I've waited for this very moment
"AI developers have immense power. They are a mobile, coveted population. Weapons and surveillance kit can only be built with their talent and assent. If they chose to wield their power for good, who knows what they could do?"
Well, I know. I'm going to take every chance I can to sneak into the sub-sub-sub routines a bit of machine code that identifies managers and targets them first. My first visual algorithm should identify software "developers" who wear dress shirts and spend more time with their fingers curled inside a coffee-cup handle than pressing the "H" key.
Software middle-managers first... Zap! Gone! Scrum lords next... Zap! Gone!
I will bake into the AI a learning routine to recognize the "type" that replaces them and then ... Zap! Gone! It will search Jenkins notifications for "targets" who receive the notifications but don't commit code... Zap! Gone!
My AI will identify those who commit code at least once a day and protect them. There are myriad problems to work out, as with any new system.
The first is that I'll never work on an AI targeting platform. I knew my plan was flawed from the start.
</sarc>The moral of this story is: there isn't a terrible technology used for bad that cannot be altered -- ever so slightly -- and changed for the good.