> But with AI (disembodied in software and embodied in robots), the relative value of being normal and predictable is rapidly evaporating.
I would argue this is precisely backwards. The normal and predictable is more valuable. Human errors are largely predictable. AI errors are not. This is one of the reasons why human code is much easier to review than AI code.
I would argue this is precisely backwards. The normal and predictable is more valuable. Human errors are largely predictable. AI errors are not. This is one of the reasons why human code is much easier to review than AI code.