"If we keep on optimizing the proxy objective, even after our goal stops improving, something more worrying happens. The goal often starts getting worse, even as our proxy objective continues to improve. Not just a little bit worse either — often the goal will diverge towards infinity. This is an extremely general phenomenon in machine learning. It mostly doesn't matter what our goal and proxy are, or what model architecture we use. If we are very efficient at optimizing a proxy, then we make the thing it is a proxy for grow worse."
Too much efficiency makes everything worse: overfitting and the strong version of Goodhart's law
Bart Simpson writing on the chalkboard: "I will not waste CPU cycles. I will not waste CPU cycles. I will not waste
look at these glucose molecules, they're holding hands 😭
It's a shame you can't just keep taking gray market stimulants like caffeine and eventually achieve infinite energy.
There are three versions of yourself: the version of you that you see, the version of you that others see, and you as you really are.
"If you crush a cockroach, you're a hero. If you crush a beautiful butterfly, you're a villain. Morals have aesthetic criteria"- Friedrich Nietzsche
(smirking while handing you a hard shell crab) have a big bite dude. of course it's a soft shell crab
Beautiful Pit Viper (Trimeresurus venustis), family Viperidae, Thailand
Venomous.
photograph by Justin Coburn
LAURA MCPHEE B. 1958 GUARDIANS OF SOLITUDE