Edit
So what are some of your opinions on the new optimus reel?
As usual, I'm bullish but practical on Optimus progress. Much like FSD, I don't think they are anywhere near final software architectures. Heck I think they are still using a lot of straightforward techniques for motion planning, and they are nowhere near a robust balance control policy (that's why it only walks very slowly) but there are some good nuggets in there.
1) Not breaking the egg. This shows their sensory feedback systems working quite well. I was a bit concered relying only on actuator torque deltas (expected vs actual) and rotation deltas wouldn't be precise, but this is an indication that it is quite precise and with minimal time delays. The actual software could still be a simple PD controller, so not sure any advanced software there. But still, validation of the sensors working well.
2) Trajectory learning. I ulimately don't think you want to explicity learn to track human movements exactly but generally mimick them. However the ability to feed in a human set of trajectories and learning robotic policy end to end from deep learning is definite progress. You are removing at least one step from the heuristic process.
3) The biggest reason I am bullish on Optimus is the abiilty to add value even without advanced refinements. The reaching / grabbing tasks shown were possible because the objects were light and the robot could easily keep center of gravity in between feet without advanced control NNets.
But there is a plethora of tasks where such restricted movement is okay. We can imagine tasks where the robot walks slowly to go pick up a smaller tool or piece of metal and walk to another spot, then calmly use the tool. Or hand material to a Kuka robot.
What this means is, unlike FSD, Optimus can begin generating real value for a subset of possible tasks well before generalized software architectures are realized.
Optimus can replace some real human workers, maybe starting next year. Workers whose total cost easily surpasses $100k / year in western countries.
I can't believe I'm saying this, but I actually believe Optimus will begin generating significant revenues before Robotaxi (not L2) does, and will never look back.