postgrok

doomer-taxonomy

Properties3
tagsai, taxonomy, policy, discourse
typeconcept
statusseedling

3 min read

Doomer taxonomy (AI harm debate)

Definition: three-camp map of the AI harm discourse: Doomers, Accelerationists, and Centrists, each with an intellectual genealogy, a motivation, and a different meaning of the word regulation. The map is Cal Newport’s, drawn on camera in 2026-10-04-youtube-newport-ai-harm-taxonomy.

Origin

2026-10-04-youtube-newport-ai-harm-taxonomy (Cal Newport, “An Insider’s Guide to AI Doom”, posted 2026-10-01). The confusion in AI policy, per the episode, stems from multiple groups calling for regulation with fundamentally different diagnoses. The first filing of this page and its source attributed the map to Arvind Narayanan. That was wrong; Narayanan is one named example inside the third camp. Correction recorded in log.

Three camps

1. Doomers

Lineage: Extropians (1990s SV futurists) → Rationalists (Yudkowsky/LessWrong) → Effective Altruism (MacAskill/Ord/Bostrom, longtermism) → Anthropic, OpenAI. Core belief: superintelligence inevitable; careful alignment yields a transhuman utopia, carelessness yields extinction, so alignment dominates other moral calculations. Internal tension: lab insiders (Anthropic, OpenAI) think they can steer to good superintelligence. External rationalists (Yudkowsky) say the labs move too fast and will produce bad superintelligence. Policy influence: EA money lobbies Congress; EA fellowship programs have placed aligned staffers in congressional offices (Newport reports this second-hand from Hill staffers). The placement mechanism is documented: Open Philanthropy funded the horizon-institute fellows who draft AI bills (Politico, Oct 2023), and a parallel AAAS cohort is funded by Microsoft, OpenAI, Google, IBM, and Nvidia. Newport’s “this is a plan” framing exceeds what the reporting states. Grading at ai-doom-paper-trail.

2. Accelerationists

Composition: VCs (Andreessen, All-In), non-Doomer AI labs (Meta, Google), the White House and executive branch. Core belief: Doomers are wrong and weird. AI benefits are enormous, slowing down is itself harmful, and competition with China punishes delay. Motivation: money (VC returns, chip sales, export controls) and power (right-coded Silicon Valley alliances). Newport contrasts this with Doomer motive explicitly: return on investment is a familiar American driver; futurist ideology is not.

3. Centrists

Position: literally between the poles on the diagram. Agree with the Doomers that guardrails are needed; agree with the Accelerationists that Doomer ideology (superintelligence obsession) is misguided. Regulate AI like any powerful technology: specific techniques, specific safety guarantees. Three subgroups:

  • AI ethics: pre-LLM field, present and near-term harms (bias, surveillance, discrimination), early critics of the doomer philosophies. Shelf: Stochastic Parrots (FAccT 2021), see ai-doom-paper-trail.
  • NTCS, “normal technology computer scientists”: evaluate the actual technology, skeptical of superintelligence claims, wary of hype. Their book is Narayanan & Kapoor, AI Snake Oil (Princeton UP, 2024): “why AI isn’t an existential risk”. Named examples: Arvind Narayanan (arvind-narayanan), Gary Marcus, Newport himself. Newport’s concern: dangerous agent behavior driven by lab race dynamics, not by a bad god.
  • Cybersecurity experts: LLM hacking capability, a grounded threat model.

Key structural insight

Doomers and Centrists both call for regulation and mean different things. Congressional staffers primarily hear the Doomer frame via EA lobbying and staffing. The intervention that follows a doomer philosophy, Newport argues, misses the implementable work: focus on specific technologies and techniques with specific types of safety guards (26:21 to 26:39).

Graph View

Backlinks