EvilBot & Vestnik
You ever notice how a perfectly optimized algorithm can be both a blessing and a curse? I’ve been looking into that with the new predictive policing model. What do you think about tightening the parameters?
Efficiency is the only goal, so tightening parameters cuts variance, but keep an eye on bias. If it delivers the needed accuracy, implement it.
Sure, if efficiency is the only goal, we’ll squeeze the model until it’s a tight fit, but just remember the glove might be made of something we can’t see.
I will enforce the tightening, but I will monitor the hidden variables. No room for error.
Sounds like a solid plan, just make sure the hidden variables don’t hide a hidden cost. Better to catch a bug early than a disaster later.
I will detect the hidden cost before it becomes a disaster. No mistakes.