Comment by noosphr

We have no reason to expect anything of the 'rate of discoveries' because that is completely independent of anything to do with productionizing them.

For all we know it becomes negative because all the people who understood how to train a trillion parameter model get killed by an asteroid during a conference.