>>12787288>cont.As long as it uses traditional motors, it won't be as power efficient as a human, so it will need the heavy battery. With the high power usage, it will also need to dissipate the power efficiently, which makes soft squishy outer coverings a problem.
There are artificial muscles (polymers that contract when you apply an electrical current) that hint at a way of making a humanoid with a similar form and low power usage, but these typically have abysmal strength. Recently I read about a new one which finally matches human muscle strength. This is very new though, so it wouldn't be practical to employ it soon. And no one has even started on the software control scheme to handle robots using muscle groups instead of stepper motors. The pivot points might be the same, but in control terms each muscle group is a separate axis. It might be complicated.
And of course, developing competent object manipulation using human-like hands is an ongoing effort.
Some other elements, like facial expressions and poses, just seem like they could be a logical outgrowth of training a neural net on movie data and linking that to a feedback loop for reproducing the expression using muscle groups instead of pixels. That part seems easier.
So to put that together:
Bots that move like humans are currently emerging
Bots that interact with objects are in their infancy, and probably won't be emerging within the next 5 years.
Bots that are fleshy basically don't exist, and probably won't be emerging within the next 10 years.
But since we do have the important elements in development, I bet we'll have something pretty good in 20 or so years.
