The goal is to make systems which AI cannot run on, yet still have value for simple compute tasks. Everything else is a cover for this truth.
As for the goal of making systems which cannot run AI - this goal has always been intrinsically fulfilled by Moores law. For as long as we were progressing towards the singularity, we were incapable of it. Whether that is of importance now, as if it were ever not, is a matter of interest in the results: do we panic about AI now, or 20 years ago?
I think that there is enough of Paula Maddox out there to prove we don’t really need AI to do great things, just great people.