Raise the Bar for Patents
A higher bar
Patents are supposed to encourage innovation by granting temporary exclusivity to the parties that researched and published the idea, but it often creates large legal burdens on companies that actually make things. I’ve previously proposed how the bar for patents should be higher than “non-obviousness”, especially for software. For the government to grant someone a monopoly on an idea, the idea should be something no one would have otherwise thought of - if someone else can think of the same idea, it would have reached the world either way.
A public mechanism
This criterion would be difficult for a patent clerk to decide, but a more market-style mechanism could be used. When a patent is submitted, the patent office can publish the problem that’s being addressed and the general area of the solution. If, within a fixed window (e.g. 90 days), another party submits a substantially similar solution, the patents “cancel” each other out and exclusivity is denied. Instead of “first-to-file” winning everything, the patent would only be granted to the “only-one-to-file”.
Incentives
For truly innovative ideas, there would still be an incentive to discover and patent the idea, since one still has a chance at getting a patent for it. However, once someone publishes a “problem” for their patent idea, why would other people have incentives to file solutions, if they won’t receive a patent? Many reasons:
Natural patent overlap - Many of these “invalidations” can occur naturally if multiple people file similar patents around the same time.
Multiple patentable solutions - If multiple unique solutions are submitted to the same problem, multiple patents can be granted.
To block the patent - Companies already operating in that space will not want to be forced to pay for ideas they were anyways working on.
For status - Other parties will still get credit for having co-discovered the idea, even if no one gets exclusivity on using the idea.
Small financial incentives - If necessary, fees that would have gone to the patent office to examine the patent can instead be given to people who co-discover the patent. (This would only be applicable if they weren’t also applying for a patent themselves.)
AI & Patents
In the short-term, AI could help facilitate the above process by reviewing the public problem statement for the patent and ensuring it’s fair, grouping similar ideas together, and helping to determine if submitted solutions match up with each other.
In the medium-term, AI could take over the process of reviewing patents. With the right formulation of the problem, the patent office can check if a powerful AI can determine the solution or not. If the AI can figure out the solution, it’s not patent-worthy, but otherwise it is. Before granting the patent, another AI would also review that the problem was formulated fairly. This method could eventually replace the current patent system independently of the above proposal.
In the long-term, AI will take over discovery and then both the existing laws and this proposed idea will no longer be relevant. The AIs themselves won’t need patents as an incentive to innovate, so the policies around patents will change.