Extremely proud to be part of this project in a small way!! ❤️
Indian stories and innovation have such a long way to go!
A story on male contraceptive!
Started reading this book and gonna list out all the amazing books mentioned here!
Absolutely enjoying the range from philosophy to science fiction!
Books :
- Ender’s game by Orson Scott Card
- The Chess Computer Handbook by David Levy
Amazing to be part this discussion on forcevla2, and the general dynamics of contact rich manipulation involving parametrizing force modalities.
Thank you @aryanmadhaverma for organizing this, and thank you @itsarnavb & @_lagrangepoint for offering the space created with such beauty & intent. 🫶
session went well, was fun to whiteboard with people again after a while. thanks @itsarnavb and @_lagrangepoint for hosting us, and everyone who joined. you’re og
quick summary of what we talked about:
started by sketching a crude version of how we’d integrate force as a modality ourselves. image tokenised through ViT, proprio, and text encoded into a pretrained VLM, latent flowing into an action head in the same policy. also talked about action chunking on the decoder helping with inference latency and temporal coherence within a chunk, though long chunks risk drifting from the world
then started picking apart the paper. the action head is actually a separate altered transformer that ingests numerical force as input and outputs (delta_p, f), directly fed into the robot’s impedance controller which converts to motor torque and makes the arm respond to feedback. different from the usual IK-based position controller logic
discussed that the improvement over pi0 might have come from the model being steered to produce a force vector as output, not the MoE in the action head. paper doesn’t show with/without MoE ablations so can’t say
also briefly touched on interrupts for robots. training time we feed deliberately bad demos and recover. inference time we use DAgger and add a human intervention label and retrain. model has no internal knowledge of whether data came from human intervention or general teleop
some questions that branched off from our discussion:
- apart from force, what other modalities of input and feedback would make action policies more aware and accurate?
- slow and fast moving policies and how they get baked together at high frequency of operation?
- flow inpainting as an inference time optimisation for cloud based vlas? (under explored)
Successful Math meetup!
I love Bengaluru! ❤️
We discussed ideas around modular forms, upper complex plane, elliptical curves and hyperbolic functions.
Next one will be more hacker-space like.
Took a lecture on entrepreneurship at Jindal Global Business School @JGBSofficial . The class was part of Prof. Sahana V Rajan's course work. I really loved Prof. Sahana's approach to teaching and course work — a beautiful blend of technicality, inter-disciplinarity, business, philosophy, humour, inclusiveness, and empathy. And I'm really grateful that I got the opportunity to teach; such a meaningful thing to do.
So honored to sit in a class by Prof. Alladi Sitaram — a giant from India in harmonic analysis & representation theory! Something beautifully chill and fun about old retired mathematicians. ❤ Sir is 78.
Wrote a lorenz attractor simulator (Butterfly effect from chaos theory).
Play with it - https://t.co/DwP5F1vaAB
written in python. Easy to interact with.
Teach yourself fundamental idea of Chaos Theory.
Congratulations @ajeetbharath@radhikagandhi & @cradlewise to receive such beautiful words of acknowledgement! So proud. ✨
Fondly remembering how extraordinarily brilliant Bharath was in college itself! ❣️
we bought a lot of silly baby things that we haven't needed
but definitely i recommend a cradlewise crib and a lot more burp rags than you think you could possibly need