Lucky Kant Nayak

Just here for the intellectual alphas

[Lucky Kant Nayak @DiamondHead101]

@shishirny @anandmahindra Kudos !!!

[Lucky Kant Nayak @DiamondHead101]

@alanrosca @robkhenderson While wealth is one of the better metrics. But I would be dumb to measure smartness just by wealth and compare ppl with different interests and different industry

[Lucky Kant Nayak @DiamondHead101]

@space_colonist Can ppl still work in frontier AI labs without PhD?

[Lucky Kant Nayak @DiamondHead101]

@warikoo @SahilBloom Isn't excel now job of claude.

[Lucky Kant Nayak @DiamondHead101]

@Bunny_ngl Mr. Beast

[Lucky Kant Nayak @DiamondHead101]

@Bunny_ngl Hercules

[Lucky Kant Nayak @DiamondHead101]

Does anyone know what happened to perplexity?

[Lucky Kant Nayak @DiamondHead101]

@jesse_vermeulen Take a walk.

[Lucky Kant Nayak @DiamondHead101]

@Kangwook_Lee How can VLM judge for skills given for weird morphology?

Is there any finding that it gives reasonable scores for prompt and the video or seq of images ?

[Lucky Kant Nayak @DiamondHead101]

@Kangwook_Lee How is it different than Eureka and it's derivatives?
And is conpatible with any skills that I prompt for a quadruped or humanoid, for example, Ronaldo suiii?

[Lucky Kant Nayak @DiamondHead101]

@michaelpiseno_ @AlecSaysThings Thanks for your insight. The MimicKit codebase implementation doesn't have history of observations afaik, I'm also using one cycle of animated reference motion like walking or bounding for go2. Is it very less, and prolly discriminator is overfitting on that?

[Lucky Kant Nayak @DiamondHead101]

@ironwagh Love the work you have been doing. Just curios to know, the general perception of VCs in India, and how they are backing the upcoming Robotics startups, or do you seek funding from outside India.

[Lucky Kant Nayak @DiamondHead101]

@Charlygotyou Is it backed by any evidence?

[Lucky Kant Nayak @DiamondHead101]

@chris_j_paxton Design is terrible, especially companies adopting simplistic looks like Sunday Robotics.

[Lucky Kant Nayak @DiamondHead101]

@ChenTessler @Sentdex @atmoio Do you think we should keep in touch with the technical details of research like the core implementation of DeepMimic or CALM for example, if we don't know the details it would be harder to know what's going wrong when you bring them to actual real world deployment.

[Lucky Kant Nayak @DiamondHead101]

@eigenron Can second that!
One continuous lock-in session for hours and u would feel like on top of the world.

[Lucky Kant Nayak @DiamondHead101]

@sean_pixel @observie Could you please share the tool?

[Lucky Kant Nayak @DiamondHead101]

@sean_pixel Omg, what is this tool?

[Lucky Kant Nayak @DiamondHead101]

@KyleMorgenstein This is the current standard pipeline and theoretically it sounds correct.

[Lucky Kant Nayak @DiamondHead101]

@KyleMorgenstein My understanding of DR in context of locomotion is collecting trajectories from multiple MDPs, where each MDP has different p(s'|s,a) due to environment factors. If the policy can figure out which MDP it is working on ( proprio history), won't it take the optimal action?