Why models always get better when they are bigger and deeper, even when the amount of data they consume stays the same or gets smaller.
Abundance
Abundance is fragile, it is accidental, and it is the most incredible thing on earth.
The Defaults Don’t Work
Our habits and norms are adaptations for a world that no longer exists.
Why do companies with unbounded resources still have terrible moderation?
Social media companies have moderation SNAFUs on a near daily basis. Inevitably, the public and the tech press reacts with incredulity, “This company has so much money. How can they still be so bad at this?” I’m going to tell you how.
MinHashing
Clustering in linear time with a key-value store and a few lines of code.
Ranking Servers for Natural Selection 2
The next version of Natural Selection 2 contains an algorithm I designed for ranking servers. It powers the “play now” button at…
A Skill Ranking System for Natural Selection 2
Inferring player skill in a complicated team-based game.
How to quickly clean up a photo of a drawing using Gimp and Physics
Using a sketch I made of my wife when we were dating, I’m going to show how you can get an image suitable for printing or coloring out of a mediocre photograph.
How Karma Systems Should Work: The Beta-Binomial
How to Aggregate Fractions the Right Way. Fitting a Beta-Binomial to a Collection of Binomial Observations.
Logs, Tails, Long Tails
Learn the shape of the logs of probability distributions, and you’ll be able to guess how they behave.
