Keep pulling the thread on Mallory McMorrow.
A primary argument for an AI tax, championed by politicians like Elizabeth Warren and Mallory McMorrow, is to generate revenue for social programs. This includes funding universal healthcare, free education, and retraining programs for workers displaced by AI-driven automation.
The discussion presents a "first principles" case for a token tax based on tax neutrality. Currently, human labor is taxed (payroll, income), while AI "labor" is not, creating a fiscal incentive to automate. A token tax could level the playing field, ensuring the tax base follows the locus of productive capacity, whether human or machine.
Critics like David Friedman detail significant practical problems with a token tax. These include tokens being a poor proxy for economic value, the "tokenizer endogeneity problem" where different languages are taxed at different effective rates, and the issue of rapidly declining per-token prices making a fixed tax unsustainable.
A major concern raised is that a token tax would stifle innovation and entrench incumbent players. By increasing the cost of experimentation, it would create a "known ROI bias," discouraging novel applications, while large firms could more easily absorb these costs or find workarounds.
Opponents like Palmer Luckey argue that a US-centric token tax would create a competitive disadvantage for American AI companies. It would make foreign AI models and services more attractive and could lead to companies domiciling elsewhere to avoid the tax.