Should robots be taxed?

Years ago, Bill Gates wanted to tax robots. See versions of that argument at Quartz, CNBC, and the New York Times.

I can foresee two major classes of problems with taxing robots. 

First, there is a problem of definition within the taxing of robots. 

It is incredibly hard to define robots and automation in practice. When you look at Automation Direct, one of the biggest suppliers of robotic machines, a lot of the components they sell have both automated and non-automated versions that are exchangeable in practice. The programmable logic controller (PLC) is heavily used in semi-automated manufacturing, which requires lots of low-skilled labor to operate machines and inspect products between manufacturing steps. 

So even if you levy a steep tax on PLCs, countless other types of general-purpose computers can perform the same tasks. 

Then, there is the problem of economically efficient taxation.

Typically, taxation comes as a consequence of a new program being built out. So what does that robot assistance program look like? What are its structures? How much is it going to cost? Because if you are just taxing robots to tax robots, then you are just taxing capital. And how is taxing capital different from taxing the income of capital owners? 

In other words, this is a question of tax incidence. And the creation of a tax necessitates lots of justification for taxing capital.