
Patents are granted for novel, non-obvious systems and methods. Would it be possible to patent a system for determining the likelihood that any given patent application would be granted and a patent assigned?
Before you dismiss this as tomfoolery, hear me out: So it's expensive to file a patent. And if there was a system for making an assessment of whether something might be patentable that would be a useful service. If you were considering spending thousands of dollars on a patent process, would you consider using a $9.99 service to make a pre-assessment of your patent's patentability? Maybe the system could even suggest ways to strengthen your application?
Well how would you do that in a way that was different than the way the patent office does? Could you train an AI system to evaluate patentability? You have all the granted patents. Presumably the patent office has all of the rejected/discontinued ones. If you used an evolutionary algorithm to randomly assess whether a patent was granted for a given patent and then train itself based on whether it was right over many thousands of patents, would the algorithm become adept at determining patentability akin to Alpha Go? Of course, you would include variables like date, the company behind the patent, etc. For surely having Microsoft behind the patent is worth a lot. When the system was mature, you could randomly have the system hack out paragraphs or claims for a given application and re-run the assessment. A user could then see that the inclusion of certain claims increased the chance of rejection and refine the patent application before submission.
I built this system one weekend and it performed admirably. I thought I might finally have invented something that I could retire on with my lovely family and spend some more time on a Polynesian island before they submerge beneath the rising seas. Sadly though when I tested this system against itself to see whether I should pursue a patent, it's probability of being patentable was very low. So I saved myself legal fees and decided not to pursue it further. Of course when I re-ran the simulation with Microsoft as the company behind it, the likelihood of patentability increased markedly. Maybe I could sell this system to them along with a bridge in Brooklyn. Do you think they would buy it? Remember this very publication of this idea allows me to pursue a patent on this for the next year. So hands off, ok?