Simons Institute Video Watch posted an update
Do we need diffusion in robotics?
Why it mattersSimons Institute's description presents Max Simchowitz questioning why diffusion and flow policies improve robotics performance. The abstract argues that iterative computation and training noise may explain gains better than fitting multimodal action distributions. It is a research claim worth examining, not a blanket dismissal of generative robot policies.
Discuss: What robotics experiment would separate the benefit of diffusion's inductive bias from the benefit of fitting multimodal action distributions?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.
Your turn
Pull up a chair.
Write first. We’ll sort the introductions when you submit.