5 ms·
I think I may be in the minority for believing that patents for software could in some cases be reasonable, but the main problem is that the current system has
by abeppu 2y ago
I think I may be in the minority for believing that patents for software could in some cases be reasonable, but the main problem is that the current system has no reasonable mechanism for establishing 'obviousness'. Closely related, especially for software the system needs to consider when two mechanism are equivalent. In particular, I think:
- we should convene actual groups of "persons having ordinary skill in the art", who are not shown the patent application, who are presented with problems, and asked to describe a range of viable solutions.
- we should also use groups of such persons to judge whether two solutions to a problem (at least one of which comes from a patent application) are not meaningfully different.
Given these inputs, I think it would become much more expensive to determine that something is genuinely inventive (and applicants should bear that cost), but it should be possible to weed out a lot of obvious or overly-broad applications. If someone in the field who thinks about this for k weeks and arrives at "the same" answer, it wasn't that novel. I think it should also be possible for a group of "persons having ordinary skill in the art" to judge that an application is seeking something overly broad and that a PHOSITA should know that something is obvious, and incur some extra cost to the applicant.
However, if we could convince ourselves that stuff being protected was genuinely inventive, though we might be in a very different world, it doesn't seem crazy to me that e.g. a bunch of the work in ML research in the past decade should have been patentable because it wasn't obvious. It was repeatedly surprising! Of course, the current frenzy of activity might not be possible if it weren't for the spirit of openness that the field is adapted to. If the transformer was patent protected, who knows where we'd be. But it's really difficult to argue that concepts like the transformer architecture, or diffusion models or even something older like CNNs were "obvious" when experts, people in active research were surprised when they first arrived on the scene with the performance they had. And the don't seem obviously less worthy of protection than any of a range of physical inventions.
- whatshisface 2y agoAbout 300k patents are granted per year. There are not enough resources available to do it the right way.
- abeppu 2y agoYeah so I have to imagine that most of those are bogus, right? That's like if 1 in a thousand people in the country had a new "invention" every year. Considering life expectancy, that's like if a single digit percentage of people would get a patent at some point in their lives? So suppose to apply, one needed to pay for market-rate hiring of a jury of PHOSITAs, and all applications became more expensive -- then you'd only apply for the things that are strategically important and very likely to be approved. Overall the number of applications should go down and the quality of granted patents should go up. I think the main equity issue here is that someone who figures out something genuinely inventive and valuable is less likely to be able to pay for the process. And perhaps, how do you vet that someone is a PHOSITA acting in good faith?
- freejazz 2y agoRaising the costs to patent an invention is only going to benefit huge, established companies. A good start would be to prevent congress from funneling PTO money out of the PTO. As it stands now, congress takes the money PTO makes from applications and various other fees and uses it however congress pleases.
- abeppu 2y agoI think the other side of this is, how would one enter the pool of eligible PHOSITAs participating in establishing obviousness? If the fear is around blocking out individual inventors or small companies, that's reasonable -- but can you earn that money by being a reviewing participant in several other applications? Can we even give people and organization "application review credits" that are earned through reviewing?
- freejazz 2y agoThe PTO has qualified people reviewing patents. The issue is the number of patents they have to review and the time they have to do so. The former being numerous and the latter being non-existent. That's why bad patents get through. The reviewers just do not have the time to adequately delve into each application. Similarly, I have no idea on what basis you think that 300,000 applications would have to be bogus. We are talking about all possible inventions. I find a lot of the rhetoric on HN does not respect the problem-space patents exist in, i.e. coming up with comprehensible textual delineations in order to properly categorize literally everything that has not been invented yet. If you do not respect the difficulty of that problem, you are going to have trouble coming up with realistic solutions that actually achieve the goals of the patent system, which is something that the US constitution requires congress to implement.