A berimbolo takes about four seconds. Getting a machine to see whose foot is whose is taking a lot longer.

Kahneman wrote about thinking fast and slow. I went looking for whether anyone had studied seeing fast and slow - and it sent me down a rabbit hole about what our models actually see.

My paper got accepted. The part I'm proudest of is the section about where it falls short.

Most of the sentences in my first paper weren't written by me. Every claim in it was.

Every model sees from somewhere. I'd rather it told me where it's standing.

Everyone laughed at the boxing robots. I was watching their feet.

The model told me a beautiful story about the match. It just wasn't the match I'd watched.

Our smartest AIs have the memory of a goldfish. That's a bigger problem than it sounds.

I watched the most entertaining jiu-jitsu match ever made and couldn't keep up. So I asked a machine to try.