The cognitive ceiling. Research by Ericsson, Mark, and Newport shows that 3-4 hours is the daily maximum for concentrated effort. Beyond that, diminishing returns.
"Diminishing returns" is not the same as zero returns. You'll get more coding done if you work eight hours a day than four hours a day. There's certainly a point where the quality gets so low that the returns are negative (by introducing bugs / technical debt / stuff you have to rewrite the next day), but in my experience 4 hours is not it.
In fact, if the problem is very complicated then it might even take you three hours just to get up to speed with what you were doing the day before.
In UI jargon, "chrome" means the non-content UI that frames what you actually care about, by analogy to the decorative chrome trim on old cars: shiny, attention-grabbing "window dressing" around the "real" thing. Mozilla documentation from 1999 talks about “window chrome” as the browser’s UI framing.
Google named their browser "Chrome" as an ironic nod to minimizing UI chrome. So the name literally comes from the use of the metal chromium on cars.
In some ways yes, but this effect would appear with any kind of reinforcement learning whether it's neural networks or just fuzzy logic. The goal is to promote certain behaviors and if it performs the behaviors that you promoted then the method works.
The problem is that, just like with KPI:s, promoting specific indicators too hard leads to suboptimal results.
We can improve society by changing the system and by addressing the root cause. Accountability, in the sense of punishment after the fact, does little to improve it and is often counter-productive (see e.g. how the US prison system breeds more crime).
You: Cool! The entrance to the subway is around the corner.
Bob: Thanks for the help, friend!
You: You’re welcome! Good luck.