A collection of opinions on dealing with the potential loss of tax revenue from AI.
>“The trend toward automation and AI could lead to a decrease in tax revenues. In the United States, for example, about 85% of federal tax revenue comes from labor income,” says Sanjay Patnaik, director of the Center for Regulation and Markets at the Brookings Institution. He suggests that governments address “the risks posed by AI” by increasing capital gains taxation rather than creating a specific tax on it, due to the difficulties in designing such a tax and the distortions it could generate. The repeated use of the conditional tense is because the impact of generative AI, the kind capable of creating content on demand, is still uncertain, both in positive terms — improved productivity and economic growth — and negative terms; job losses.
>…
>Daniel Waldenström, a professor at the Stockholm Institute for Industrial Economics, rejects the idea of a specific tax on AI, arguing that there has been no significant increase in unemployment, even in the United States, the birthplace of these new technologies and a leader in their implementation. He also emphasizes the difficulty in defining it precisely: “What are automation, robots, or AI? A chip, a humanoid machine, an application, or a computer program? We will never be able to define it precisely. We should continue taxing what already exists: income from labor, consumption, and capital gains.”
>…
>In the best-case scenario, the new jobs created by AI could be “more productive, better paid, and more accessible,” offsetting job and tax losses, predicts Patnaik. However, the latent — and very likely — risk remains that the process will not be automatic. Job creation could be delayed, less-skilled professionals could struggle to adapt, and a gap could emerge between countries — and within them — and across productive sectors.
zcleghern on
Land value tax would solve this
OrbitalAlpaca on
Every industry in America rushing to try to implement AI in its workflow to decrease headcount. No one is talking about what to do with all the unemployed.
Carlpm01 on
Consumption taxation would solve this.
ImmortalAce8492 on
I’m reminded of the conversation Tucker Carlson had with Ben Shapiro in which Carlson said he would support something akin to anti-automation policies.
At first glance, that sounds ridiculous. But the more you think about it, the more it raises serious questions; I don’t think neoliberalism, as a framework, is well equipped to answer.
If people are laid off at scale, what happens to them? I highly doubt we’ll meaningfully tax corporations to compensate for that displacement—we aren’t doing it now, and there’s little reason to believe we will in the future. That brings us back to Carlson’s underlying point: while inefficient, employing 100 people may outweigh full automation. Not in GDP or output terms, but in human terms. That kind of value isn’t captured by standard economic metrics. Ping Andrew Yang lol.
I still believe truly transformative AI is a long way off, but the trajectory alone raises uncomfortable questions. If GDP maximization remains the sole objective, then we should be prepared for backlash, political and economic alike, as people increasingly resist changes that threaten their livelihoods.
ycpa68 on
This feels like a Luddite headline
saltyoursalad on
Yes.
mostanonymousnick on
Taxing productivity gains seems (literally) counter productive, but then, I also believe the same about payroll taxes.
Demortus on
This is dumb. No one is trusting AI to do the entire job of a person (unless it was something very basic, like data entry), for the simple reason: you can’t hold a bot accountable if it makes a mistake that causes major losses for the firm. What’s happening is that people are offloading tasks that used to be very time consuming, like debugging code, to LLMs. That’s increasing the productivity of workers themselves, not replacing them, per se. Additionally, there are no individual “AI workers” whose labor you can tax. While you could impose a tax on LLM API transactions, this would be a direct tax on productivity (i.e., a really bad idea).
9 Comments
A collection of opinions on dealing with the potential loss of tax revenue from AI.
>“The trend toward automation and AI could lead to a decrease in tax revenues. In the United States, for example, about 85% of federal tax revenue comes from labor income,” says Sanjay Patnaik, director of the Center for Regulation and Markets at the Brookings Institution. He suggests that governments address “the risks posed by AI” by increasing capital gains taxation rather than creating a specific tax on it, due to the difficulties in designing such a tax and the distortions it could generate. The repeated use of the conditional tense is because the impact of generative AI, the kind capable of creating content on demand, is still uncertain, both in positive terms — improved productivity and economic growth — and negative terms; job losses.
>…
>Daniel Waldenström, a professor at the Stockholm Institute for Industrial Economics, rejects the idea of a specific tax on AI, arguing that there has been no significant increase in unemployment, even in the United States, the birthplace of these new technologies and a leader in their implementation. He also emphasizes the difficulty in defining it precisely: “What are automation, robots, or AI? A chip, a humanoid machine, an application, or a computer program? We will never be able to define it precisely. We should continue taxing what already exists: income from labor, consumption, and capital gains.”
>…
>In the best-case scenario, the new jobs created by AI could be “more productive, better paid, and more accessible,” offsetting job and tax losses, predicts Patnaik. However, the latent — and very likely — risk remains that the process will not be automatic. Job creation could be delayed, less-skilled professionals could struggle to adapt, and a gap could emerge between countries — and within them — and across productive sectors.
Land value tax would solve this
Every industry in America rushing to try to implement AI in its workflow to decrease headcount. No one is talking about what to do with all the unemployed.
Consumption taxation would solve this.
I’m reminded of the conversation Tucker Carlson had with Ben Shapiro in which Carlson said he would support something akin to anti-automation policies.
At first glance, that sounds ridiculous. But the more you think about it, the more it raises serious questions; I don’t think neoliberalism, as a framework, is well equipped to answer.
If people are laid off at scale, what happens to them? I highly doubt we’ll meaningfully tax corporations to compensate for that displacement—we aren’t doing it now, and there’s little reason to believe we will in the future. That brings us back to Carlson’s underlying point: while inefficient, employing 100 people may outweigh full automation. Not in GDP or output terms, but in human terms. That kind of value isn’t captured by standard economic metrics. Ping Andrew Yang lol.
I still believe truly transformative AI is a long way off, but the trajectory alone raises uncomfortable questions. If GDP maximization remains the sole objective, then we should be prepared for backlash, political and economic alike, as people increasingly resist changes that threaten their livelihoods.
This feels like a Luddite headline
Yes.
Taxing productivity gains seems (literally) counter productive, but then, I also believe the same about payroll taxes.
This is dumb. No one is trusting AI to do the entire job of a person (unless it was something very basic, like data entry), for the simple reason: you can’t hold a bot accountable if it makes a mistake that causes major losses for the firm. What’s happening is that people are offloading tasks that used to be very time consuming, like debugging code, to LLMs. That’s increasing the productivity of workers themselves, not replacing them, per se. Additionally, there are no individual “AI workers” whose labor you can tax. While you could impose a tax on LLM API transactions, this would be a direct tax on productivity (i.e., a really bad idea).