Andrew Davison on Twitter: "By the way I'm a huge fan of the original Bitter Lesson argument about how algorithms scale with hardware; see for instance https://t.co/DYgeZVaBHO" / Twitter

Good "Bitter Lesson 2.0" argument, but I'm not sure big data is the main thing robotics needs right now. Main challenge is still perception -> efficient/accurate/physical 3D representation. I suspect the methods that work will be hand-designed or will train locally in real-time.

Quote Tweet

Karol Hausman

@hausman_k

Jan 9

Bitter lesson by @RichardSSutton is one of the most insightful essays on AI development of the last decades. Recently, given our progress in robotics, I’ve been trying to predict what the next bitter lesson will be in robotics and how can we prevent it today. Let me explain

Show this thread

Andrew Davison

@AjdDavison

Jan 12

i.e. if your perception system can turn raw sensor measurements into a 3D simulation of the world with enough fidelity, rapid real-time simulation allows planning or local learning, even for completely new actions, and robots could do new, creative actions.

Andrew Davison

@AjdDavison

By the way I'm a huge fan of the original Bitter Lesson argument about how algorithms scale with hardware; see for instance gaussianbp.github.io

5:29 PM · Jan 12, 2023

1,294

Views

Likes

Conversation