I would think there are many changes that will be coming to the entire patent process. In particular, AI discovery of new inventions will throw a wrench into the system at it's core. What happens when 6 months from now, 90% of new inventions were originated by AI.
My take is a lot of patents are really poor quality. (Worked on a patent search engine so I have looked at a lot of them)
This year I took an interest in Heart Rate Variability biofeedback and developed my own system which I haven't found in the literature (patent or otherwise) [1]
I have looked at a lot of patents on the subject and to me they are trash. They blather on with literature reviews, describe complex systems which violate all the principles of biofeedback (e.g. like they have you do "resonant breathing" at a fixed frequency or compute scores over a 2 min interval rather than look at the biosignal in real time which is the basic principle of biofeedback -- if I just want some metric to measure my health I will take my blood pressure the way my doc wants me to!)
It's shocking because I can't believe any of these people ever seriously tried it!
Maybe a percent or so of patents are well-written, clear, beautifully illustrated and describe genius inventions. My favorite is this one
which is for the plastic PET bottle of which half a trillion or so are made a year -- something brilliant and demonic (in terms of waste) at the same time!
---
[1] to document as prior art: a healthy heart has variations of HR over a time scale of 10s or so called the Mayer Oscillation. The amplitude of this oscillation can be estimated by averaging the metric known as SD1 over a 20s interval:
You can increase the amplitude of you Mayer oscillation by: plot instantaneous heart rate (1/R-R interval) as a function of time; watch the slope beat by beat, if your heart rate falls breathe out, if it rises, breathe in. if your Mayer oscillation is weak you might have a hard time doing it and feel short of breath for a moment but usually your Mayer oscillation quickly strengthens (less than 1 min) and once it is strong it is very easy to keep pacing.
I built Grounded Discovery Labs, a system that takes a science or engineering problem statement and produces a ranked portfolio of candidate concepts, along with decisive experiments and kill criteria.
This is not just a thin AI wrapper around a model. The point is not simply to generate ideas, but to force candidate concepts through a structured workflow that pressure-tests them, repairs weak points where possible, and tries to separate stronger directions from noise and hallucination.
