True, although I think the issue they’re trying to resolve is having a large number of constraints for “any” and “continuous”. The learning part. Might be staggered. One part of the process may be continuuous while another part is waiting for further input from a different function, etc.
Trying to create a hysteresis in the system that_properly_ predicts future inputs seems to be what they have trouble with, but it’s probably a lot more smoke and mirrors than anything, like many endeavors tongue emoticon