Festo is.[1] Over the last decade, they've built a robotic bird, a robotic fish, a robotic kangaroo, and a robotic hand. Many are pneumatic, and their technology for precision pneumatic control is available for industrial applications.
I don't follow cycling as much as I used to, but in an attempt to keep rich teams from building bicycles out of unobtanium, they put some minimum weight restrictions on the bikes. One of the early results was that some telemetry gear started getting built into these bikes. The biggest one I know of was building a torque meter into the rear wheel. For the cost of a fraction of a percent of power output, it could tell the coach if you're really going all out or dogging it. The difference in those numbers would hint at you being sick, a subtle equipment failure, or an aerodynamics problem.
It should take a certain range of force to open a door, drill a hole, tighten a bolt, or raise a platform. If you are using more than that, something is jammed, your bit is wearing faster than expected, the bolt is cross threaded, or something is stuck on the platform and you're trying to ram it through the roof.
Some tools are very difficult to scale up to high torque situations, but it seems like this data should be available to a robot, via sensors on the power supplies if nothing else.
I think of Legion from the Mass Effect games. It's a robot with a blank circle for a face... but with plates around that circle that are animated in a distinctly eyebrows-like way.
The future will probably be some sort of artificial muscle derived actuator, absolutely. But for now the only way to do anything useful with a humanoid robot is with hydraulics (Atlas), huge electric motors that are either too slow or water-cooled, or sacrificing capabilities to focus on just a few things (Cassie)
Have you seen Digit? (https://www.youtube.com/watch?v=LSk8uCHN5CY)
The article mentions that we (robotics researchers) don't think about movement first but the fact of the matter is that those of us working on humanoids can't STOP thinking about movement and how hard of a problem it is to solve. The issue is that we don't have anywhere near the technology in either electric motors or hydraulic actuation that is required to solve the problem of smooth, high-fidelity, accurately-controlled motion that is also safe for human co-operation (which typically means torque control or some other scheme for feedback control over the impedance between the robot and the environment). Hydraulics are far too "stiff" and dangerous to be used in robots that work directly alongside humans, and electric motors can't hit the torque vs. velocity requirements for dynamic motions that allow the robots to be both safe and expressive and dynamic like a living organism.
But nobody is funding research in to actuation because it doesn't have short term payoff and it isn't sexy. So here we are.