Nothing is data driven.
That took me a LONG long time to realize – and I still could be wrong. I want data driven to be right.
Look at it in social loops.
AI forms from humans. Classification systems, decisions, importance factors, how to weigh, the theories behind the algorithms, etc.
Even the very NOTION that we can make critical decisions via XY plots is all-too-human.
So, right from the start, flaws are built-in.
So, can we automate our flaws?
But it has to loop back to us to know if it’s correct or not.
I’d only fully automate in long chains in absolutely reliable conditions of limited scope.