Toru

I won’t make it to CoRL this year, but our robot peeling work will be presented at Monday’s poster session! Paper, code and data are fully updated: toruowo.github.io/peel/

This work is almost a year old, and a lot has happened since. I checked this afternoon, and was relieved to find that peeling potato with a knife still isn’t something that Astra can do out of the box (yet)!

Anyways, most publications eventually become obsolete as new advances emerge, and AI is accelerating that cycle. While the specific methods used in this work might not last, I hope the idea does: a robot policy shouldn’t just get the task done, but also do it the way people prefer. How to incorporate human preferences into robot policies still feels underexplored, and I’m excited to see more work in this direction.

For any related discussion or follow-up, feel free to reach me by email :)

How should we approach research, when the very field we work on is going through rapid and potentially fundamental changes?

I recently gave a talk at the Stanford Vision Seminar, reflecting on my PhD journey and how AI is changing the way I think about research. I adapted part of it into a new blog post, organized around three questions:
1. What should we work on?
2. How should we work with AI?
3. What is a PhD for?

I hope some of my reflections are helpful. The talk ended with an open discussion, and I’d love to continue it here; curious how others are thinking about these questions.

Personally -- as much as there's FOMO, I also think life is very long, and we have time.

toruowo.github.io/blog/posts/202…

Inspired by recent thoughts from senior researchers whom I deeply respect (e.g. @JitendraMalikCV @Michael_J_Black @Ken_Goldberg @phillip_isola), I wrote down some thoughts as a junior researcher too.

The recent Astra demos in dexterous manipulation are truly impressive. But they also raise two questions that I don’t think academia has good answers to yet:
- How should credit be assigned when an agent synthesizes many prior works into a new research result?
- How can academia attract and retain talents when so much of its incentive structure relies on credit?

I reflect on these questions in more detail in a blog post. Curious to hear what others think, and debates are very welcome!

toruowo.github.io/blog/posts/202…

With all the discussions around Astra + robotics, I keep hearing the same question from friends in the field: what should we work on -- that won’t be solved by the next LLM?

I think there’re still plenty to explore!

One direction I’ve been looking into is hardware-algorithm co-design. @RewardAI_'s release today is a particularly exciting example: check out the cross-embodiment dexterity enabled by the Omnibody Hand & OMI-1! I've been a fan since @chenwang_j's DexCap; huge congrats to @chenwang_j @zipengfu @yifengzhu_ut and other friends at Reward AI for scaling things up to such an impressive system!

(Also planning to share my study notes on hardware for dexterity in a new blog post soon. Stay tuned!!)

Recently, I’ve been thinking about which fundamental skills are worth developing for the AI-native age.

One observation stuck with me: many people can speak fluently; but much rarer is the ability to convey a deep idea in just a handful of precise words.