Cost depends on scale. In the past it didn't make sense to make a humanoid robot, because you had to have a human control the robot. Their essay reads somewhat like they believe the mental barriers of robots will be resolved faster than the physical aspects. My guess is that it's the opposite. Highly capable machines that do dumb tasks, because their brains are dumb.
A lot of the work they describe as being capable of being performed by robots but not yet cost-competitive are tasks done in discrete environments. Meaning the grail for automation is being able to stick cameras all over the factory and be able to translate externally monitored actions into an internally understandable perspective of the robot.
Their discussion of drywall work was pretty funny. Either I'm misunderstanding or they are. They seem to think the skim coat is substantially different material than the mud. If your goal is to accelerate and automate drywall it seems like developing a mud pump that regulates mud flow wall smoothing over joints as a purpose built tool isn't a big ask. Taping also doesn't have to be paper. Walls also don't have to be drywall.
Drywall is easily my least enjoyed project at my house. I am not good at it at all.
Quote:
Tasks for tapers are split between unexposed and highly exposed. Workers building interior walls use paper tape and a paste called mud to smooth over seams, joints, and screws in drywall sheets. Mud and tape are messy materials, so robots cannot perform the task "Press paper tape over joints to embed tape into sealing compound and to seal joints." After the first tape coat dries, workers "Apply additional coats to fill in holes and make surfaces smooth." This task is rated E3: an autonomous drywall robot uses AI to scan walls, spray additional coats of sealant, and sand walls once dry. Overall, by work time, 41% of tapers' tasks are exposed at E3, while robots cannot do another 26%.
.
