Re: Knowing what to measure
In re "...If AI/ML is worth anything, surely the machines will soon have the capability to perform their own self-tweaking?"
Yes, yes, yes, Lyndon. This is the logical conclusion, if we take all the breathless hype around AI at its face value. The algorithms that can teach themselves create ever greater efficiencies and savings have it handled!
I don't know if this actually true. But the flavor-of-the-month way the industry is talking about artificial intelligence and machine learning certainly suggests they believe it.
As AI/ML is substituted for human endevor, one of the multiple problems I see coming involves quality. Businesses (and lots of public-sector agencies also) are continuously looking for a cheap fix in a kind of desperate drive to reduce ongoing costs. AI/ML may work for some substitutions, but with the level of current technology, only up to a point. Nevertheless, automated substitutes are often being implemented if they're "almost as good" as the human performance they're replacing. Basically, technology that should be an adjunctive aid to human work, and a way to boost productivity, is being used as an "almost as good as" replacement for the humans themseves.