A 'compute tax' is the wrong answer for the future of artificial intelligence and work

TEXT ANALYSIS: "A 'compute tax' is the wrong answer for the future of artificial intelligence and work"

The Dissection

This article performs two simultaneous operations. First, it argues against a specific policy instrument (the compute tax) using efficiency-framed economic objections. Second, and more fundamentally, it functions as a delay-and-deflect operation against structural acknowledgment of labor market collapse. The author wants you focused on tax design nuances while the mechanism she claims to doubt is already detonating in real-time.

The Core Fallacy

The Gradualism Assumption.

The article's entire architecture rests on this premise: "even when they look rapid in hindsight, take place incrementally. The importance of this distinction cannot be overstated, because it gives labor a chance to evolve."

This is the mechanistic error. The Discontinuity Thesis holds that AI automation is different in kind, not degree from previous technological transitions. Previous automations targeted physical labor; AI targets cognitive labor across all domains simultaneously. The author cites medical advances, language translation, and engineering problem-solving as AI benefits—precisely the cognitive domains where human employment historically concentrated. She cannot claim AI boosts productivity and that workers will smoothly transition. These claims are in direct conflict.

The historical precedent she's invoking (industrial revolution, steel taxation) involved physical substitution with cognitive labor remaining valuable. This transition involves cognitive substitution with no remaining large-scale cognitive domain where humans maintain durable superiority. The "incrementally" that "cannot be overstated" is precisely the variable that has changed.

Hidden Assumptions

1. Competitive parity is not mandatory. The author frames adoption as a policy choice. Under DT mechanics, this ignores the coordination impossibility: if competitors automate and you don't, you cease to exist as an economic entity. "Voluntary restraint" is not a real option in competitive markets.

2. Productivity gains distribute. "AI is...contributing to the GDP...raising overall prosperity in ways that can offset disruptions." This is the central distribution blindness. Under post-WWII capitalism, productivity gains accrue to capital owners. The "prosperity offset" argument only functions if you assume wage growth tracks productivity growth—and that link has been severed for decades.

3. Tax policy is the relevant variable. The author frames this as a fiscal design question. DT states the problem is structural—the circuit between mass employment, wages, and consumption breaks. Tax code modifications cannot preserve productive participation where productive participation no longer exists.

4. Retraining is a viable exit. The "retraining, ingenuity, and entrepreneurship" escape hatch assumes retrained humans can out-compete AI in whatever domain they enter. This requires either predicting AI limitations decades ahead (the author criticizes others for this) or believing humans will always find new niches faster than AI fills them.

Social Function

Ideological anesthetic and elite self-exoneration.

The article performs a precise class function: it allows readers who benefit from AI capital accumulation (broadly, anyone with financial assets, but specifically technology sector participants and their defenders) to:
- Acknowledge "disruption" without acknowledging destruction
- Accept "concerns are valid" while blocking every proposed response
- Frame structural collapse as premature speculation ("we don't know")
- Position themselves as epistemically humble while demanding action

The "we don't know with anything approaching certainty" formulation is the crucial hedge. It shifts the burden of proof to whoever predicts collapse while allowing present accumulation to continue unimpeded. This is not humility—it's the most convenient possible stance for those currently winning.

The Verdict

The article's objections to the compute tax may be partially valid on tax-design grounds while being entirely irrelevant to the underlying structural prognosis.

The DT framework does not require compute taxes as a policy. It requires acknowledgment that mass employment-to-wage-to-consumption circuits are collapsing under AI cognitive automation, and that this collapse is already in progress, not some speculative future requiring certainty before action.

The author's core error is not her specific critique of compute tax design. It's the framing of uncertainty as a reason for inaction while the mechanism she claims may not materialize is demonstrably operating. Coders are being displaced. Legal researchers are being displaced. Medical diagnostics are being displaced. The "incremental" she invokes to reassure is not what the workers experiencing this feel from their position inside the discontinuity.

The article wants you to feel sophisticated for recognizing complexity. The Discontinuity Thesis says complexity is not your protection. Structural mechanics do not pause for policy debates.