taken seriously. Perhaps its just a commentary on the absurdity of the patent system. If that's the goal, then cool, I think he's made a point.

not to be a buzzkill, but it'd be beneficial to reduce the data set examiners (and defendants) have to sift through to find relevant prior art. speaking as an ex-examiner, that's the biggest problem, separating the signal from noise from an already poorly written data set.

Digital representation of everything that was & will ever be done is already public domain: π.

Now to look for it… https://github.com/fenollp/minepi/blob/master/README.md

"The numbering scheme of the prior art is as follows, the first 10 digits is UNIX epoch time" ... what about prior art before 1970 ?

Applications are open for YC Winter 2020

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact