tomahunt

Oh, I thought he had derived a math identity.

The story of Feigenbaum presented in Gleik's Chaos is the reason I chose to do physics.

"Hacker school" rebranded to "The Recurse Center"

www.recurse.com

Theoretical physics and applied maths. I occasionally see people at the back of seminars with their laptop open, pugging away at their current puzzle with Mathematica.

I've used it to test and prototype calculations and algorithms.

The Stackexchange page for Mathematica is pretty amazing, and quite a good community.

The main sections of the talk without note are:

- Bert and Natural Language processing, - Tesla Autopilot Hardware v2+: NN at scale, - AdaNet: AutoML with Ensembles, - AutoAugmentation, - Training Deep Networks with Synthetic data, - Segmentation Annotation with Polygon-RNN++, - DAWNBench: Training fast and cheap, - Big GAN: state of the art in image sythesis, - Video to Video Synthesis, - Semantic segmentation, - AlphaZero and openAI Five, - Deep learning frameworks

"Can you make the error bars a bit smaller please?".... "um no, I'm affraid they are as I've calculated "

A medical doctor friend of mine pointed out that thinking slow was fine, except in an emergency or indeed any time--limited situation where the rate of response needs to be above some threshold. Think: baraster responding to opponent, a and e doctor, board-room negotiation etc.

Exactly the question I would love answered.

Edit: this seems roughly answered to a question by user r4um