Taylor W. Killian

@tw_killian

Assistant Professor @BYUCS | Member of Technical Staff @LilaSciences, interested in RL, Decision Making under Uncertainty & Generalization

Provo, UT twkillian.github.io Joined September 2015

987 Following 3.4K Followers 7K Posts

Latest posts 5

Pinned post

Taylor W. Killian @tw_killian Aug 3

Today is my first day @BYU! As promised in the 👇 thread (a few weeks late, c'mon I had to move while continuing to work, etc ;) ), I've written a post explaining my decision to return to academia _and_ my home town: twkillian.github.io/returning-to-a…

I'm excited for the journey ahead!

Taylor W. Killian @tw_killian May 14

📣 There's never a "best" time to share important updates, especially after sitting on this for so long...

I'm joining the faculty @BYU + @BYUCS this Summer as an Assistant Professor in preparation for the upcoming school year. Lots of excitement and a fair bit of nerves. 🧵

View quoted post on X

Taylor W. Killian @tw_killian 2d

How I look when I go to bed vs. when I get up in the morning #gettingold

Agneya Tharun @AgneyaTharun 3d

What if a decision model could replace the reward function in RL?

I trained two humanoids to walk from scratch. One used 12 hand-tuned rewards (things like staying upright, avoid jerky movements). The other used elephant-1 to judge how naturally it walked.

Results:

View quoted post on X

Taylor W. Killian @tw_killian 4d

As someone who welcomed their 4th child during their PhD, I cannot more strongly endorse this. Kids and family were a major part of what made my PhD program (through COVID) manageable.

Sarath Chandar Verified account @apsarathchandar 4d

Delaying having kids to focus on career is such a bad advice and I see it as a normal thing in North America. Kids bring purpose to life!
View quoted post on X

Taylor W. Killian @tw_killian 6d

This is really fascinating work that after seeing this result is making me re-think some things about my lab's core RL emphasis. If you can't beat an LLM coding the weights itself then it's 1) too easy of a task and 2) not the right algorithm.

Rikiya Takehi @rikiyatakehi Nov 2

4/6 Writing from scratch: given MinAtar game rules and code, Astra wrote a Breakout network's weights in a single reply.

With zero training or tools, it scored 20× higher than PPO trained for 500M steps!

View quoted post on X

Taylor W. Killian @tw_killian Oct 29

RL’s best venue for publishing is getting an upgrade!

Marlos C. Machado @MarlosCMachado Oct 29

We have been working on this for a while, so I'm excited to finally announce it. The RLJ is now accepting submissions on a rolling basis. You can submit your paper at any time, and papers accepted by a predefined date are expected to be presented at the next RLC conference!
View quoted post on X