Kushal Pisavadia

Six days into the new year and I'm doomscrolling pretty hard. How 'bout you?

It is very strange reading @allconsuming_20 with no author attribution. It's nice, and I've read stuff I wouldn't normally. But I'm so used to finding more from the same people whose writing I enjoy that it's a hard response to fight.

I do wonder if part of this is related to the lockdowns. Hard to do wayfinding research when people: 1. Aren't going outside much 2. Can't meet other people easily

Tiny observation: there haven't been many big wayfinding/mapping research papers making waves the past year. Where have all the techy/mapping people disappeared to? Are they all at @googlemaps working on private things? If you know of any curious papers, please send them my way

I made this thread, and I'm now getting emails asking me about "AI" and to distill my thoughts on it. I didn't mention those two letters anywhere I like reading academic papers and posting some notes now and again.https://twitter.com/KushalP/status/1345001080111980544 …

Plan for this afternoon: put on a warm jacket and walk around outside listening to MF Doom.

Fin thread

What research are they doing to make proceeding training runs and experiments cheaper to run? Are TPUs worth it or are there better hardware paths that could be taken? How much of Google's funding is dependent on @deepmind falling in line with certain corporate goals?

How easy was it to dump the state of the system to do point-in-time recovery? How are they making it cost effective? How much of the system is actually utilised vs. idle?

How did they measure their progress? Did they write any tests (unit, integration, other) to check their working? What did they use to monitor it? Could they jump into strace to debug the running system and "change" it's path? Did they use distributed tracing or just images?

This is impressive work. I have so many (immediate) questions, but less on the side of RL. I'd really like to understand how the team went about operationalising this system.

I would like to see @deepmind publish more concrete examples of their findings when ready. It has been 5 years since AlphaGo was shown to the world. Could they use tools like @replicateai to help others to reproduce, learn from, and build on their work?

The team have published pseudocode to arXiv, but it's not enough to reproduce their work: https://arxiv.org/abs/1911.08265 

You might be wondering how you can run MuZero to play against it or understand how it works. Sadly, the best that we have is the Nature paper and the accompanying blog post.

A comment on the publishing approach with @nature: the paper was received 3rd April 2020, accepted 7th October 2020, and published 23rd December 2020.

MuZero is now at a point where I wonder if it could be used for more abstract (but related) fields of study. Take it from discrete fields, to continuous fields like wayfinding and orienteering. Can it observe mountaineers and figure out how to traverse the wilderness?