Last time I checked, I did not suffer from dementia. I can clearly list the insights I had from the book I’m reading now, but I remember nothing from the fifty something articles I read today on the internet. I can vaguely remember their titles though. One was “How the most successful people do something in the morning”, the other was “7 ways to be successful at something”, it was about making money I think. The rest were combinations of how-tos, lists, and general advices about life, business, and computer science.
But why the heck don’t I remember anything from them?
It’s probably because I just skimmed through them, and never actually read any of their content. The internet has taught us to skim. You already know that, right? We also know that skimming is rewiring our brains. We are spoon-fed more content than we can consume. It’s like an all-you-can-eat restaurant. You want to taste everything before deciding what to indulge, but while tasting the 53rd dish, your stomach started to ache, and you forgot to eat. When consuming content, my brain is torn between two modes: the explore mode and the exploit mode.
The explore vs exploit dichotomy
In machine learning — reinforcement learning in particular —you want the algorithm to discover new things, but you also don’t want it to spend too much time exploring dead ends, while it can exploit the successful options it has already found. I am skimming through dozens of articles trying to gauge if they they are worthy of my time. And while doing so, I forgot to actually read the good ones I stumbled upon. Computer scientists found a way to deal with the the explore and the exploit paradox. They let the algorithm spend sometime exploring, but make sure to spend some time exploiting as well. In practice, it spends most of its time exploiting and usually dedicate just 5% to exploration. We, on the…
