Bill Gates Did Not Propose A Robot Tax. He Described A Subsidy
The coverage focused on his remedy and largely ignored his diagnosis, which is the part even his critics accept.
Sign up for my Substack daily AI newsletter here.
See my AI Training course portfolio for corporate Business Leaders here.
Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/
Bill Gates published an essay in August warning that the world has no plan for the disruption artificial intelligence is about to cause. Within days, the debate had collapsed into a single question: whether governments should tax robots, and the answer arriving from industry, from economists and from the chief executive of the company selling the hardware was no. Almost nobody engaged with the observation underneath the proposal, which is a great deal more interesting and considerably harder to argue with.
CONTEXT AND BACKGROUND
The observation is about incentives. Employers who hire a person pay tax on that person’s earnings, every year, for as long as they are employed. Employers who buy the machine that replaces that person write the cost off against profit. Gates put it plainly, noting that the tax system nudges you toward replacing people with machines, and argued that a tax would slow the rush away from human labour a little while raising money for retraining and a stronger safety net.
The rejection was immediate. The International Federation of Robotics said the proposal aims to solve a problem that does not exist, arguing that taxing production tools rather than profits would damage competitiveness and employment.
INSIGHT AND ANALYSIS
What almost nobody disputed was the asymmetry itself. Even Oren Etzioni, the computer scientist who told reporters that Gates has the diagnosis right and the prescription wrong, and who dismissed the parallel proposal to tax AI tokens as measuring effort rather than displacement of workers, said a tax on robots that replace workers makes sense.
The strongest reply came from Jensen Huang, who said he loves Gates but does not see what he sees, and that his remedies would be different. His argument is that when companies become more productive they hire rather than dismiss, since most companies have ambitions for growth. He expects AI to be a net job creator at a scale never seen. Readers should weigh that Nvidia sells the hardware in question, just as they should weigh that Gates retains financial interests in the technology industry. Notice, though, that Huang disputes the consequence, not the mechanism.
There is a precedent that is routinely described backwards. South Korea is widely credited with the world’s first robot tax. It introduced no such thing. Under its Restriction of Special Taxation Act, companies investing in automation could deduct three per cent of the cost, or seven per cent for smaller firms, against corporate tax. The government reduced those rates to one per cent and five per cent. The measure was simply a reduction of an already existing benefit.
For South African executives, the point lands harder than it does in America. New and unused machinery used in a process of manufacture is depreciable here at forty per cent in the first year of use and twenty per cent in each of the three following years. Software and cloud services that displace white-collar work are treated more favourably still, deductible immediately as ordinary operating expenditure and carrying none of the levies that attach to a salary.
IMPLICATIONS
For directors, this is a procurement observation rather than a political one. When an automation business case reaches your board, part of the return is coming from the tax treatment rather than from the technology. That does not make the investment wrong, but boards should know which part is which.
For policymakers, the Korean example suggests the cheaper instrument. Reducing an existing allowance requires no new definition of a robot, no new administration and no new legislation of the kind that has failed twice in nine years. It is also reversible.
For everyone else, a note of caution. The objections to a robot tax are serious, and Etzioni’s point that taxing to slow AI in a competitive environment risks being overtaken deserves an answer. None of those objections touches the underlying asymmetry, which sits in the tax code whether or not a new tax is ever introduced.
CLOSING TAKEAWAY
The most quotable thing in the essay was the remedy, and the remedy is probably unworkable in the form proposed. The most important thing in it was the description of a distortion that no government designed and no government has examined.
South Africa runs an accelerated allowance for machinery, and immediate deductibility for the software replacing office work, in a country that cannot employ the people it already has. That may still be the right policy. What is not defensible is that nobody has asked the question, and that an argument about a new tax has crowded out the far cheaper option of looking at what the existing rules already pay for.
Author Bio: Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net
Comments