The cat's kinda out of the bag at this point. But a non-profit public trust that acted as a patent-store to enforce the public dissemination of any derivative works based on the ideas maintained by the patent-store could make a huge difference ten, twenty years down the road. It would need an initial endowment to get started, retain a lawyer or two to manage it.

And then, publicize the hell out of it, evangelize the foundation over every college campus with a CS department. When students have established new state of art with ML, they can toss the design to the foundation in addition to arxiv, and where ever else they might publish.