What Computers Still Can't Do

Knowledge Graph: What Computers Still Can't Do (Hubert Dreyfus, 1992)

Editorial spotlight: ↑ the body's know-how cannot be reduced to know-that

Concepts

• Heidegger's ready-to-hand (Zuhandenheit) (importance 5): Equipment that withdraws from conscious awareness during skillful use. A hammer becomes transparent to the skilled carpenter — the focus is on the nail, not the tool.. Source: (from training memory of book).
• Merleau-Ponty's body schema (importance 5): Pre-reflective bodily awareness that orients us in space and enables skillful action without explicit calculation. The body 'knows' how to move before thought intervenes.. Source: (from training memory of book).
• Dreyfus's situatedness (importance 5): Human intelligence is always already embedded in concrete situations that provide meaningful context. You cannot separate cognition from its worldly embeddedness.. Source: (from training memory of book).
• Dreyfus's commonsense knowledge problem (importance 4): The vast background of everyday understanding humans possess cannot be made explicit or formalized. We know how to navigate the world without encoding millions of facts.. Source: (from training memory of book).
• Dreyfus Stage 5: Expertise (importance 4): Intuitive, non-deliberate, fluid performance. The expert perceives what needs to be done and does it without conscious rule-following or calculation.. Source: (from training memory of book).
• Heidegger's Dasein (being-in-the-world) (importance 4): Human existence is fundamentally being-in-the-world, not a detached mind accessing an external world. Cognition is grounded in practical involvement.. Source: (from training memory of book).
• Dreyfus's background practices (importance 4): Shared bodily skills and cultural know-how that make meaning possible. Cannot be made fully explicit or programmed — the enabling condition for all cognition.. Source: (from training memory of book).
• Merleau-Ponty's motor intentionality (importance 4): The body is directed toward tasks pre-reflectively. Grasping a cup doesn't require calculating angles — the body 'intends' the action directly.. Source: (from training memory of book).
• Dreyfus Stage 1: Novice (importance 3): Follows context-free rules and recognizes objective features. This is the only stage that matches symbolic AI's operation.. Source: (from training memory of book).
• Dreyfus Stage 3: Competence (importance 3): Begins choosing goals and perspectives. Hierarchical planning emerges, but still largely deliberate and rule-guided.. Source: (from training memory of book).
• Dreyfus Stage 4: Proficiency (importance 3): Situations are grasped holistically from experience. Intuitive pattern recognition emerges, though response may still be deliberate.. Source: (from training memory of book).
• Dreyfus's context-free elements (importance 3): Features that can be recognized without understanding the overall situation. GOFAI can only work with these, but expertise transcends them.. Source: (from training memory of book).
• Dreyfus's semantic primitives critique (importance 3): The idea that meaning can be built from basic atomic concepts. Dreyfus argues meaning is holistic and context-dependent, resisting decomposition.. Source: (from training memory of book).
• Heidegger's present-at-hand (Vorhandenheit) (importance 3): The detached, theoretical stance where things become explicit objects of contemplation. This is the breakdown mode, not primary engagement.. Source: (from training memory of book).
• Dreyfus's breakdown phenomenon (importance 3): When skillful coping fails, we shift from ready-to-hand transparent use to present-at-hand explicit reflection. GOFAI models only the breakdown mode.. Source: (from training memory of book).
• Dreyfus's critique of atomism (importance 3): The view that complex phenomena can be built from simple independent elements. Meaning and skill resist this — they are context-dependent wholes.. Source: (from training memory of book).
• Dreyfus's pattern-completion mechanism (importance 3): Neural nets generalize through similarity, not symbolic inference. This is closer to human cognition but still disembodied in 1992 implementations.. Source: (from training memory of book).
• knowledge engineering bottleneck (importance 3): The difficulty of extracting expert knowledge into rules. Dreyfus argues this fails because expertise is not rule-based at stage 5.. Source: (from training memory of book).
• intentionality (aboutness) (importance 3): Mental states are about things in the world. Dreyfus argues this emerges from embodied coping, not symbol manipulation.. Source: (from training memory of book).
• Wittgenstein's forms of life (importance 3): Shared practices and ways of going on that ground language and rule-following. Dreyfus sees this as allied with phenomenology's background.. Source: (from training memory of book).
• Dreyfus's transparency in skillful action (importance 3): In expert performance, both tools and body disappear from awareness. Attention is on the task, not the means. GOFAI makes everything explicit, blocking this.. Source: (from training memory of book).
• Merleau-Ponty's maximal grip (importance 3): The body tends toward optimal orientation without calculation — e.g., adjusting distance to see a painting clearly. Pre-reflective bodily intelligence.. Source: (from training memory of book).
• Merleau-Ponty's habit body (importance 3): Acquired skills sediment into the body, expanding its repertoire. Driving becomes second nature; the body acquires new 'organs'.. Source: (from training memory of book).
• Merleau-Ponty's body as zero-point (importance 3): The body is the origin of spatial perspective — here vs. there. Not an object in space but the condition for a spatial world.. Source: (from training memory of book).
• Dreyfus's perception-action loop (importance 3): Perception and action are coupled, not sequential. Perceiving is exploring, acting is guided by what shows up. Disembodied AI lacks this coupling.. Source: (from training memory of book).
• Harnad's symbol grounding problem (1990) (importance 3): How do symbols acquire meaning? Dreyfus anticipates this: symbols must be grounded in embodied interaction, not just defined by other symbols.. Source: (from training memory of book).
• Moravec's Paradox (1988) (importance 3): Hard problems for humans (chess, logic) are easy for AI; easy problems for humans (perception, motor control) are hard for AI. Supports Dreyfus's embodiment thesis.. Source: (from training memory of book).
• Dreyfus's commonsense reasoning critique (importance 3): Everyday reasoning uses vast background knowledge and context-sensitivity. Cannot be captured in first-order logic or production rules.. Source: (from training memory of book).
• McCarthy's qualification problem (importance 3): Cannot list all conditions under which a rule applies — e.g., 'turning the key starts the car' unless battery dead, out of gas, etc. Dreyfus sees this as fatal to rule-based AI.. Source: (from training memory of book).
• Dreyfus's intelligent behavior definition (importance 3): Context-sensitive, flexible, skillful coping that draws on vast background understanding. Not reducible to discrete symbol manipulation.. Source: (from training memory of book).
• situated cognition movement (1980s-90s) (importance 3): Cognitive science emphasizing context, embodiment, and action. Dreyfus influenced and is cited by this movement.. Source: (from training memory of book).
• Dreyfus Stage 2: Advanced Beginner (importance 2): Recognizes situational aspects from prior experience. Rules become more nuanced but still explicit.. Source: (from training memory of book).
• backpropagation learning (1986) (importance 2): Training method for multilayer networks. Dreyfus acknowledges this as progress toward continuous, similarity-based learning vs. symbolic rules.. Source: (from training memory of book).
• Wittgenstein's family resemblance (importance 2): Concepts are defined by overlapping similarities, not necessary/sufficient conditions. Challenges GOFAI's definitional approach.. Source: (from training memory of book).
• Gibson's ecological psychology (importance 2): Perception is direct pickup of information, not inference from sense data. Dreyfus sees this as alternative to computational theories.. Source: (from training memory of book).
• Brooks's fast, cheap, and out of control (importance 2): Slogan for behavior-based robotics. Dreyfus endorses this as moving toward embodied AI, though still far from human-level skill.. Source: (from training memory of book).
• defeasible (default) reasoning (importance 2): Reasoning with rules that have exceptions. AI approaches use non-monotonic logics; Dreyfus argues the exceptions are infinite and context-dependent.. Source: (from training memory of book).
• distributed cognition (Hutchins 1995) (importance 2): Cognition extends beyond individual mind into tools and social practices. Allied with Dreyfus's anti-individualism.. Source: (from training memory of book).
• Clark & Chalmers's extended mind (1998) (importance 2): Mind extends into notebook, smartphone, etc. Post-1992 but builds on embodied cognition Dreyfus championed.. Source: (from training memory of book).
• Varela's enactive cognition (1991) (importance 2): Cognition is enacting a world through sensorimotor coupling. Close to Dreyfus's phenomenology-informed view.. Source: (from training memory of book).

Claims

• Dreyfus's embodied skill argument (importance 5): Human expertise depends on bodily know-how that cannot be captured in explicit rules or representations. The body's pre-reflective coping with situations grounds all higher cognition.. Source: (from training memory of book).
• GOFAI's formalizability assumption (importance 5): Good Old-Fashioned AI assumes all intelligent behavior can be captured as symbol manipulation following explicit rules. Dreyfus targets this as fundamentally mistaken.. Source: (from training memory of book).
• Dreyfus's frame problem critique (importance 4): AI systems cannot determine what is relevant in a situation without infinite regress. Humans solve this through embodied familiarity, not formal rules about relevance.. Source: (from training memory of book).
• Dreyfus's representation assumption critique (importance 4): GOFAI assumes intelligent behavior requires internal models of the world. Dreyfus argues skillful coping often proceeds without representation — direct responsiveness.. Source: (from training memory of book).
• Dreyfus's four GOFAI assumptions (importance 4): Biological, psychological, epistemological, and ontological assumptions underlying symbolic AI. Each is philosophically suspect and empirically challenged.. Source: (from training memory of book).
• Dreyfus's epistemological assumption critique (importance 4): All knowledge can be formalized. Dreyfus argues know-how (skills, background understanding) resists formalization — it's not degraded knowledge but a different kind.. Source: (from training memory of book).
• Dreyfus's regress-of-rules argument (importance 4): If all skill is rule-following, you need rules for applying rules, and rules for those rules — infinite regress. Must ground in non-rule-based coping.. Source: (from training memory of book).
• Dreyfus's 1990s embodied AI turn (importance 4): By 1992 second edition, Dreyfus sees new hope in connectionism and behavior-based robotics — but insists full situatedness and background remain missing.. Source: (from training memory of book).
• Dreyfus's 'what is' assumption critique (importance 3): GOFAI assumes intelligence requires decomposing the world into atomic facts (what-is). Dreyfus argues skilled action works with holistic, meaning-laden situations.. Source: (from training memory of book).
• Dreyfus's biological assumption critique (importance 3): The brain processes information like a digital computer. Dreyfus argues neural processing may be analog, holistic, and fundamentally non-discrete.. Source: (from training memory of book).
• Dreyfus's psychological assumption critique (importance 3): The mind works by manipulating mental symbols. Dreyfus argues much cognition is sub-symbolic, embodied, and non-representational.. Source: (from training memory of book).
• Dreyfus's ontological assumption critique (importance 3): Reality consists of atomic facts. Dreyfus argues the world as experienced is meaningful, context-dependent wholes, not fact-assemblies.. Source: (from training memory of book).
• Dreyfus's critique of AI planning (importance 3): Classical AI planning assumes complete world models and means-ends reasoning. Real-world action is opportunistic, context-sensitive, not exhaustive search.. Source: (from training memory of book).
• Dreyfus's alchemy-to-chemistry analogy (importance 3): GOFAI is like alchemy — not a science on the wrong track but a misconceived enterprise. True cognitive science requires rethinking foundations, not incremental progress.. Source: (from training memory of book).
• Dreyfus's digital computer limits (importance 3): Digital computers may be inherently unsuited for human-like intelligence due to discrete, context-free symbol processing. Analog, embodied systems may be necessary.. Source: (from training memory of book).
• Dreyfus's 1992 future AI prediction (importance 3): AI will succeed only when it becomes embodied, situated, and grounded in skillful coping. Pure symbol manipulation will never achieve human-level intelligence.. Source: (from training memory of book).
• Dreyfus's 1992 partial vindication claim (importance 3): Second edition argues GOFAI's failures vindicate his 1972 critique. But new approaches (connectionism, behavior-based robotics) offer hope if pursued with phenomenological insight.. Source: (from training memory of book).
• Dreyfus's early learning critique (1972) (importance 2): In first edition, skeptical of machine learning as too slow and brittle. By 1992, more optimistic about neural network learning if embodied.. Source: (from training memory of book).
• Dreyfus's continental-analytic bridge (importance 2): Uses continental phenomenology (Heidegger, Merleau-Ponty) to critique analytic AI. Unusual cross-tradition work in 1970s-90s philosophy.. Source: (from training memory of book).

Empirical results

• Dreyfus's microworld success concession (importance 3): AI succeeds in toy domains with fixed rules and closed worlds (chess, blocks world). But this success doesn't scale to open-ended real-world situations.. Source: (from training memory of book).

Methods

• Dreyfus's five-stage skill acquisition model (importance 4): Learning progresses from novice (rule-following) through competence and proficiency to expert (intuitive, context-sensitive performance). Only the first stage fits GOFAI.. Source: (from training memory of book).

Entities

• Newell & Simon's Physical Symbol System Hypothesis (importance 4): The claim that a physical symbol system has the necessary and sufficient means for general intelligent action. Dreyfus's primary theoretical opponent.. Source: (from training memory of book).
• Winograd's SHRDLU (1971) (importance 3): Natural language system for manipulating blocks. Dreyfus cites it as paradigmatic microworld success that doesn't generalize.. Source: (from training memory of book).
• Schank's scripts (1977) (importance 3): Stereotyped event sequences (restaurant script, etc.) meant to encode commonsense knowledge. Dreyfus critiques this as brittle and infinite-regress-prone.. Source: (from training memory of book).
• Lenat's CYC project (1984-) (importance 3): Attempt to encode all human commonsense knowledge in logical form. Dreyfus predicts it will fail because background knowledge is not a finite set of facts.. Source: (from training memory of book).
• connectionist networks (1980s-90s) (importance 3): Pattern-matching systems using distributed representations. Dreyfus sees these as more promising than GOFAI but still lacking embodiment and situatedness.. Source: (from training memory of book).
• Rumelhart & McClelland PDP models (1986) (importance 3): Parallel Distributed Processing frameworks. Dreyfus is cautiously optimistic but insists they still lack grounding in bodily action.. Source: (from training memory of book).
• expert systems (1970s-80s) (importance 3): Rule-based AI for specialized domains (MYCIN, DENDRAL). Dreyfus critiques these as brittle stage-1 novice systems that never reach expertise.. Source: (from training memory of book).
• Searle's Chinese Room (1980) (importance 3): Thought experiment showing syntax doesn't constitute semantics. Dreyfus allies with Searle against strong AI but emphasizes embodiment over intentionality.. Source: (from training memory of book).
• Plato-Descartes rationalist tradition (importance 3): The philosophical lineage that views knowledge as explicit, detachable from the body, and formalizable. GOFAI inherits this; Dreyfus opposes it.. Source: (from training memory of book).
• Husserl's phenomenology (importance 3): Study of conscious experience from the first-person perspective. Heidegger and Merleau-Ponty radicalize this toward embodiment; Dreyfus follows them.. Source: (from training memory of book).
• Wittgenstein on rule-following (importance 3): Rules require background practices to be applied — infinite regress threatens unless grounded in forms of life. Dreyfus uses this against GOFAI's rule obsession.. Source: (from training memory of book).
• Polanyi's tacit knowledge (1966) (importance 3): We know more than we can tell. Dreyfus cites this as evidence that expertise resists complete articulation.. Source: (from training memory of book).
• Ryle's knowing-how vs. knowing-that (1949) (importance 3): Skill knowledge (knowing how) is not reducible to propositional knowledge (knowing that). Dreyfus uses this against the idea that all intelligence is symbol manipulation.. Source: (from training memory of book).
• Gibson's affordances (1979) (importance 3): Environment offers action possibilities directly perceived — chair affords sitting. Dreyfus aligns this with phenomenology's ready-to-hand.. Source: (from training memory of book).
• Brooks's subsumption architecture (1986) (importance 3): Layered reactive control without central planning. Dreyfus sees this as vindication — embodied, situated robots succeed where GOFAI failed.. Source: (from training memory of book).
• Heidegger's Being and Time (1927) (importance 3): Foundational phenomenology text. Dreyfus uses its analysis of ready-to-hand, presence-at-hand, and being-in-the-world throughout.. Source: (from training memory of book).
• Merleau-Ponty's Phenomenology of Perception (1945) (importance 3): Central source for Dreyfus's embodiment arguments. Body schema, motor intentionality, and pre-reflective coping.. Source: (from training memory of book).
• blocks world domain (importance 2): Simplified environment with geometric blocks on a table. Classic AI testbed that Dreyfus argues is fatally toy-like.. Source: (from training memory of book).
• chess-playing programs (pre-1992) (importance 2): Brute-force search systems that Dreyfus acknowledges as successful — but in a domain with complete information and fixed rules, unlike everyday life.. Source: (from training memory of book).
• Schank's restaurant script (importance 2): Canonical example: enter → sit → order → eat → pay → leave. Dreyfus asks: what handles variations? Script for script variations? Infinite regress.. Source: (from training memory of book).
• Hopfield networks (1982) (importance 2): Recurrent neural networks for associative memory. Dreyfus notes their holistic pattern-completion but questions whether they capture skill.. Source: (from training memory of book).
• MYCIN (1972) (importance 2): Medical diagnosis expert system using ~600 rules. Dreyfus notes it performs at competent level but cannot match expert physicians' intuition.. Source: (from training memory of book).
• Weizenbaum's ELIZA (1966) (importance 2): Pattern-matching chatbot. Dreyfus and Weizenbaum both critique the illusion of understanding it creates.. Source: (from training memory of book).
• Turing Test (1950) (importance 2): Behavioral criterion for machine intelligence. Dreyfus argues passing it requires embodiment and situatedness, not just symbol manipulation.. Source: (from training memory of book).
• Cartesian mind-body dualism (importance 2): Mind as separate from body, a thinking substance. Dreyfus argues this mistake enables GOFAI's disembodied approach.. Source: (from training memory of book).
• STRIPS planner (1971) (importance 2): Classic AI planning system using preconditions and effects. Dreyfus notes it works in toy domains but fails to scale.. Source: (from training memory of book).
• Shakey the Robot (1966-72) (importance 2): Mobile robot using STRIPS planning. Dreyfus cites it as embodied but still symbol-manipulating, not genuinely skillful.. Source: (from training memory of book).
• Brooks's Genghis robot (1989) (importance 2): Six-legged insect-like robot using subsumption. Walks robustly without planning or world model — pure reactive coupling.. Source: (from training memory of book).
• What Computers Can't Do (1972 first edition) (importance 2): Original critique during GOFAI's height. Widely dismissed by AI community as philosophically confused.. Source: (from training memory of book).
• AI winter (1974-1980, 1987-1993) (importance 2): Periods of reduced funding and interest after unmet promises. Dreyfus's critiques gain credibility during these downturns.. Source: (from training memory of book).
• Lighthill Report (1973) (importance 2): UK government report criticizing AI progress. Dreyfus cited as influence; leads to UK AI funding cuts.. Source: (from training memory of book).
• Minsky & Papert's Perceptrons (1969) (importance 2): Showed limitations of single-layer networks. Contributed to first AI winter. Dreyfus notes irony: critics of neural nets now face critiques of symbolic AI.. Source: (from training memory of book).
• Wittgenstein's Philosophical Investigations (1953) (importance 2): Source of rule-following arguments and forms of life. Dreyfus uses against idea that meaning is formal definition.. Source: (from training memory of book).
• Varela's The Embodied Mind (1991) (importance 2): Buddhist-phenomenology synthesis arguing for embodied, enactive cognition. Dreyfus sees this as ally.. Source: (from training memory of book).
• Hubert Dreyfus (1929-2017) (importance 1): UC Berkeley philosophy professor, Heidegger scholar, phenomenologist. Became AI's most prominent philosophical critic.. Source: (from training memory of book).

Relations

• Dreyfus's embodied skill argument builds-on Heidegger's ready-to-hand (Zuhandenheit)
• Dreyfus's embodied skill argument builds-on Merleau-Ponty's body schema
• Dreyfus's embodied skill argument requires Dreyfus's situatedness
• Dreyfus's embodied skill argument contradicts GOFAI's formalizability assumption
• GOFAI's formalizability assumption exemplifies Newell & Simon's Physical Symbol System Hypothesis
• Dreyfus's frame problem critique refutes GOFAI's formalizability assumption
• Dreyfus's frame problem critique requires Dreyfus's situatedness
• Dreyfus's commonsense knowledge problem refutes GOFAI's formalizability assumption
• Dreyfus's commonsense knowledge problem requires Dreyfus's background practices
• Dreyfus's five-stage skill acquisition model motivates Dreyfus Stage 1: Novice
• Dreyfus's five-stage skill acquisition model motivates Dreyfus Stage 2: Advanced Beginner
• Dreyfus's five-stage skill acquisition model motivates Dreyfus Stage 3: Competence
• Dreyfus's five-stage skill acquisition model motivates Dreyfus Stage 4: Proficiency
• Dreyfus's five-stage skill acquisition model motivates Dreyfus Stage 5: Expertise
• Dreyfus Stage 1: Novice requires Dreyfus's context-free elements
• Dreyfus Stage 5: Expertise exemplifies Dreyfus's embodied skill argument
• Dreyfus Stage 5: Expertise contradicts GOFAI's formalizability assumption
• GOFAI's formalizability assumption exemplifies Dreyfus Stage 1: Novice
• Dreyfus's microworld success concession exemplifies Winograd's SHRDLU (1971)
• Dreyfus's microworld success concession exemplifies blocks world domain
• Dreyfus's microworld success concession exemplifies chess-playing programs (pre-1992)
• Dreyfus's microworld success concession contradicts Dreyfus's situatedness
• Dreyfus's semantic primitives critique supports GOFAI's formalizability assumption
• Dreyfus's embodied skill argument contradicts Dreyfus's semantic primitives critique
• Schank's scripts (1977) exemplifies Dreyfus's semantic primitives critique
• Schank's restaurant script exemplifies Schank's scripts (1977)
• Lenat's CYC project (1984-) contradicts Dreyfus's commonsense knowledge problem
• Dreyfus's embodied skill argument refutes Lenat's CYC project (1984-)
• Dreyfus's representation assumption critique supports GOFAI's formalizability assumption
• Heidegger's ready-to-hand (Zuhandenheit) contradicts Dreyfus's representation assumption critique
• Heidegger's present-at-hand (Vorhandenheit) exemplifies Dreyfus's breakdown phenomenon
• Dreyfus's breakdown phenomenon contradicts Heidegger's ready-to-hand (Zuhandenheit)
• Heidegger's Dasein (being-in-the-world) supports Dreyfus's situatedness
• Dreyfus's 'what is' assumption critique supports GOFAI's formalizability assumption
• Dreyfus's critique of atomism supports Dreyfus's 'what is' assumption critique
• Dreyfus's embodied skill argument contradicts Dreyfus's critique of atomism
• Dreyfus's four GOFAI assumptions motivates Dreyfus's biological assumption critique
• Dreyfus's four GOFAI assumptions motivates Dreyfus's psychological assumption critique
• Dreyfus's four GOFAI assumptions motivates Dreyfus's epistemological assumption critique
• Dreyfus's four GOFAI assumptions motivates Dreyfus's ontological assumption critique
• Dreyfus's epistemological assumption critique supports GOFAI's formalizability assumption
• Dreyfus's embodied skill argument contradicts Dreyfus's epistemological assumption critique
• connectionist networks (1980s-90s) contradicts GOFAI's formalizability assumption
• Hopfield networks (1982) exemplifies connectionist networks (1980s-90s)
• Rumelhart & McClelland PDP models (1986) exemplifies connectionist networks (1980s-90s)
• backpropagation learning (1986) enables Rumelhart & McClelland PDP models (1986)
• Dreyfus's pattern-completion mechanism exemplifies connectionist networks (1980s-90s)
• Dreyfus's pattern-completion mechanism supports Dreyfus's embodied skill argument
• expert systems (1970s-80s) exemplifies GOFAI's formalizability assumption
• MYCIN (1972) exemplifies expert systems (1970s-80s)
• knowledge engineering bottleneck requires expert systems (1970s-80s)
• Dreyfus Stage 5: Expertise contradicts knowledge engineering bottleneck
• Weizenbaum's ELIZA (1966) exemplifies GOFAI's formalizability assumption
• Turing Test (1950) motivates GOFAI's formalizability assumption
• Searle's Chinese Room (1980) refutes GOFAI's formalizability assumption
• intentionality (aboutness) requires Dreyfus's embodied skill argument
• Dreyfus's background practices enables Dreyfus's situatedness
• Dreyfus's background practices supports Dreyfus's commonsense knowledge problem
• Plato-Descartes rationalist tradition motivates GOFAI's formalizability assumption
• Cartesian mind-body dualism exemplifies Plato-Descartes rationalist tradition
• Dreyfus's embodied skill argument contradicts Cartesian mind-body dualism
• Husserl's phenomenology motivates Heidegger's Dasein (being-in-the-world)
• Wittgenstein on rule-following supports Dreyfus's regress-of-rules argument
• Wittgenstein's forms of life supports Dreyfus's background practices
• Polanyi's tacit knowledge (1966) supports Dreyfus's embodied skill argument
• Wittgenstein's family resemblance contradicts Dreyfus's semantic primitives critique
• Merleau-Ponty's motor intentionality supports Dreyfus's embodied skill argument
• Dreyfus's transparency in skillful action exemplifies Heidegger's ready-to-hand (Zuhandenheit)
• Merleau-Ponty's maximal grip exemplifies Merleau-Ponty's motor intentionality
• Merleau-Ponty's habit body exemplifies Merleau-Ponty's body schema
• Ryle's knowing-how vs. knowing-that (1949) supports Dreyfus's embodied skill argument
• Dreyfus's regress-of-rules argument refutes GOFAI's formalizability assumption
• Merleau-Ponty's body as zero-point supports Merleau-Ponty's body schema
• Dreyfus's perception-action loop supports Dreyfus's embodied skill argument
• Gibson's affordances (1979) supports Heidegger's ready-to-hand (Zuhandenheit)
• Gibson's ecological psychology contradicts Dreyfus's representation assumption critique
• Dreyfus's critique of AI planning refutes GOFAI's formalizability assumption
• STRIPS planner (1971) exemplifies Dreyfus's critique of AI planning
• Shakey the Robot (1966-72) exemplifies STRIPS planner (1971)
• Harnad's symbol grounding problem (1990) supports Dreyfus's embodied skill argument
• Moravec's Paradox (1988) supports Dreyfus's embodied skill argument
• Brooks's subsumption architecture (1986) contradicts GOFAI's formalizability assumption
• Brooks's subsumption architecture (1986) supports Dreyfus's embodied skill argument
• Brooks's Genghis robot (1989) exemplifies Brooks's subsumption architecture (1986)
• Brooks's fast, cheap, and out of control exemplifies Brooks's subsumption architecture (1986)
• Dreyfus's 1990s embodied AI turn evidences Brooks's subsumption architecture (1986)
• Dreyfus's 1990s embodied AI turn evidences connectionist networks (1980s-90s)
• Dreyfus's alchemy-to-chemistry analogy refutes GOFAI's formalizability assumption
• What Computers Can't Do (1972 first edition) motivates Dreyfus's embodied skill argument
• AI winter (1974-1980, 1987-1993) supports Dreyfus's embodied skill argument
• Lighthill Report (1973) builds-on What Computers Can't Do (1972 first edition)
• Dreyfus's commonsense reasoning critique exemplifies Dreyfus's commonsense knowledge problem
• McCarthy's qualification problem supports Dreyfus's frame problem critique
• defeasible (default) reasoning exemplifies McCarthy's qualification problem
• Dreyfus's early learning critique (1972) refutes GOFAI's formalizability assumption
• Minsky & Papert's Perceptrons (1969) contradicts connectionist networks (1980s-90s)
• Dreyfus's intelligent behavior definition requires Dreyfus's embodied skill argument
• Dreyfus's digital computer limits refutes GOFAI's formalizability assumption
• Dreyfus's 1992 future AI prediction motivates Dreyfus's 1990s embodied AI turn
• Heidegger's Being and Time (1927) motivates Heidegger's Dasein (being-in-the-world)
• Heidegger's Being and Time (1927) motivates Heidegger's ready-to-hand (Zuhandenheit)
• Merleau-Ponty's Phenomenology of Perception (1945) motivates Merleau-Ponty's body schema
• Merleau-Ponty's Phenomenology of Perception (1945) motivates Merleau-Ponty's motor intentionality
• Wittgenstein's Philosophical Investigations (1953) motivates Wittgenstein on rule-following
• Wittgenstein's Philosophical Investigations (1953) motivates Wittgenstein's family resemblance
• Dreyfus's continental-analytic bridge enables Dreyfus's embodied skill argument
• situated cognition movement (1980s-90s) builds-on Dreyfus's situatedness
• distributed cognition (Hutchins 1995) supports Dreyfus's background practices
• Clark & Chalmers's extended mind (1998) builds-on Dreyfus's embodied skill argument
• Varela's enactive cognition (1991) supports Dreyfus's perception-action loop
• Varela's The Embodied Mind (1991) exemplifies Varela's enactive cognition (1991)
• Dreyfus's 1992 partial vindication claim refutes GOFAI's formalizability assumption
• Dreyfus's 1992 partial vindication claim supports Dreyfus's 1990s embodied AI turn
• Heidegger's ready-to-hand (Zuhandenheit) requires Heidegger's Dasein (being-in-the-world)
• Merleau-Ponty's body schema requires Dreyfus's situatedness
• Dreyfus Stage 2: Advanced Beginner builds-on Dreyfus Stage 1: Novice
• Dreyfus Stage 3: Competence builds-on Dreyfus Stage 2: Advanced Beginner
• Dreyfus Stage 4: Proficiency builds-on Dreyfus Stage 3: Competence
• Dreyfus Stage 5: Expertise builds-on Dreyfus Stage 4: Proficiency
• Dreyfus's situatedness requires Dreyfus's background practices
• Newell & Simon's Physical Symbol System Hypothesis requires Dreyfus's representation assumption critique
• Winograd's SHRDLU (1971) requires blocks world domain
• Schank's scripts (1977) contradicts Dreyfus's commonsense knowledge problem
• Lenat's CYC project (1984-) contradicts Dreyfus's background practices
• expert systems (1970s-80s) contradicts Dreyfus's five-stage skill acquisition model
• connectionist networks (1980s-90s) contradicts Dreyfus's representation assumption critique
• Brooks's subsumption architecture (1986) contradicts Dreyfus's representation assumption critique
• Brooks's subsumption architecture (1986) supports Dreyfus's critique of AI planning
• Brooks's Genghis robot (1989) contradicts Shakey the Robot (1966-72)
• Hubert Dreyfus (1929-2017) motivates Dreyfus's embodied skill argument
• Dreyfus's continental-analytic bridge requires Heidegger's Being and Time (1927)
• Dreyfus's continental-analytic bridge requires Merleau-Ponty's Phenomenology of Perception (1945)
• Dreyfus's continental-analytic bridge requires Wittgenstein's Philosophical Investigations (1953)
• situated cognition movement (1980s-90s) supports Dreyfus's 1990s embodied AI turn
• Varela's enactive cognition (1991) supports Dreyfus's 1990s embodied AI turn

© 2026 LOOMUS. Curation, importance scoring, edge typing, and visualization. Source paper © respective authors, quoted under fair use.