I’ve been thinking about the paper Understanding Deep Learning Requires Rethinking Generalization.

Fiction gives concrete examples of abstract lessons, and your brain doesn’t care if it’s made up.¹

This book is about making better decisions. Unlike most books with that theme, it emphasizes thinking rationally.

See SMBC and Nick Bostrom for a longer version of my views.

Use port forwarding.

TLDR abstraction isn’t free or always desirable.

Did you know Vietnamese has never had its own native alphabet? They were all based off Chinese characters.

Misalignment is when the reward function you specify induces an optimal policy that you don’t like. At least that’s how I define it.

You can think of semigroups, monoids, and groups as the minimal structures needed to define and extend exponents.

This is an opinionated and incomplete book. These are sloppy notes.

One of the best interviews I’ve seen. Check it out.

These are so handy.

Beijing does 1 thing better than anywhere else I’ve been to: skewers.

To Show

Knowing lots of facts without anything to unite them is like having a crappy Tetris board.

The ladder of abstraction is more a jungle gym.