Hubert Dreyfus's central thesis in "What Computers Can't Do" is that the symbolic, rule-based approach to artificial intelligence, dominant in his time, cannot replicate human intelligence because human understanding and action rely on implicit, embodied, and context-dependent knowledge, not explicit, codified rules. He argues that the attempt to create intelligent machines by formalizing knowledge into symbols and algorithms is fundamentally flawed.
Dreyfus distinguishes between calculable, rule-governed tasks and intelligent, human-like activities. He asserts that human skills like recognizing a face, grasping a joke, or navigating a complex environment are based on a non-symbolic "know-how" acquired through embodied experience and a deep, intuitive grasp of context. Readers learn that superficial simulations of intelligence, while capable of specific tasks, will not achieve the flexible, general intelligence characteristic of humans.
Full text isn't indexed yet — this overview draws on general knowledge of the book and its metadata, and chat works the same way.
Key concepts
- Philosophical behaviorism — The idea that mental states are defined by behavioral dispositions, which Dreyfus criticizes as insufficient for explaining human intelligence.
- Tacit knowledge — Knowledge that is difficult to articulate or formalize, often acquired through practice and experience.
- Embodied cognition — The theory that the mind is shaped by the body's physical interaction with the world.
- Formalism — The AI approach that treats intelligence as the manipulation of symbols according to explicit rules.
- Phenomenological understanding — Knowledge derived from direct experience and conscious awareness of the world.