Accelerationism vs. Safety (e/acc vs. Decel)
The same disagreement, performed at higher temperature.
Kernel
By 2023 the AI policy debate had organized itself into two performative camps: e/acc, demanding faster deployment and less regulation; and "decel," the catch-all e/acc label for anyone arguing for slowing down, mandatory evaluations, or capability moratoria. The fight is partly philosophical and very largely tribal. Both camps share more premises than either admits, but their public registers do not allow it.
Frontline
FLI's 2023 "pause" letter (Yudkowsky escalates to "shut it all down"). The Biden 2023 executive order and its 2024 successor. The California SB 1047 fight. The 2024 OpenAI board crisis. Every product launch that uses words like "superintelligence" or "AGI" without scare quotes.
Doctrine — e/acc
The expected loss from slowing AI is greater than the expected loss from speeding it up. Markets and competition are the actual alignment mechanism. AI safety as a discipline is largely the worry of an unrepresentative academic elite. The 21st century needs more building and less PowerPoint about risk.
Doctrine — Decel
Frontier-AI capabilities are advancing faster than alignment understanding. The expected loss from misalignment is bounded by extinction. Voluntary pauses, compute-licensing, dangerous-capability evaluations, and selective deployment are minimum-viable governance. The label "decel" is a rhetorical attack, not a self-description.
Stakes
The 2024 U.S. election repositioned the entire alignment-policy debate. The new administration's posture is closer to e/acc than to the 2023 executive order. The EU and UK are moving the other direction. Whoever wins the regulatory layer at the national level for the next four years gets to set the global default.
Outlook
The most likely 2027 endpoint is a quiet compromise neither camp will claim: mandatory evaluations for the largest training runs, no moratorium, expanded export controls, public-interest interpretability research at the national-lab level. Both sides will declare victory and the underlying questions will remain.