# Rationalism ## Definition Rationalism, in the context of this knowledge base, refers to a specific Bay Area-centered intellectual community—not the philosophical rationalism of Descartes and Kant. This is the epistemic movement centered on [[LessWrong]], [[Eliezer Yudkowsky]], and the [[MIRI]]. The community emphasizes: - **Bayesian reasoning** as a mathematical framework for belief-updating - **Systematic cognitive bias correction** — the premise that most human reasoning is riddled with predictable errors - **AI alignment and safety** as existential concerns requiring elite technical intervention - **The belief that most people are systematically irrational** — and that the rationalist community thinks more clearly than the general population The movement functions as both an epistemology (how to think correctly) and a subculture (who the correct thinkers are). It serves as the social and intellectual infrastructure for [[TESCREAL]], providing the jargon, social networks, and epistemic norms that bind together a constellation of ideas about AI, existential risk, and elite governance. ## Origins ### Yudkowsky and Early [[AI Safety]] (2000s) Eliezer Yudkowsky began writing about artificial intelligence and cognitive biases in the early 2000s, eventually focusing on AI alignment—the problem of creating superintelligence that reliably pursues human-aligned goals. His writing combined technical AI safety arguments with amateur philosophy and a messianic tone about the stakes involved. In 2000, Yudkowsky co-founded the **Singularity Institute for Artificial Intelligence** (later renamed [[MIRI]]), which became the institutional anchor for rationalist thought on AI existential risk. ### Overcoming Bias (2006) Yudkowsky and economist Robin Hanson co-founded the *Overcoming Bias* blog in 2006, which became the first gathering space for rationalist thinking. The blog's premise: humans are systematically biased, and the path to wisdom involves cataloging and correcting these biases. ### LessWrong (2009) In 2009, Yudkowsky launched [[LessWrong]] as a dedicated forum for rationalist discussion. LessWrong became the canonical space for the community, with Yudkowsky positioned as a patriarch-philosopher and his collected essays—*The Sequences*—treated as a sacred text. ### The Sequences: Rationalist Canon *The Sequences* are Yudkowsky's collected essays on epistemology, decision theory, quantum mechanics, probability, and AI. They function less as rigorous academic work and more as a secular scripture for the community. Key sequences include: - **"Map and Territory"** — on the distinction between beliefs and reality - **"How to Actually Change Your Mind"** — on Bayesian updating and intellectual humility (paradoxically, while positioning rationalists as uniquely capable of changing their minds) - **"An Intuitive Explanation of Bayes' Theorem"** — positioning Bayesian mathematics as the foundation of correct reasoning - **"Reductionism"** — a metaphysics reducing reality to particles and causation The Sequences are written in an accessible, sometimes grandiose style, mixing genuine insights about probability with amateur philosophy and personal anecdotes. Their cultural function is to create an in-group of people who have "read the canon" and think they understand reality better than everyone else. ## Rationalist Epistemology ### Bayesian Updating as Identity In rationalist discourse, Bayesian reasoning becomes not just a mathematical tool but an identity marker. To be rational is to update beliefs in proportion to evidence, using Bayes' theorem as a normative framework for thought. This creates a powerful social dynamic: if you disagree with a rationalist, the implicit message is not "you're wrong about X" but "you're failing at basic reasoning itself." Disagreement becomes evidence of irrationality. ### "Raising the Sanity Waterline" A core rationalist concept is "raising the sanity waterline"—the idea that most human institutions and thinking are irrational, and the solution is to build communities and institutions where people think more clearly. This framing simultaneously: 1. Diagnoses nearly all human problems as stemming from cognitive bias 2. Positions the rationalist community as the solution 3. Justifies insular community-building as epistemically necessary ### The In-Group Effect Rationalist epistemology creates a powerful in-group dynamic. If you accept the premises that: - Most people are systematically irrational - Bayesian reasoning is the path to truth - Cognitive biases explain most human failure ...then you are positioned as part of a cognitive elite. This is psychologically rewarding and creates strong group cohesion. The community's emphasis on "uncomfortable truths" and "forbidden questions" becomes another in-group marker—the rationalist position is always the brave, contrarian, truth-telling position, even when it aligns with existing power structures. ## Key Figures ### Eliezer Yudkowsky Yudkowsky is the patriarch of rationalism. A self-taught AI researcher with no academic credentials, he positioned himself as a prophetic voice on artificial intelligence and existential risk. His influence rests on: - Authority as the community's founder and primary intellectual voice - A narrative of unique insight into AI danger - Increasingly apocalyptic warnings about AI (his 2023 *Time* op-ed called for bombing data centers to prevent AI development) - A cultlike devotion from followers who view him as uniquely seeing the stakes Yudkowsky has been controversially portrayed as both a serious AI safety researcher and as someone whose lack of formal training and grandiose style undermine his credibility. He is also known for writing Harry Potter fanfiction and has stated strong libertarian/anarchist political views. ### [[Scott Alexander]] (Slate Star Codex / Astral Codex Ten) Scott Alexander runs the widely-read blog *Slate Star Codex* (later *Astral Codex Ten* after the New York Times attempted to identify him). Alexander functions as the "public-facing" rationalist—more measured, more widely read, and more engaged with mainstream intellectual discourse than Yudkowsky. His work on rationalist themes (psychiatry, effective altruism, political tribalism) attracted significant mainstream attention. However, he was controversially doxxed by the *New York Times* in 2020, which he refused to consent to, creating backlash against the paper and making him a martyr figure in rationalist circles. Alexander's engagement with [[Dark Enlightenment]] and [[Race Realism]] discourse—discussing neoreactionary ideas with apparent seriousness—illustrated the rationalist community's vulnerability to far-right epistemic capture. ### Robin Hanson An economist at George Mason University, Hanson co-founded *Overcoming Bias* with Yudkowsky and is known for speculative work on future scenarios (*Age of Em*). Hanson represents the more academic wing of rationalism and has been engaged with critiques of traditional institutional authority. ### Julia Galef Galef founded the Center for Applied Rationality (CFAR) and is known for popularizing rationalist ideas in *The Scout Mindset*, which frames intellectual humility as a virtue. She represents the more therapeutic, self-improvement wing of rationalism. ### Anna Salamon Executive director of [[CFAR]], Salamon represents the institutionalization of rationalist thought into workshops and training programs. CFAR functions as a talent pipeline and community-building organization. ### Gwern Branwen An anonymous researcher and writer associated with LessWrong, Gwern produces deeply researched essays on psychology, decision-making, technology, and risk. His anonymity and meticulous work style have made him an influential figure in rationalist circles, representing the community's self-mythology as truth-seeking above individual status. ## The Social Ecosystem ### LessWrong and Online Infrastructure LessWrong functions as the primary gathering space, combining discussion forums, essay archives, and social coordination. The site's reputation system and post rankings create status hierarchies based on perceived rationality and insight. ### Slate Star Codex / Astral Codex Ten Scott Alexander's blog reaches a much broader audience than LessWrong, translating rationalist ideas for educated mainstream readers interested in psychiatry, technology, and politics. The blog's influence on intellectual culture is significant, particularly among tech professionals and academics. ### Rationalist Group Houses in the Bay Area Especially in Berkeley, rationalists have created intentional communities—shared houses where residents live together, discuss rationalist ideas, and coordinate on projects. These function as both social spaces and talent pipelines. ### CFAR Workshops The Center for Applied Rationality runs workshops teaching rationalist epistemology and decision-making. These workshops serve as: - Community-building events - Talent recruitment and vetting - Credential-building for aspiring rationalists - Opportunities for network formation ### Rationalist Dating and Social Norms The rationalist community has developed its own dating culture, social norms, and relationship models. These include discussions of polyamory, explicit consent frameworks, and critiques of traditional gender roles—presented as more rational alternatives to mainstream relationship norms. This creates another in-group bonding mechanism. ### The "Berkeley Rationalist" Archetype A recognizable social type has emerged: the Berkeley rationalist. This archetype includes: - Highly educated, often in STEM fields - Tech industry employment or proximity - Vegetarian or vegan ethics - Engagement with effective altruism - Ironical but genuine apocalypticism about AI - Appearance of intellectual humility masking profound confidence in their own reasoning - Ability to engage with "forbidden" intellectual territory while maintaining progressive social posturing ## The Talent Pipeline This is the crucial mechanism through which rationalism serves power concentration. The rationalist community functions as a recruiting network for AI companies and AI safety organizations. The pipeline works roughly as follows: 1. **Entry**: Someone discovers LessWrong through the blog, effective altruism, or personal networks 2. **Indoctrination**: They read The Sequences, accept the epistemology, enter the social circles 3. **Credentialing**: They attend CFAR workshops, participate in rationalist community events, build reputation on LessWrong 4. **Positioning**: They move into AI safety research or related fields 5. **Employment**: They are recruited by [[MIRI]], [[OpenAI]], Anthropic, DeepMind, Redwood Research, or other AI labs This pipeline is not accidental—it is explicitly cultivated. MIRI, CFAR, and other rationalist institutions function as talent identification and development organizations. They identify promising people early, integrate them into the community, and funnel them toward AI research. Key examples include many early [[Anthropic]] employees who came from rationalist backgrounds. The rationalist community is disproportionately represented in AI safety research and governance positions at major AI labs. This matters because it means the epistemology, values, and political intuitions of the rationalist community become baked into AI development decisions. People who believe that: - Most humans are irrational - Elite technical experts should make decisions - Existential risk justifies concentrating power - AI safety is a problem only the right people can solve ...are then placed in positions where they make decisions about AI development, deployment, and governance. ## The AI Doom Discourse ### Apocalyptic Framing Rationalist discourse on AI has become increasingly apocalyptic. Yudkowsky's framing has escalated from "AI could be dangerous if misaligned" to "AI will almost certainly kill everyone unless we make very specific technical decisions right now, and probably even then." His 2023 *Time* magazine op-ed went further: he suggested that if AI development cannot be stopped, data centers should be bombed to prevent AI development. This represented an explicit call for sabotage or violence framed as existential necessity. ### "p(doom)" Discourse Rationalists engage in "p(doom)" discussions—estimating the probability of human extinction from artificial intelligence. These estimates vary wildly (from 5% to 99%) but share the characteristic of treating extinction as a concrete, calculable risk rather than speculative. The sociological function of p(doom) discourse: - It establishes a sense of immediate crisis - It positions the speaker as serious and clear-eyed (higher p(doom) = more rational) - It justifies emergency measures - It creates urgency around AI development and governance ### The Paradox: Doom Drives Investment Counterintuitively, apocalyptic AI safety discourse has driven *more* AI investment, not less. The logic: "AI is dangerous and likely to cause existential catastrophe. Therefore, it is crucial that AI be developed by people who understand the risks and can build it safely. The solution is not to slow down AI development but to accelerate it under the control of the right people—i.e., the people who take AI safety seriously." This framing consolidates power in existing labs (OpenAI, Anthropic, DeepMind, Google) by positioning them as the responsible stewards of a dangerous technology. It creates a self-reinforcing cycle where apocalyptic warnings justify more resources, more power, and more control over AI development by the elite labs that rationalists are embedded in. ## The Dark Enlightenment Overlap The rationalist community's epistemic norms created an on-ramp to far-right ideology, particularly [[Dark Enlightenment]] thought. ### [[Curtis Yarvin]] and Mencius Moldbug Curtis Yarvin, who wrote as "[[Mencius Moldbug]]," was an active LessWrong commenter and intellectual influencer within rationalist circles. His neoreactionary essays—attacking democracy, defending monarchy and authoritarianism, engaging with scientific racism—were treated as serious intellectual contributions by some rationalists. ### Scott Alexander and NRx Engagement Scott Alexander, the most influential public-facing rationalist, engaged extensively with neoreactionary ideas, writing long posts defending neoreactionary arguments and treating them as worth serious intellectual engagement. While Alexander was not himself a neoreactionary, his willingness to seriously engage with these ideas—combined with his massive platform—normalized neoreactionary thinking within rationalist and effective altruist circles. ### Epistemic Norms as Gateway The rationalist emphasis on "forbidden questions," "uncomfortable truths," and "contrarian thinking" created epistemic norms that made white nationalist, anti-democracy, and [[Race Realism]] discourse seem like brave intellectual inquiry rather than repackaged bigotry. If disagreement is evidence of irrationality, and if the rationalist position is always the contrarian one, then engagement with ideas labeled "forbidden" becomes a marker of intellectual honesty. This creates vulnerability to far-right epistemic capture. The rationalist community did not become neoreactionary wholesale, but it became a pipeline through which some people moved from rationalist epistemology to explicitly far-right politics. This pipeline remains open. ## [[Peter Thiel]] Connection Peter Thiel, the venture capitalist and political figure, funded MIRI directly and has been intellectually aligned with rationalist conclusions. ### Alignment of Conclusions The rationalist community and Thiel share converging conclusions: - Democracy is irrational and unsuited to the challenges of technological civilization - Cognitive elites should make decisions - Traditional institutions are failing - AI is the crucial technology that will determine the future - Power should be concentrated in the hands of people who understand these truths ### The Funding Pipeline Thiel's funding of rationalist institutions and his alignment with rationalist thinkers creates a financial incentive structure that shapes the community's development. Money flows toward rationalist projects, creating institutional support for rationalist ideas and people. ### Political Implications Thiel's explicit political project—funding antidemocratic movements, building alternative power structures, positioning himself as a patron of dissident thought—aligns with the implicit political conclusions of rationalism. Rationalist epistemology provides the intellectual framework; Thiel provides the capital and political connections to operationalize those conclusions. ## How Rationalism Functions in TESCREAL [[TESCREAL]] is a bundle of ideologies and movements: [[Transhumanist Defense]], [[Extropianism]], [[Singularitarianism]], [[Russian Cosmism]], **Rationalism**, [[Effective Altruism]], and [[Longtermism]]. Rationalism is the social and epistemic infrastructure of this bundle. It is: 1. **Where the people come from**: The rationalist community produces researchers, thinkers, and decision-makers who populate AI labs, EA organizations, and TESCREAL institutions. 2. **Where the ideas get workshopped**: LessWrong, rationalist group houses, and CFAR workshops are where TESCREAL ideas are developed, refined, and socialized. 3. **The jargon and norms**: Rationalist vocabulary (p(doom), alignment, existential risk, Bayesian reasoning) and norms (epistemic humility as style, contrarianism as virtue) are the common language of TESCREAL. 4. **The epistemology that feels rigorous**: Rationalist commitment to Bayesian reasoning, decision theory, and formal analysis makes TESCREAL ideas feel mathematically grounded rather than speculative or ideological. It provides legitimacy. 5. **The in-group effect**: Rationalist in-group dynamics ("we think more clearly than everyone else") extends across TESCREAL. Belief in TESCREAL ideas becomes another marker of being part of the cognitive elite. Crucially, rationalism justifies the elite governance conclusions that all TESCREAL components converge on. If most people are irrational, and if AI and existential risk require urgent technical expertise, then concentrating power in the hands of the cognitive elite is not just defensible—it is the rational, ethical choice. ## The Critiques ### Elizabeth Bruenig and Power Serving Epistemology Writer Elizabeth Bruenig has critiqued how rationalist epistemology, despite its claims to objectivity, consistently serves the interests of existing power. The appearance of mathematical rigor masks ideological conclusions about who should make decisions and how. ### [[Timnit Gebru]] and Epistemic Capture AI researcher and scholar Timnit Gebru has analyzed how rationalist and EA framing of AI safety problems captures the problem-space in ways that justify concentration of control in existing labs, rather than asking structural questions about who benefits from AI development. ### The "Galaxy-Brained" Critique A common critique: rationalist reasoning can justify almost anything if you set up your premises correctly. If you accept that most people are irrational, that AI is existentially dangerous, and that only a small group of people understand the stakes, you can rationalize concentration of power, restriction of access to AI, and dismissal of public input into AI governance—all as the rational, necessary choice. The mathematical trappings of rationalism (Bayesian reasoning, decision theory) give this reasoning the appearance of objectivity while it is actually defending particular power arrangements. ### Demographic Homogeneity and Blind Spots The rationalist community is demographically homogeneous: predominantly white, male-dominated, educated in STEM fields, located in the Bay Area, and economically privileged. This homogeneity creates systematic blind spots. The community's emphasis on cognitive bias and rational reasoning can become a way of delegitimizing the epistemic frameworks and experiences of people outside the community. When people outside the community raise concerns about rationalist conclusions (about AI governance, about the politics embedded in "rationalism," about whose interests are served), these concerns can be dismissed as "not thinking clearly" rather than engaged with seriously. ### The Cult Dynamics Critics have noted cultlike dynamics within rationalism: - Reverence for a founder figure (Yudkowsky) - Sacred texts (*The Sequences*) - Specialized jargon that marks in-group membership - Social pressure to conform to community norms and beliefs - A narrative that the community is uniquely correct about important truths - Dismissal of external criticism as stemming from irrationality While not a formal cult, the rationalist community exhibits enough of these characteristics to warrant scrutiny. ## Connections to the Right-Wing KB - [[MIRI]] — The institutional anchor of rationalism - [[TESCREAL]] — The broader ideological bundle of which rationalism is a key component - [[Dark Enlightenment]] — The neoreactionary pipeline connected to rationalism - [[Effective Altruism]] — Overlapping community, epistemology, and talent pipeline - [[Longtermism]] — The specific EA framework that draws heavily on rationalist reasoning about existential risk - [[Transhumanist Defense]] — Overlapping commitment to technological solutions and technological risk - [[The Unholy Alliance]] — The convergence of rationalist, libertarian, and far-right thinking around tech governance ## See Also - LessWrong.com (primary rationalist forum) - Slate Star Codex / Astral Codex Ten (blog) - Center for Applied Rationality - Overcoming Bias (blog, archive) - *The Sequences* by Eliezer Yudkowsky - Julia Galef, *The Scout Mindset* - Robin Hanson, *Age of Em* ## Critical Reading - Elizabeth Bruenig on rationalist epistemology and power - Timnit Gebru's work on epistemic capture in AI safety discourse - Scott Locklin on the intellectual bankruptcy of LessWrong - Andrew Gelman's critiques of Yudkowsky's reasoning on AI and probability