Robotic Tai Chi
- Thread starter Taiji Rebel
- Start date
Xue Sheng
All weight is underside
interesting....taiji, sure....master...no.... bad taiji yes.... but it is interesting
Taiji Rebel
2nd Black Belt
Yes, it made me laugh. Taijiquan teachers have nothing to fear. But it is interesting to see a robot clunk out some taiji-styled movements
The robot doesn't have flexibility waist.
If you don't move your waist and shift your body weight from one leg to another leg and back, you are not training "Taiji waist - silk reeling".
If you don't move your waist and shift your body weight from one leg to another leg and back, you are not training "Taiji waist - silk reeling".
Sammy Naur
Blue Belt
No offense meant, but given Tai Chi is one of those martial arts that few people seem to actually do well, i.e. "bad" (and I'm basing that on many opinions shared here by more experienced people), is this an example of people who don't understand Tai chi well teaching AI? As in was this AI fed on movements by a legit (read: hard to find) expert, or some average Joe or (worse) YouTube content?
As I understand AI, it sinks or swims based on the quality of material it consumes. Ergo, since there is so much "bad" out there, should we be surprised this looks subpar?
Similar robots are been taught to box, same issue. This is a slow, over long period of time training process. And if you watch this teleoperation video closely (e.g. 1:15) you'll even see the robot learning, making decisions that favor its own balance and stability over raw movement (in other words, just because the operator quickly shifts footwork, the robot goes slower because it knows better)
And then of course I remember you need to be cautious about videos like the OP, they might not even be real, like this one. And a lot of these videos, along with the OP, come from state run media. They are designed to make Chinese robots superior to all Western ones (e.g. Boston Dynamics).
As I understand AI, it sinks or swims based on the quality of material it consumes. Ergo, since there is so much "bad" out there, should we be surprised this looks subpar?
Similar robots are been taught to box, same issue. This is a slow, over long period of time training process. And if you watch this teleoperation video closely (e.g. 1:15) you'll even see the robot learning, making decisions that favor its own balance and stability over raw movement (in other words, just because the operator quickly shifts footwork, the robot goes slower because it knows better)
And then of course I remember you need to be cautious about videos like the OP, they might not even be real, like this one. And a lot of these videos, along with the OP, come from state run media. They are designed to make Chinese robots superior to all Western ones (e.g. Boston Dynamics).
