Re:Ask HN: Should the teaching robots, like ATLAS, Martial A...
by smoldesu on Sunday June 20, 2021 @08:12PM
(#27574987)
I'll take the bait. I used to drive for an FRC team, and we'd write demo programs similar to (but much less impressive than) Boston Dynamic's demos. Those demos are static though, and have a limited degree to which they can react to outside intervention. The field of inverse kinematics has a long ways to go before we need to be worried about this sort of thing.
[ Parent
| on HN ]
Re:Ask HN: Should the teaching robots, like ATLAS, Martial A...
by PaulHoule on Sunday June 20, 2021 @08:16PM
(#27575012)
It's true... Even in 2021 robots are not graceful. I think of those old videos with Honda's Asimo that look a little painful. Asimo solves for the configuration space that the joints move through, it can't take advantage of the momentum of the body, so it works really hard to move awkwardly.
Newer robots are better but still have a way to go.
Re:Ask HN: Should the teaching robots, like ATLAS, Martial A...
by samstave on Sunday June 20, 2021 @08:26PM
(#27575058)
Thanks to both of you for the reply.
Agreed that we are "not there yet" - but given that they are physically capable of the moves we have seen in this demo, irrespective of it being a static demo, the balance and complexity of poses, pose-transitions is what is scary to me -- all the other AI stuff mentioned is already here, in possibility, no?
Re:Ask HN: Should the teaching robots, like ATLAS, Martial A...
by smoldesu on Sunday June 20, 2021 @08:49PM
(#27575192)
The issue is mostly that a humanoid robot is a horribly inefficient way to fight someone. Sure it's uncanny, but you'd have a lot more luck (and save a lot of money/R&D) if you just taped some plastic explosive onto a quadcopter.