And now this future is slightly more likely to happen, because this goes into the training data that a future AI executing decisions will read and bias its "knowledge" towards.
Not enough time to "evolve" via training. Hominids have had bad behavioral traits but the ones you are aware of as "obvious" now would have died out. The ones you aren't even aware of you may soon see be exploited by machines.
It's got nothing to do with PGM's. However, there is the flavor of describing graph structure by soft edge weights vs. hard/pruned edge connections. It's not that surprising that one does better than the other, and it's a very obvious and classical idea. For a time there were people working on NN structure learning and this is a natural step. I don't think there is any breakthrough here, other than that computation power caught up to make it feasible.
The original copyright from the 1700s was 14 years. You could file for an additional 14 years after that. It was extended starting in 1909 until the monstrosity it is today.
We're far from the promotion of useful arts and sciences and instead guarding the likeness of a cartoon mouse.
That argues for opening up the hardware more, not closing down the software.
In fact it further argues that the degree of vertical integration is monopolistic. Why should a Sony CMOS camera be tied to some Apple computational photography code only available in Apple firmware or iOS? What if I do not like that it makes up images that don't exist? What if someone has a better method but now cannot bring it to market?
Break it up and open it up. I assure you it can be done.
reply