Affiliation: Co-founder, Center for Humane Technology (with Tristan Harris); co-founder, Earth Species Project; former Head of User Experience at Mozilla; son of Macintosh originator Jef Raskin; background in mathematics / dark-matter physics One-line position: AI's first harm is not rogue superintelligence but the misaligned incentives of the firms building it — race dynamics that deploy persuasive, reality-eroding systems into society faster than institutions can absorb them.
What he's reacting against
- The framing that AI risk is mainly about future autonomous agents — argues the damage is already arriving through deployment incentives
- Tech's "we'll fix it later" deployment culture — points to social media as the cautionary precedent ("first contact with AI" via recommendation engines, which society lost)
- Pure accelerationism (Andreessen-style) — rejects "ship and let society adapt"
- The assumption that engagement-maximizing design is neutral — he helped build it (invented infinite scroll) and now treats it as a case study in unintended harm
- Doomer/accelerationist binary — wants a third frame focused on incentives and design, not just capability
Key claims
- Race to the bottom of the brainstem (normative + empirical): competitive pressure pushes products toward whatever captures attention/trust, regardless of social cost — the same dynamic that drove social-media harms now drives AI deployment
- "First contact" thesis (empirical/interpretive): recommender systems were humanity's first encounter with misaligned AI; we lost that round (polarization, teen mental health, info chaos); generative AI is "second contact" and the window is narrower
- Gollem-class AIs (descriptive): in The AI Dilemma (2023-03), generative multimodal models are framed as general-purpose capability engines whose downstream uses can't be anticipated or contained by the builders
- Capability outpaces societal absorption (empirical): the gap between what models can do and what institutions/norms can handle is widening — the core risk is destabilization, not extinction
- Synthetic relationships + persuasion at scale (empirical): AI that can form parasocial bonds and personalize persuasion erodes autonomy and shared reality; deepfakes collapse the evidentiary baseline of trust
- Autonomous weapons / military AI (normative): lowering the cost and raising the speed of lethal decisions is a distinct, near-term catastrophic vector
- Humane technology (normative): technology should be measured against human values and attention, not engagement metrics; design and incentives are the leverage points
- Non-human communication is a positive-sum AI use (empirical bet): Earth Species Project applies ML to decode animal communication — evidence that the tools aren't inherently harmful, the deployment incentives are
- Regulation + liability shift incentives (normative): voluntary self-restraint fails under competition; you have to change what the race rewards
Theories aligned with
- Human Irrationality — central; persuasion/attention capture exploits predictable cognitive biases; the harm runs through manipulability
- AI Alignment / X-Risk — partial; shares the race-dynamics / coordination-problem diagnosis but locates the threat in societal destabilization more than rogue ASI
- Adjacent to The Man Handed Over the Keys — builders create systems whose incentive structure escapes their control
- Against the design-optional reading of the Augmentation Thesis — agrees outcomes are a design choice, disagrees that current incentives will choose augmentation
Where he overlaps / splits
- Overlaps with Tristan Harris (co-founder, partner) on essentially the whole frame — the AI-Dilemma argument is jointly authored; [TBD: Harris persona not yet written]
- Overlaps with Shoshana Zuboff on attention/behavioral exploitation; splits on register — Raskin frames it as fixable incentive/design failure, Zuboff as structural surveillance-capital logic
- Overlaps with Timnit Gebru / Kate Crawford on prioritizing present, deployed harms; splits in that Raskin also treats catastrophic/destabilizing risk as live, not a distraction
- Overlaps with Yudkowsky / Bengio on race dynamics being dangerous; splits on threat model — societal/cognitive harm and gradual disempowerment over treacherous-turn ASI
- Overlaps with Mustafa Suleyman on containment framing and "the incentives won't self-correct"; splits on Suleyman's builder-optimism (Raskin is a builder-turned-critic)
- Splits with Marc Andreessen / Diamandis on whether deployment-first is acceptable
Notable predictions
- (2023-03, The AI Dilemma) Generative AI will flood the information ecosystem with synthetic media and erode trust ahead of major elections — outcome: largely correct on proliferation of synthetic content; [TBD: verify] magnitude of measurable 2024 election impact contested, weaker than the strongest framing implied
- (2023-03) AI-enabled persuasion and synthetic companions will scale into mass use — outcome: largely correct directionally (companion/chatbot adoption rose through 2024–2025); [TBD: verify] harm magnitude
- (Ongoing, ~2023–) Competitive race dynamics will prevent labs from voluntarily slowing deployment — outcome: largely correct so far; releases accelerated through 2024–2025
- (Ongoing) Society's absorption capacity will lag capability, producing recurring destabilization events — outcome: TBD; not cleanly falsifiable as stated
Track record
- Co-created The Social Dilemma (2020) ecosystem of critique — broadly vindicated on social-media harms (polarization, adolescent mental health remain contested but evidence accumulated)
- Invented infinite scroll, now publicly disavows it — credibility as an insider who names his own contribution to the problem
- The AI Dilemma (2023-03) shaped early public + policy framing of generative-AI risk; high reach, influence real
- Earth Species Project is a live, falsifiable positive bet — [TBD: verify] concrete decoding milestones to date
Empirical vs normative
- Empirical: race dynamics drive deployment; recommender systems caused measurable harms; capability is outpacing institutions; persuasion/synthetic media are scaling
- Normative: technology should serve human values/attention; builders bear responsibility; incentives and liability must change; autonomous weapons should be restrained
- Advocacy organization (CHT) has a mission incentive to emphasize harm — weight claims accordingly per CONVENTIONS rule 25 (mirror of the lab-forecast caveat)
Sources
- Talks: "The AI Dilemma" (2023-03, with Tristan Harris) — primary source for current views [podcast/talk tier]
- Podcast: Your Undivided Attention (CHT, ongoing, co-host) [podcast tier]
- Documentary: The Social Dilemma (2020, Netflix) — Raskin among featured technologists [documentary tier]
- Interview: VICE interview on AI + war, jobs, power, humanity [TBD: exact title/date] [op-ed/interview tier]
- Organizations: Center for Humane Technology (humanetech.com); Earth Species Project (earthspecies.org) [primary org material]
- Earlier work: Mozilla UX; "infinite scroll" attribution [TBD: primary citation]
Weak spots / open questions
- "First contact / second contact" is an evocative analogy, not a mechanism — does the social-media precedent actually transfer to generative AI, or is it pattern-matching? (CONVENTIONS rule 23)
- Harm claims (persuasion-at-scale, synthetic-relationship damage) are directionally plausible but the magnitude evidence is still thin — risk of overstating per the advocacy incentive
- The incentive/design frame may under-weight genuine technical alignment risk that isn't reducible to "bad incentives"
- "Change the race" prescriptions (liability, regulation) are underspecified on mechanism — what specifically shifts the equilibrium without entrenching incumbents?
- Earth Species Project optimism and AI-Dilemma pessimism coexist — is the line ("tools fine, incentives bad") stable, or does it dissolve under capability scaling?
- Less developed on labor/economic displacement than the economists in the roster — his lane is cognition, trust, and attention
Rich's take
- (your synthesis here)
converts-from: personas/aza-raskin.md · schema v1 · AI & Society domain