I do kinda feel sometimes like all that's wrong in modern society is a few decades of goodharting everything from shareholder value to election winning to everything else.
I agree, but "goodharting" is too narrow a word. There's something bigger at play in objectification and optimization, and goodharting is only one of many harmful symptoms.
Yes I agree. It's not just overoptimization but it's looking at optimization to solve problems that aren't best treated as optimization problems in the first place.
Just happened to read about Jill Lepore's new book, "The Rise and Fall of the Artificial State" before I read this piece. Hard to tell from the back cover, but it seems like she's chasing some of these ideas. That bodes well for what you and your fellow cultural AI travelers are up to, as she is good at timing her topics for the moment (i mean that as a compliment).
Glad to see you put the suggestion into action! Now, just rinse and repeat.
In an old PBS documentary “Commanding Heights”, it is said in episode 1 that Indian statisticians were too good and thought they can optimize the whole economy! They were optimizing the joint diet of millions. It was the time of centrally planned economies. There is a great scene showing Mahalanobis around 56:00 in https://youtu.be/gfRTpoYpHfw?si=pLZc2MZa7F4KXfE8&t=55m50s explaining his magic equation
I do kinda feel sometimes like all that's wrong in modern society is a few decades of goodharting everything from shareholder value to election winning to everything else.
I agree, but "goodharting" is too narrow a word. There's something bigger at play in objectification and optimization, and goodharting is only one of many harmful symptoms.
Yes I agree. It's not just overoptimization but it's looking at optimization to solve problems that aren't best treated as optimization problems in the first place.
Maybe we can get together and figure out more optimal ways not to optimize everything so much.
I think James Scott's notion of legibility (not just for states) gets at this pretty well.
Just happened to read about Jill Lepore's new book, "The Rise and Fall of the Artificial State" before I read this piece. Hard to tell from the back cover, but it seems like she's chasing some of these ideas. That bodes well for what you and your fellow cultural AI travelers are up to, as she is good at timing her topics for the moment (i mean that as a compliment).
Glad to see you put the suggestion into action! Now, just rinse and repeat.
I should probably put it at the top and not in a footnote, but baby steps, right?
Looking forward to reading Lepore's new book!
In an old PBS documentary “Commanding Heights”, it is said in episode 1 that Indian statisticians were too good and thought they can optimize the whole economy! They were optimizing the joint diet of millions. It was the time of centrally planned economies. There is a great scene showing Mahalanobis around 56:00 in https://youtu.be/gfRTpoYpHfw?si=pLZc2MZa7F4KXfE8&t=55m50s explaining his magic equation
Diet optimization and game theory are topics that tickle my fancy! On the former, can't resist mentioning this old blogpost of mine: https://datta-analysis.blogspot.com/2014/07/a-nutritious-breakfast.html
Later I wrote a script to call an optimizer instead, with bizarre results as you can guess.