Cognitive Science Chapter 12 & 13
Acting rationally
- achieve the best outcome; act best when uncertainty exists; produce the best expected outcomes
Goal-based
- actions depend on goals; - retain goal information with desirable situations.
Cognitive / behavioral model
- From K. Craik
Goal-based agents
- goal planning
Simple reflex agents
- survival
Evolutionary computational
- A concept parallel to AGI that depends on a collection of computational methods that have been modeled on the principles of biological evolution
Turing test
- A machine entity that passes this test responds in such a way that it cannot be distinguished from a human - A machine's ability to exhibit behavior in a way that the machine appears to be human • Conversation is not face-to-face: carried out in writing questions and reading the answers to those questions
Behavior
- A mapping or translation of sensory inputs into a pattern of motor actions intended to complete a task
Actuators
- A mechanical device for moving or controlling a mechanism or system
Depth-first algorithm
- A methodology for exploring a knowledge base in which information is examined in an increasingly specialized manner
Paradigms
- A philosophy or approach for developing theories for, analyzing, and evaluating a class of problems
Intelligent Agent (IA)
- A rational autonomous entity that observes and acts on an environment, directs its activity toward achieving goals, and is capable of learning to achieve its goals through experience; includes cooperating entities - complete machine implementation of human thinking, feeling, speaking, symbolic processing, remembering, learning, knowing, problem solving, consciousness, planning, and decision-making.
Reactive / subsumptive
- A robotic architecture characterized by direct connection between perceptive elements (sensory information) and actuating elements (sense-action processes)
Plan
- A robotic primitive element encompassing the corresponding human attributes of reasoning and cognition
Sense
- A robotic primitive that includes that part of a robotic system that converts elements of an environment into information that is made available to other parts of the robotic system
Programs
- A series of instructions that an intelligence devises for the computer to execute
Telepresence
- A set of technologies that gives a person the feeling of a physical presence in a real or virtual setting
Firefighting
- A simple multiagent example
Expert systems
- A software program that is dedicated to solving problems and providing "expert-quality" advice to users; demonstrates intelligence similar to that of a human
Behaviorsim
- A way of conceiving of empirical constraints on psychological states
Reasoning
- Ability to draw inferences appropriate to the facts and the situation
Autonomous entity
- Acting without human intervention or control
Simple reflex
- Actions based on existing precepts (survival)
Top-down
- An abstract representation of mathematical models that aims to describe functionality at a very high level, then partition it into more detailed levels until the detail is sufficient for machine implementation
Programming
- Argument: A machine can never be programmed to do anything new - Evaluation: It can also not be demonstrated that a human idea is not the result of teaching.
Mathmatical
- Argument: Some theorems can neither be proved nor disproved - Evaluation: Technically true, though no such limitations apply to the human intellect
Nervous system
- Argument: The nervous system is not a discrete-state machine. A machine cannot mimic nervous system behavior - Evaluation: A digital computer could be programmed to produce results indicative of a continuous organization
Theological
- Argument: Thinking is a function of man's (God-given) immortal soul. - Evaluation: This argument is a serious restriction of the omnipotence of the Almighty (a name or title for God)
Consciousness
- Argument: Universal Computing Machine can never reproduce consciousness - Evaluation: This is solipsist point of view. How do you define thinking? *Solipsism: The view or theory that the self is all that can be known to exist
Mind as a machine
- Attempts to emulate the human brain • May require a comprehensive theory of mind
Subsumptive architecture
- Based on parallels, or loosely coupled processes, intelligence emerges from a large number of parallels and is coordinated through interaction with the environment
If-then rules
- Conditional execution of a group of statements (consequents) depending on the value of an expression (antecedents)
Overall challenges to an IA
- Considerable criticism of "________" AI has come from the neuroscientific community - Coding of models - Categorization requirement (facts, rules) - Procedure (algorithmic processes)
Computational
- Considerable criticism of "________" AI has come from the neuroscientific community (Edelman & Reeke, 1988) - "Artificial intelligence is a science that finds itself in somewhat the same epistemological position as Aristotelian dentistry. Aristotle stated that women have fewer teeth than men and attributed this characteristic to women's supposed lesser need, men being stronger and more choleric, but he never bothered to look in Mrs. Aristotle's mouth to verify his theory. Similarly, AI has developed as an almost entirely synthetic enterprise, quite isolated from the complementary, analytic study of the biology of natural intelligence represented by psychology and the neurosciences."
Finite state model
- Contemporary implementation of Turing's universal computing machine
Central processor unit (CPU)
- Digital computer: Calculations, logical decision, program sequence control. - Turing's finite state descriptor: determines state transitions. - Craik behavioral model: makes cognitive decisions (cognitive manipulation)
Input / Output
- Digital computer: Sensor information, control of all external system elements (equipment) - Turning's finite state descriptor: Receives sensory information ("x"), and provides control ("y") to external world changes - Craik behavioral model: Signals: from external sensors; to external actuators; conversion to internal representation; conversion to action signals
Memory
- Digital computer: Stores programs, results, temporary results, and data - Turing's finite state descriptor: Stores state definitions (SO,...), external information ("x"), transition (IF-THEN), and rules ("x/y") - Craik behavioral model: Memory: facts, cognitive rules, cognitive methods
Communication (Bus)
- Digital computer: communication between other elements of the computer - Turing's finite state descriptor: communications with external world - Craik behavioral model: communications with external world
Architecture
- Functional organization of an entity, possibly a computer
Hawkins AI model
- Hawkins proposes an architecture based on the neocortex. Our brains compare perceptual inputs to expectations.
Human brain replica
- HumanBrainConsistsof: • ~ 100,000,000,000 Neurons • ~ 100,000,000,000,000 Synapses • C. elegans brain consists of 302 neurons • Drosophila brain contains ~100,000 neurons - A model of the human brain would have to represent all neurons, all neuronal connections, and keep track of ongoing processes
Historical AI
- Mechanical: Calculating machines (Pascal, Leibnitz, Newton Babbage) • Ada Lovelace first programmer (created an algorithm to be carried out by a computer: Bernoulli numbers) - Intellectual/Philosophical: Logic (Aristotle); mathematical calculus (Leibnitz, Newton); Knowedge-based agent: (Craik); computation (Turing). - Electronic and computer: computer (Zuse, Eckart, IBM, Intel); integrated circuit (Shockley, Kilby)
Technological singularity
- Invention of an artificial superintelligence that will result in rapid technological growth, changing humanity • Each new generation would improve on what came before at incomprehensible rates • Supported by observed exponential growth in computing power • Kurzweil predicts 2045; others predict far sooner • What will become of humans?
Rote learning
- Learning that focuses on repetition or memorization
Strong AI
- Machines can act intelligently AND have minds in the same sense that people do - Intelligence is required to: • Reason • Represent knowledge • Plan • Learn • Communicate • Solve problems • Though intelligence alone is not strong AI - Also includes: • Consciousness: subjective experience and thought • Self awareness: awareness of existence as a separate entity and of one's own thought • Sentience: subjective perception of feelings (qualia)
IAs
- Numerous practical examples of ____ provide encouraging evidence that the disciplines of psychology, biology, computer science, and engineering may eventually lead to a machine that "exceeds human intelligence."
DNA
- One of the emerging technologies to attempt strong AI - Description: Based on human biology. Trillions of DNA molecules within a test tube. Each performing a given operation on differing data - 660 trillions calculations per second (cps)
Spin (quantum computing)
- One of the emerging technologies to attempt strong AI - Description: Computing with the spin of electrons. Spin is a quality of electrons within an atom. Subject to laws of quantum mechanics - Potential capacity: Mainly for memory - retains information when power is removed
Nanotubes
- One of the emerging technologies to attempt strong AI - Description: Hexagonal network of carbon atoms rolled up into a seamless cylinder - Potential capacity: High density, high speed (1000 Gigahertz; thousand times a modern computer; logical switch size 1 x 10 nanometers)
Light
- One of the emerging technologies to attempt strong AI - Description: Laser beams perform logical and arithmetic operations - Potential capacity: 8 trillion cps
Molecules
- One of the emerging technologies to attempt strong AI - Description: To switch states, change the energy level of the structure within a "rotaxane" molecule. - Potential capacity: 10 to the power of 11 (10の11乗) bits per square inch
Robots
- Particularly, the mechanical embodiment of an IA as a mechanical machine that can function autonomously
Trial-and-error learning
- Proceeds from random responses to a behavior that satisfies the requirements of the task
Applet
- Prototyping software tool application
Robotic
- Representative Sensor: Cameras, infrared range finders, tactile sensors, odor detectors - Representative Actuators: Motors and other actuators
Human
- Representative Sensor: Eyes, ears, tactile, hands, legs, mouth, nose - Representative Actuators: Hands, legs, mouth, arms
Cognitive (software)
- Representative Sensor: Keystrokes, file contents, network packets - Representative Actuators: Display devices (optical, audio), file outputs, packet transmission
Membership function
- Represents the degree of truth of a variable. - They are not equivalent to the probability of the variable, rather membership in the universe of outcome
AI & consciousness
- Requires a solving the descriptive and combinatory problem of Marr's levels (computational, algorithmic, implementation) - Human thinking is flawed (use of both logic and heuristics), but computers are completely logical • Yet humans are conscious... • Those that argue that consciousness is an illusion would argue there is no need for "strong" AI
Artificial general intelligence
- Seeks to explain how real-world minds achieve general intelligence using limited resources by appropriately organizing and utilizing their memories - A model envisioned by Minsky, McCarthy and others - A "thinking machine" with human-like "general intelligence" - To include: self-awareness, will, attention, creativity as well as human qualities we take for granted. To date, only formative thinking characterizes AGI.
Russell / Norvig generic IAs
- Simple reflex - Model-based - Goal-based - Utility-based - Learning agents
Conscience
- Startle blinks - Emotional words: kill, maim, joy
Aristotelian logic
- Syllogisms (deductions) • Suppose two principles and a conclusion: • Socrates is a man • Man is mortal • Therefore, Socrates is mortal - Induction: argument from the particular to the universal
Foraging
- The act of looking or searching for food
Embodiment
- The idea that cognitive processes exist within and interact with an external environment through sensory inputs and motor outputs
Limbic system
- These scans show brain activity in empathy-generating centers of the ______ _____ in normal individuals (left) and in psychopathic individuals (right) when they are exposed to violent images.
Testing AIs
- Turing test - Voight-Kampff questions - Hare psychopathy test
Universal computing machine
- Turing's vision of a machine that could solve a problem if it was mathematically solvable
Evolutionary AI
- VirgilGriffith • Evolutionary algorithms as a shortcut to AI • Let AI evolve on its own • Competition as a drive • Limitations: costs to characteristics • Cost for brain size • Energy loss - Strong AI may combine weak AI with some kind of evolutionary drive • The intelligent agents must "want" something • The intelligent agents must have the processing power of human brain • Theory of mind • Possibly initially modeled on brain
AI
- ___ is emerging as a central element of cognitive science; methodologies lend themselves to study in biological modeling, principles of intelligent behavior, robotics.
Utility-based
- enhanced goal-based agents - add a quality factor.
Utility-based agents
- goal-based agents focused on quality and optimum solutions
Model-based
- keep track of changing precepts; - maintains an internal state that it uses to develop responses.
Model-based reflex agents
- keep track of information to solve problems
Learning agents
- keep track of results to improve learning
Emotional words
- kill, maim, joy - In tests in which people have to separate real words from nonsense words, normal volunteers process these emotionally charged words more slowly than neutral words, such as 'table' and 'butter'. - Psychopaths process both categories of word at the same speed.
Acting humanly
- knowledge representation, reasoning, learning
Bottom-up
- methods that describe mathematical models as proceeding from fundamental elements by starting with the smaller, more fine details of the model and then building upward until a comprehensive model ensues
Learning agents
- outgrowth of Turing (universal computation); - build a learning machine and then "teach it." (preferred model)
Coding of models
- programmer must find a suitable representation of the information; - what symbolic manipulations may be required; - what antecedent requirements on the representation; - human cognition may not even rely on symbolic computation at all.
Thinking rationally
- solve any problem described in logical notation - exemplified by Aristotelian principles*
Thinking humanly
- subsumes psychological elements (introspection, neurological actions of brain using brain imaging)
AI
- the computational elements of IAs.
Categorization requirement (facts, rules)
- the programmer must specify a sufficient set of rules to define all the categories that the program must support.
Procedure (algorithmic processes)
- the programmer must specify in advance the actions to be taken by the system for all combinations of inputs that may occur. - The number of such combinations is enormous and becomes even larger when the relevant aspects of context are taken into account.
Alicebot
- worksbyabruteforcesolution:trying to figure out every possible question, and come up with every possible answer. • If a bot like this passed the Turing test, would that mean it is intelligent? • No one would argue that the alicebot is conscious. • But...you could program a bot to argue that it is conscious
Hare psychopathy test
These are the 20 criteria: - Do you exhibit glib and superficial charm? - Do you have a grandiose (exaggeratedly high) estimation of self? - Do you have a constant need for stimulation? - Are you a pathological liar? - Are you cunning and manipulative? - Do you have lack of remorse or guilt? - Do you have shallow affect (superficial emotional responsiveness)? - Are you callous, and do you lack empathy? - Do you have a parasitic lifestyle? - Do you have poor behavioral controls? - Are you sexually promiscuous? - Did you display early behavior problems? - Do you lack realistic long-term goals? - Are you overly impulsive? - Are you irresponsible? - Do you fail to accept responsibility for own actions? - Have you had many short-term marital relationships? - Do you have a history of juvenile delinquency? - Have you experienced a revocation of conditional release? - Do you display criminal versatility?
Multiagent IAs
• A cooperative (or noncooperative) group of IAs capable of sophisticated information processing activity. • Based on mechanisms that specify the kinds of information they can exchange and their method for doing so.
IA classifications
• Acting humanly: • Thinking humanly: • Thinking rationally: • Acting rationally:
Approaches to Artificial General Intelligence (AGI)
• Combine many modular AI programs • Chatbots (Alice) • Build a mechanical human brain replica • Evolve AGI • Build a full model of human intelligence/consciousness
Weak AI
• Machines can act intelligently (Chinese room only) • A perfect alicebot would fall under this category. • Capable, not creative, can not act outside of fixed programming
Startle blinks
• People blink when startled. This response becomes more intense if they are placed in a threatening environment — (i.e., showing a picture of an angry face.) • The effect in psychopaths is unchanged.
Turing's finite state machine
• Sn = State (condition) definition of the system with a number (n) indicating the specific state. • x/y = "x" indicates what stimulus (from the external world) is detected; "y" what action is to be taken when "x" occurs. The action "y" will move the state of the system to a new state (or possibly retain the original state).
Evolutionary computing
• Some similarity to AGI but modeled on the principles of biological evolution. • Aims to solve real world problems: finance; software design; robotic learning • Model and understand natural evolutionary systems existing in: economics, immunology, ecology • A metaphor for the operation of human thought processes - singularly germane to achieving an IA
Evolutionary programs
• Sub programs compete for resources in an artificial environment • Sub programs try different solutions to maximize fitness • Offspring continue to work to maximize fitness
Issues in the singularity
• Will the AI have similar goals to humans? • What if the AI's goals are not the same as our? • Will we be "plugged in"? • What if only some people can afford bio-digital fusion? • Will the new world be incomprehensible to us? • Critics like Searle and Pinker believe the singularity to be unlikely
Voight-Kampff questions
• You've got a little boy. He shows you his butterfly collection plus the killing jar. What do you do? • You're watching television. Suddenly you realize there's a wasp crawling on your arm. How would you react?