First-to-file just determines which application has priority. But it’s still evaluated by the same rules, including vulnerability to prior art. These descriptions read more like free verse, or the monologue of someone smoking cannabis; they’re not rational descriptions of inventions and I question whether they’d carry much weight, so I’m seeing these as a first-generation run at the concept. One thing that would have value is to treat all past publications as a database (maintenance manuals, instructions, marketing brochures, blogs, science fiction etc.) and generate “invention” descriptions for each feature and combination of features. Add anticipated technologies too, for additional combinations. There’s so much obviousness in applications that still get through, this would be a service to almost everyone.
I wonder if we could pipe the output of this to a 3d printer and have it make everything?
Exactly. I could build a system in ten minutes that includes all prior art, immediately. It just takes a POST request with an identifier for the patent you’re looking for. The number is then rendered in BASE 64 encoding as letters, numbers & punctuation. Every possible set of letters, numbers and punctuation has some identifier that maps to it, so you every piece of prior art (as well as every book, scientific paper, etc.) is “in the system”. (As are all the gibberish sequences, and all the sequences that are almost right, but impossible in some subtle way.)
Using very low resolution, say 100 x 100 x 100 elements, every possible configuration is 2^1,000,000 or about 9.9006562292958982506979236163019032507336242417875 × 10^301029 different prints. Of course most of them will be a useless collection of loose bits of plastic. And we can divide that down further by eliminating symmetrical or rotational results, but still, that’s a LOT of PLA! 
The reverse actually. My understanding is that the general rule in the US is that if ANYBODY has published more than a year before filing, it is unpatentable, The creator* has a one year grace period after the time he publishes when he can still file. If he or she doesn’t make that deadline before filing, he has created prior art invalidating his own application.
*and who exactly can publish is apparently a poorly written part of the law.
ISTR that about 20 years ago the USPTO made a conscious decision to deny fewer patents. They were getting sued by people who had been denied patents, and that gets expensive. However when they grant a patent that seems fairly obvious, they are not sued, the holder of the patent is. So suddenly the “strike zone” for patentability increased greatly.
This project actually makes the patent process more expensive, since now patent applicants will have to argue that these generated “inventions” actually qualify as prior art. The invention here is to hinder start ups and small businesses who have a limited budget and depend on a patent portfolio to protect their business and get funding to stay in business
My understanding also, (from an expensive consult with a patent lawyer) once something is published,
(or manufactured and sold) it means there is prior art and not patentable.
So if you have an idea or product, have it patented first before selling or even showing it. Which is tricky since a design will likely have many changes and it is quite expensive getting a patent.
These aren’t inventions. They’re randomly generated word salad without any correspondence to actual concepts or products.
Now, as a research analyst in the patent field, I fully admit that the claim language of patent applications often appears to be randomly generated word salad without any correspondence to actual concepts or products. But there’s actually something behind the words, generally, even if the claims seem to be drafted by developmentally disabled monkeys. By their very nature these random word combinations cannot have an inventive concept. Even if I found one of these to, by some extreme coincidence, correspond with words to a patent claim I was investigating, I would not for a minute consider using it as prior art. These are by definition not analogous art. I would argue that they’re not even art, using the patent definition of the term. (The program that generated them is certainly art though.)
Art has to be analogous to read on a patent. This is by its very nature analogous to nothing.
That’s not prior art. It’s randomly producing a string that happens to be identical to another defined string.
it would be cool if it also randomly recombined the diagrams
In theory, the point of patents is to encourage publication of inventions which otherwise would be kept as trade secrets, so it’s entirely reasonable that if you already published it anyway, you shouldn’t be granted a patent.
In practice, as you say, read (or pay your lawyer to have read) the fine print for your jurisdiction, and any other jurisdiction you might hope your patents to have some validity in.
At the moment it is far too random.
But let’s say this algorithm becomes much refined to the point where it is itself capable of determining whether the word salad it has generated is a feasible concept and can decide whether to publish or not. Or even trawl the internet looking for real world problems and issues that need solving and attempt to publicly provide inventive solutions. Would it need a human overseer or intervention in the publication process in order to make its solutions prior art so that someone couldn’t just construct the invention it describes and patent it themselves?
The flip side is that such an algorithm could be used in private hands to patent useful things and make someone rich for inventing an algorithm to invent things rather than actually inventing the things they patent.
But have these irrational numbers been published?
It’s not randomly producing a string; the string is specified by the input (the identifier). Actually, https://libraryofbabel.info/ is already an implementation of this.
No. Most do, however.
It’s perfectly possible to imagine an infinite sequence of digits that never repeats but does not contain all subsequences. For example, 0.01100011110000011111…
A number that does contain all possible finite sequences of digits is called a “rich number” or a “disjunctive number”. There is also the concept of a “normal number”, which means that all finite subsequences occur with the same asymptotic frequency.
ETA: Does $\pi$ contain all possible number combinations? - Mathematics Stack Exchange
But as for patents, how explicitly do the claims of a patent application have to be spelt out?
I’ve seen claims of the form: “The system of claims 1 - 20, using Y instead of X”.
How about:
- A system comprising 42 widgets.
- The system of claim 1, with a different pre-determined number of widgets
Is that OK, or is it invalid because it implicitly contains an infinite number of claims which are not specified explicitly?
Well, how about the following patent application:
- A system created by taking a natural number N, creating a base-256 representation of said number N, interpreting said base-256 representation of said number N as a sequence of UTF-8 encoded characters, and interpreting said sequence of characters as the specification for the system.
Or how about:
- A system consisting of a sonic screwdriver and a bowtie.
- Claim 1, with the text “a sonic screwdriver and a bowtie.” replaced by an arbitrary sequence of Unicode characters.
I also like that you have blithely glossed over the moral implications of creating a simulated universe containing beings who think they are sentient (prove it!) just to invent stuff for you.
I think I have described the modern corporation.
I worked at a company where I was tasked with writing up information on our patents for the patent attorneys. I was coached by the attorneys in the patent process.
Unfortunately, the real process is WAY more complicated than people think.
Prior to my schooling, I thought patents protected the little inventor, like this: Invent something, apply for patent, patent office does a review, get patent, no one can make your thing.
Now I know patents protect the big companies with big pockets, like this:
Invent something
Apply for a patent on it and anything similar you can think of
Patent office does a review. They have their heads up their asses. They give you some kind of crazy response that shows they have no idea what you are talking about.
You rebut.
Wash, rinse, repeat for about 10 years.
You end up with a patent for something completely different from what you started out with.
Anyone can make your thing and you need to have lots and lots of cash to defend your own patent.
