Media

Todd Sundsted @toddsundsted

another great office game night!

the best writing about artificial intelligence on the web

distill.pub/

the challenge of last-mile logistics in a city like NYC

dear god, someone sent me a username/password in an email!

I can’t decide if the problem here is humans, ancient email technology, or something else...

our new kickstarter came this week - el dorado.

superficially i thought it looked like catan, but other than the existence of resources it doesn’t play like catan at all.

enjoying it!

Bill Dally - Scaling of Machine Learning - ScaledML 2018

excellent dive into deep learning hardware, GPUs (obviously), lower precision weights, trained quantization of weights, pruning of neurons and weights...

youtube.com/watch?v=h3QKvUPg_A

researching the history of artificial intelligence via magazines and newspapers... side tracked by the old hardware in the advertisements...

when i think about autonomous vehicles i always hear red barchetta playing in the distance...

norvig.com/cacm-hybrid.html

"Google's Hybrid Approach to Research" - Spector, Norvig, Petrov

finally getting around to reading "Playing Atari with Deep Reinforcement Learning". you'd assume, given how I spent my youth, it would have happened earlier...

i just love these interactive overviews/presentations of recent research

distill.pub/2018/building-bloc

plan for today: stir pot, organize resulting chaos, hack hack hack, nap, write, study... (kidding about the nap)

`stack build` is pronounced "stack set your laptop on fire"

@seanl @phenethylamine @lain

we went with haskell for a bunch of reasons including solid multicore support (which ocaml lacked) and support for docker (which impacts jvm-based languages), but learning to use haskell effectively has been hard, hard, hard. it really feels like you're starting over again.

a loaf of bread
a jug of wine
a typeclass
and thou

#pointfreeverse

i wonder how long until someone designs a cryptocurrency to run on one of these machine learning chips...?

IBM explains how much faster their machine learning solution is (apparently to kindergarteners)...