Introduction
You likely view your mind as a general-purpose engine capable of mastering any task, from poetry to quantum physics. This intuition treats the brain as an unprogrammed computer or a blank slate shaped entirely by culture. Steven Pinker argues in "How the Mind Works" that this common view is profoundly mistaken. The human mind is instead a system of organs of computation designed by natural selection to solve ancestral survival problems.
Specialized mental modules make up your cognitive architecture, each engineered for specific domains like face recognition, language acquisition, or mate selection. You do not possess a single all-purpose intellect, but rather a Swiss army knife of functional adaptations.
- How did natural selection build such intricate mental machinery?
- What makes human intelligence distinct from animal cognition?
- Why do emotions and consciousness serve evolutionary functions?
1. Standard Equipment
Reverse-Engineering the Psyche
Psychology and biology can be understood as reverse-engineering complex organic systems produced by natural selection.
You can analyze an engineered machine by figuring out what problems it was designed to solve. The human mind and biological organs must be reverse-engineered in the exact same way to understand their adaptive functions. Steven Pinker builds on the works of George Williams, Richard Dawkins, and Daniel Dennett on reverse-engineering biology and psychology.
Psychology must use the logic of reverse-engineering, pioneered by Charles Darwin, to determine the specific adaptive problems that our mental faculties were designed to solve. Evolutionary psychology combines the cognitive revolution with evolutionary biology.
Biologist Margie Profet used reverse-engineering to show that pregnancy sickness is an adaptation that protects the developing embryo from plant toxins.
The mind is composed of many specialized mental modules, each uniquely tailored by natural selection to solve specific problems. Just as the body has specialized organs like the heart and lungs, the mind relies on distinct computational modules equipped with built-in assumptions for specific domains. This view opposes the Standard Social Science Model.
The Computational Theory of Mind
Building a functioning robot that can see, move, and navigate the physical world is vastly more challenging than space exploration. This endeavor exposes the hidden computational complexity of everyday human perception.
An electronic camera or eye merely records brightness values in an array of tiny patches. This presents an ill-posed problem where edge detection, lighting variations, and depth must be deciphered without homunculi inside the head. The brain solves this by relying on built-in assumptions about the physical world, such as even illumination.
Pawan Sinha and Edward Adelson devised a drawing of a ring of light and dark gray tiles that actually represents a rectangular cutout in a black cover revealing different background objects.
The mind is not the physical brain itself, but rather what the brain does through information processing and computation. This process bridges the gap between physical matter and mental states. Beliefs and desires are represented as physical states of matter acting as symbols that interact through logical rules, ultimately triggering muscle movements. This computational theory of mind explains how immaterial intentions can cause physical actions without violating physical laws.
2. Thinking Machines
Intelligence is a form of information processing where symbols simultaneously carry information and cause physical events to happen.
Natural Computation
Information is a correlation produced by a lawful process. A computer or information processor uses symbols that stand for states of the world and trigger physical operations, ensuring that true premises lead to true conclusions without requiring a soul or special energy. A machine that scans tree rings on a stump and burns corresponding marks onto a wooden plank acts as a rational machine producing true conclusions.
Thinking does not occur in spoken languages like English, but in a rich, unambiguous internal language of thought called mentalese. Spoken languages are ambiguous, brief, and cluttered with grammatical boilerplate unsuited for internal thought. Production systems use memory bulletin boards and knee-jerk reflexes called demons to manipulate mentalese symbols and deduce new facts. A production system uses Long-Term Memory, Short-Term Memory, and a Goal column to deduce that Edgar is the uncle of Me.
The human brain utilizes at least four major formats of representation to process information efficiently. These formats include:
- visual images in a picture-like mosaic;
- phonological representations in a sound-based tape loop;
- grammatical representations structured as hierarchical trees;
- mentalese as the conceptual language of thought.
Michael Posner's experiments show that matching physically identical letters like A-A is faster than matching physically different but alphabetically identical letters like A-a, which require conversion into an alphabetic format.
Neural networks use distributed patterns of activity over sets of units to implement fuzzy logic, memory storage, and automatic generalization. Unlike rigid logic gates, auto-associators connect every unit to every other unit, creating content-addressable memory with graceful degradation, constraint satisfaction, and similarity-based generalization. Auto-associators can automatically complete the missing parts of a word or recognize patterns despite noisy input or missing data.
The Defending Champion
Raw neural networks struggle with higher cognitive feats like representing distinct individuals, compositionality, and recursion. Connectionist associationism fails to distinguish identical individuals, compute syntactic composition without role confusion, handle quantification, avoid catastrophic forgetting, or process recursive nested propositions without structured symbolic architecture. Identical twin brothers Shawn Blick and Jonathan Blick were both charged after one bit a police officer's ear, but justice requires identifying the specific individual who committed the act.
Philosophical attacks on the computational theory of mind, such as Searle's Chinese Room and Penrose's quantum arguments, fail to disprove computational intelligence. Searle appeals to common sense through the Chinese Room thought experiment to claim symbol manipulation lacks understanding, while Penrose invokes Gödel's theorem and quantum mechanics in microtubules. Both arguments are empirically sterile and overlook that slow simulations obscure fast computational understanding. John Searle imagines a man inside a room following instruction sheets to manipulate Chinese characters without understanding the language.
3. Revenge of the Nerds
Human uniqueness is explained by the cognitive niche, an evolutionary adaptation involving complex reasoning, social cooperation, and technology.
Natural selection is the only known physical mechanism that creates complex adaptive design without requiring a conscious designer. By combining replication, finite resources, competition, and random copying errors where beneficial traits are passed down, natural selection simulates forward-directed teleology through ordinary forward causation.
Computer scientists Dan Nilsson and Susanne Pelger simulated the random mutation of a primitive light-sensitive skin slab. It naturally evolved into a complex fish-like eye in just four hundred thousand generations.
Alternative theories of evolution such as Lamarckism and macromutations fail because they cannot account for the emergence of functional complexity. Several hundred generations of circumcision in Jewish boys have not caused them to be born without foreskins.
Brain evolution is heavily constrained by biological costs, proving that intelligence is not an inevitable goal of evolution. Neural tissue is metabolically expensive and bulky, requiring compromises in human anatomy, increased childhood dependency, and slower reaction times for complex cognitive processing compared to lower-tech animals.
The average man's IQ is 107 while a brown trout's IQ is 4, yet a man cannot easily catch a brown trout. Human brains take up two percent of body weight and consume twenty percent of energy and nutrients. Cognitive experiments reveal a seventy-milliseconds unaccounted transmission time between stimulus and response.
The ability to learn during an organism's lifetime can guide and accelerate the evolution of innate neural structures. Geoffrey Hinton and Steven Nowlan demonstrated this Baldwin effect using computer simulations of neural networks requiring twenty correct connections. Two to the twentieth power gives the number of genetically distinct organisms needed to find a fit mutant in a non-learning population.
Life's Designer
Animals possess highly sophisticated, specialized information-processing algorithms rather than general-purpose intelligence or simple reflexes. Tunisian desert ants measure distance and direction on their foraging journeys using path integration or dead reckoning to make a direct line back to their nests. A desert ant's nest has a diameter of one millimeter, and an ant may travel fifty meters from it. Birds require learning the celestial pole to account for the twenty-seven thousand years wobble cycle of the earth's axis of rotation.
Humans entered the cognitive niche by using abstract intuitive theories and causal reasoning to outwit the defenses of other organisms. Unlike species that adapt only over evolutionary time through physical defenses, you analyze the world and invent novel, goal-oriented manipulations that overcome biological constraints.
!Xõ trackers in the Kalahari Desert read subtle signs of trampling, footprint ages, and behavior patterns to track and locate wild game.
The Cognitive Niche
Human intelligence was made possible by a unique combination of primate visual adaptations, social living, precision hands, and hunting. Out of an estimated fifty million animal species on earth, only one followed the route to human intelligence. The human brain tripled in size over a timescale of five million years. Chimpanzees use stones to crack open nuts and stems to fish out termites, demonstrating early manual tool use.
Human evolution occurred gradually over millions of years alongside tool manufacture, bipedalism, and brain expansion. Upright posture began four or four and a half million years ago. Homo habilis and evidence of butchery appeared two million years ago. Anatomically modern humans and mitochondrial Eve emerged between 200,000 and 100,000 years ago. A population bottleneck reduced ancestors to ten thousand people sixty-five thousand years ago. Archaeologists exploring sites in Zaire found finely crafted bone implements and grindstones dated at seventy-five thousand years ago.
Human behavior in modern society often diverges from Darwinian survival imperatives because your mind remains adapted to ancestral foraging bands. You spend ninety-nine percent of human existence living in small foraging bands.
People watch pornography instead of seeking a mate, forgo food to buy drugs, and consume excessive sugar and fat despite modern abundance.
4. The Mind's Eye
Visual perception is an ill-posed problem solved by the brain using built-in assumptions about the physical world.
The visual system works backward from a two-dimensional retinal image to reconstruct a three-dimensional environment by betting on regularities like lighting from above and rigid shapes. Because projection destroys depth information, an infinite number of three-dimensional arrangements can produce the same two-dimensional retinal image. The visual system converts these ill-posed problems into solvable ones by applying premises such as that matter is cohesive, surfaces are uniformly colored, and viewpoints are generic rather than coincidentally aligned.
Christopher Tyler created autostereograms, and Adelbert Ames, Jr. built distorted rooms like the Ames room, where irregular shapes and slants force the visual system's assumptions to fail, creating striking illusions. David Marr defined vision as a process that transforms images of the external world into a useful description free of irrelevant information.
Introduction to Vision
Stereoscopic vision calculates object depth by combining the slightly different projections of the two eyes using high school trigonometry. Because the eyes are separated in the skull, objects at different distances fall on different retinal locations through binocular parallax.
Charles Wheatstone demonstrated this in 1838 by designing the stereoscope, placing two pictures taken from slightly different vantage points in front of each eye to produce a three-dimensional illusion. Bela Julesz proved that object recognition is not a prerequisite for stereo matching by generating computer-made random-dot stereograms. When shifted patches of random dots are viewed stereoscopically, the camouflaged shape levitates into the air purely through low-level disparity computations. Bela Julesz used computer graphics to create random-dot stereograms resembling television snow, proving that the brain solves the correspondence problem without needing to recognize objects like lemons or cherries first.
David Hubel and Torsten Wiesel raised kittens and baby monkeys with one eye covered, resulting in permanent functional blindness in the deprived eye due to the loss of cortical real estate.
Stereo Vision and Shading
The visual shape analyzer combines prior probabilities about the world with likelihoods derived from perspective projection using Bayes' theorem. To recover shapes from two-dimensional lines and angles, the shape analyzer bets on the most probable physical object by multiplying prior probabilities of encountering regular, parallel, or symmetrical structures by the likelihood of the projection. Edward Adelson designed zigzag box drawings demonstrating how identical lines are interpreted as paint boundaries in one context and shape-and-shading folds in another based on 3-D shape cues. Edwin Land developed the Retinex theory, positing that gradual changes in brightness and color across a visual field are attributed to lighting gradients, whereas abrupt transitions signal material boundaries between distinct objects. Edwin Land tested his color constancy theory using artificial 2-D rectangular patches called Mondrians, demonstrating that people perceive consistent colors despite changing illumination.
Human vision does not produce a complete three-dimensional scale model of the world, but rather a view-dependent representation restricted to visible surfaces and two-dimensional geometry. Vision produces a 2.5-D sketch, which is a pixel-like mosaic representing surfaces and edges from the viewer's perspective rather than a 3-D model. Depth is downgraded to half a dimension because it must be computed from active data rather than hitting the retina directly. Each cell in the array holds information about depth, slant, tilt, and color for surfaces, or boundary types and orientations for edges. Closing one eye and nudging the other with a finger causes the world to jump because the frame-shifter is bypassed, whereas voluntarily moving the eyes copies the muscle command to shift the reference frame and keep the world steady.
10 Key Ideas
- Psychology and biology can be understood as reverse-engineering complex organic systems produced by natural selection.
- The mind is not the physical brain itself, but rather what the brain does through information processing and computation.
- Intelligence is a form of information processing where symbols simultaneously carry information and cause physical events to happen.
- Human uniqueness is explained by the cognitive niche, an evolutionary adaptation involving complex reasoning, social cooperation, and technology.
- Visual perception is an ill-posed problem solved by the brain using built-in assumptions about the physical world.
- The visual system works backward from a two-dimensional retinal image to reconstruct a three-dimensional environment by betting on regularities like lighting from above and rigid shapes.
- Human behavior in modern society often diverges from Darwinian survival imperatives because your mind remains adapted to ancestral foraging bands.
- The visual shape analyzer combines prior probabilities about the world with likelihoods derived from perspective projection using Bayes' theorem.
- Human vision does not produce a complete three-dimensional scale model of the world, but rather a view-dependent representation restricted to visible surfaces and two-dimensional geometry.
- Animals possess highly sophisticated, specialized information-processing algorithms rather than general-purpose intelligence or simple reflexes.