Jeff Dean's NIPs slides

Notes and take away ideas from this seem to be popping up all over the place today. Jeff Dean’s (from Google Brain) presentation slides from NIPS. Here’s a link to a hacker news thread on it (with link to the slides). If anyone has a video link yet that would be amazing.
It seems kind of incredible what is starting to be done there (granted the compute power of those TPU clusters is a little out of reach for mere mortals for the time being).


Very impressive examples. They have cracked how to reformulate engineering problems to be solved much more efficiently using reinforcement learning.

good stuff, and along the lines of Karpathy’s Software 2.0

Yes exactly. It’s great to see a real example.

Looks similar to a lot of the talks he gave this year, that should be up on YouTube, but unfortunately I don’t have the NIPS link to this not sure who has posted it.

1 Like

thanks - will look for one of the other talks

1 Like

Ya I’d also double check Twitter feeds to see too.

Everything is a model - Delip Rao chimes in, echoing Jeremy :slight_smile:


Thank you for linking this!