Because a human body has no capability of controlling a non-human design. My fingers bend the way human fingers bend, I can't make them do anything else.

If you all design an interface to emulate human behavior then it needs to have human capabilities and human limits otherwise I can't control it.

The idea isn't to be hyper specialized to a specific task. It's to be hyper generalized to fit into spots already being filled by human workers. The goal is for the machine to be placed in the role of a paid human worker without the need to specialize anything else in the environment, a drop-in automation solution.

"All we're missing is the single hardest piece, which we have been failing to make work in cars for years"

Based on my experience with how destructive a robot vacuum can be, there is 0% chance I would let a Tesla developed robot exist in my house.

The automation tax that gates/etc proposed to fund UBI/social support networks is making more and more sense.

I'm all for UBI, but the automation tax is a quagmire.

In this theoretical new tax, tell me what qualifies to be taxed?

• An Atlas autonomous robot? Sure, absolutely. How about instead a hydraulic arm that is controlled by a human? Previously there were 4 humans that moved the widget from A to B, but now they have 1 human operating a joystick for a net loss of 3 jobs. Is that taxed?
• How about an Excel macro? Prior to the macro, there was one person filling in the spreadsheet the entire 8 hour workday. Now that person was replaced with an Excel macro that runs in 5 minutes with one click. That is automation too right? What would you tax? The cost of the person replaced?
• Who pays the tax? A company that buys an Atlas robot after the law is passed? Absolutely. How about a company that bought Atlas robots 24 hours before the law passed? How about the company that bought them a year before the law passed? Now apply the Excel macro automation. Excel macros have around since the 1990s. Are you going to go back to the first macro run and tax every company retroactively? How about if the macro only does part of the work?

Automation tax is a nice idea but a nightmare to try to make in policy. Additionally, it will have a stifling effect on any business efficiency efforts after it exists.

If the tax is based upon workers losing their jobs to automation, it will have a massive knock on effect limiting new hires. A company would be very leery of hiring a worker if they could be accused (and taxed) of automation replacement when that worker is let go.

The point with factory work is that you don’t need half of what this robot can do if you change the plan of the factory a bit.

So no I don’t think the idea here is for standard factory work.

You're changing the problem that is being solved. The CURRENT work process is to use a human with all the benefits and detriments of a human. The idea would be to drop one of these Atlas robots in without changing the work or work environment. Perhaps there is a more efficient human doing the work from 8am-5pm and only some work needed from 5pm-8am. An Atlas robot would be perfect use case here. You don't have to redesign the work or the environment for a human or robot to switch out to do the same work.

What you're describing is changing the nature of the task or the environment to optimize for a robot.

• Flat floors? Just use wheels instead of legs.
• Short distance to cover? Drop the entire torso and head and just be an arm with a camera.

Boston Dynamics already has that robot. Its called Handle:

As you can see, its a wheeled robot with an arm, but this robot couldn't do the task that the Atlas robot can in the video because it doesn't have the fine motor control or fingers to grasp the engine covers, nor does Handle have the ability to deal with those soft pliable racks where Atlas is placing the covers.