Sep 18, 2026 · 22 min
AI’s growth model makes slowing down nearly impossible
AI can’t afford to slow down | Ed Zitron
The episode shifts attention from speculative superintelligence to the present harms, financial commitments, and accountability gaps shaping AI development.
- 1Superintelligence warnings can distract from practical oversight of AI systems already affecting people.
- 2Operational constraints and economic commitments may look like a slowdown without reflecting deliberate restraint.
- 3Regulation must address current harms rather than waiting for speculative existential risks to materialize.
Don't miss
Ed Zitron connects the industry’s compute commitments and revenue expectations to the difficulty of voluntarily slowing AI development.
The brief
Ed Zitron argues that superintelligence and recursive self-improvement claims can obscure the immediate harms of existing AI systems and the need for practical accountability.
Isaac Pound and Zitron test whether legislation, industry choices, or operational limits could produce a genuine slowdown in AI development.
Data-center delays, shortages of quality training data, and strategic messaging raise a harder question: are companies slowing down, or rebranding constraints as restraint?
The conversation traces superintelligence fears through rationalist and effective-altruist communities, then contrasts that narrative with the case for immediate oversight.
Massive compute commitments, projected revenues, and pressure for venture returns make continued expansion economically difficult for the AI industry to abandon.
The episode’s central conclusion is stark: AI may not be able to afford a slowdown, making regulation and accountability more urgent.