Art Kuo

Mostly science. Likely mishmash of biomechanics, neural control, reinforcement learning, machine learning.

[Art Kuo @KuoArt]

@chubicki Yeah the public interactions (incl bartender) all obviously tele-op. Dancing behind glass was autonomous but robots potted. Walking down street was real. All quite impressive, even tele-op was good but spoiled by the deceit. (Also too bad many ppl so gullible.)

[Art Kuo @KuoArt]

@francoisfleuret Extraordinary. I look forward to your upcoming "Little book of TikZ," where you share secrets of generating a 3D manifold from 2D commands and a very clever use of gradient defined along a line.

[Art Kuo @KuoArt]

@gabrielpeyre Kernel trick CAN be infinite-dimensional but need not be. A finite-dimensional polynomial kernel can also yield curved boundaries.

[Art Kuo @KuoArt]

@chubicki You're missing the truly amazing innovation here, which is ensuring more than six batteries are charged and working at the same time, and more than ten things on the same network! In the lab we thank the gods anytime that happens.

[Art Kuo @KuoArt]

@PKudzia_BioMech @lazarjov I'm skeptical of this. Max is too young to know what a floppy disk is. Actually you are too. You may want to check with it's actually an old vinyl Pink Floyd record.

[Art Kuo @KuoArt]

@BrookeOdle @JCashaback @Hope_Engineers @HopeKinesiology I was surprised that it’s easier to build a passive brachiating machine than a walking machine. “Yo magnets!” (Paraphrasing bc actual expression from Breaking Bad was more colorful)

[Art Kuo @KuoArt]

@BrookeOdle @JCashaback @Hope_Engineers @HopeKinesiology Here’s a video of a tinkertoy brachiating machine from Gomes and Ruina. https://t.co/R98WdiyZn1

[Art Kuo @KuoArt]

@PKudzia_BioMech @lazarjov Pawel please check your github, you have a pending invitation to a repo.

[Art Kuo @KuoArt]

@PKudzia_BioMech Nice! Hope you can take a bunch of pics/video. We should be able to produce a tutorial that any lab's undergrads could use. (We are working on a github repo…)

[Art Kuo @KuoArt]

We've long meant to share code, but lacked time to package well. Would you be willing to lead an open access project on this? We can support w/ sample code, but need some documentation, maybe some videos & samples. We could start a Github & a Discord channel. 1/

[Paweł Kudzia, PhD @PKudzia_BioMech]

G&P 2009: "A simple method for calibrating force plates and force treadmills using an instrumented pole" Does anyone have code/experience with this calibration they would be willing to share with our lab. @StevenHCollins @KuoArt @adamczyk_peter @Ferris_vball @maxdonelan

[Art Kuo @KuoArt]

@Miles_Brundage Yeah well Boston Dynamics $1.1B valuation, OpenAI $29B, Facebook $0.5B (was >$1B), Google $1.2T.

Boston Dynamics moves atoms around, the others mostly just electrons. Hydrogen weighs 1800x as much as an electron. (Not saying this is a useful comparison either.)

[Art Kuo @KuoArt]

@joshrbaxter @PKudzia_BioMech @StevenHCollins @adamczyk_peter @Ferris_vball @maxdonelan Very cool project, great to have drawings & code open access.
(Actually I don't like Collins C matrix convention. I prefer using transpose, so A*x = b is solved by x = A\b. More typical, plus Matlab \ usually selects the best numerical method automatically.)

[Art Kuo @KuoArt]

@KarlZelik @PKudzia_BioMech Very cool. Idea for Vandy senior capstone design project: Fully open-source instrumented pole <$500. Shopping list for off-the-shelf parts, drawings for fittings e.g. orderable from Shapeways. I think cheap uniaxial load cell possible using Aliexpress sensors + clever design.

[Art Kuo @KuoArt]

We should have mentioned in the paper: same calibration method can also determine force plate alignment/origin to better accuracy than standard L method, including potentially sub-pixel accuracy. With no extra effort. End^2/

[Art Kuo @KuoArt]

@PKudzia_BioMech if you're up for leading, we could consider making whole project public, and anyone could participate & contribute (& possibly co-author). End/