Hanyang University Il Hong Suh 2012.4.17 1
4.5 Ga : 1.8 Ma = 1 day : 34.6 Sec Evolution of Species Mitochondria cyanobacteria eukaryote trilobite fish Bacteria polymerization Darwinian Creature Skinnerian Creature 3.8 billon y 1 billon ye 700 million y 4 billon ye 2 billon ye 4.5 billon y ears Simple cel Complex c ars ears ars ars ears ls ells Amphibia Dinosaur Mammalia Human Gregorian Creature Popporian Creature 542 million y 2.5 million Origin of L Australopithecus 1.8 million ears Cambrian ex anguage years years plosion 2
Four Kinds of Minds • Tower of Generate-and-Test – By a process of evolution by natural selection – Important advances in cognitive power Daniel Clement Dennett Gregorian creature (1942~ ) • American philosopher • Evolution biology and cognitive science • Kinds of Minds: Toward an understand Popperian creature of consciousness • Darwin’s Dangerous Idea: Evolution and the meanings of life • Intentional stance(beliefs and desires) Skinnerian creature Darwinian creature Tower of Generate-and-Test 3
Darwinian Creature • Darwinian evolution of species by natural selection – Generated by recombination and mutation of genes Charles Darwin – Field-tested, and only the best designs survived (1809~1882) • English naturalist • The origin of species • Evolution resulted from nat ural selection Gregorian creature Popperian creature Skinnerian creature Sensory Signal Darwinian creature (Data) Tower of Generate-and-Test 4 4
Darwinian Creature • Behavior-based Intelligence Intention Prediction Specialized sensor-base Preservation of sp d Simple Prediction ecies ( Simple bottom-up pro cessing ) Gregorian creature Popperian creature Skinnerian creature Darwinian creature Tower of Generate-and-Test Perception Model/Learning Diving beetle se 1 or 2 High performance S Fixed & tightly coupled Sen nses waves of 1 - pecialized Sensors sory-motor Coordination / N 9 m ( Very Low entropy, atural selection Very simple data ) 5
Skinnerian Creature • Property of phenotypic plasticity • Simple sort of “experience” Burrhus Frederic Skinner – Getting a positive or negative signal (1904~1990) • American psychologist – Adjusted probability of that action • Operant conditioning (Skinner bo x) • Evolution resulted from natural s election Gregorian creature Popperian creature Skinnerian creature Related data Learning Darwinian creature (reinforcement) Tower of Generate-and-Test 6
Skinnerian Creature • Operant conditioning, reinforcement Prediction Stimulus-Response lear Intention ning-based Prediction Egocentric Surviva ( Complex bottom-up pr l ocessing ) Gregorian creature Popperian creature Skinnerian creature Learning (reinforcement) Darwinian creature Tower of Generate-and-Test Perception Model/Learning High performance Specializ Plastic & loosely coupled S ed Sensors ensory-motor Coordination / Reinforcement Learning ( Low entropy, simple data ) 7
Popperian Creature • Preselection and prediction • “Permits our hypothesis to die in our stead” Sir Karl Raimund Popp – The inner environment contains about the er (1902~1994) • Austrian and British philosophe outer environment and its regularities r • Critical rationalism • Filtered Pattern • Scientific method by advancing empirical falsification Gregorian creature Mind rehearsal Popperian creature Filtered Pattern (image based) Skinnerian creature Learning Darwinian creature (reinforcement) Tower of Generate-and-Test 8
Popperian Creature • Pattern-based Prediction Intention Prediction Partially Prediction by Pattern-ba sed Simulation Altruism ( Mostly bottom-up & si mple top-down proces sing ) Gregorian creature Popperian creature Mind rehearsal (image based) Skinnerian creature Learning (reinforcement) Darwinian creature Tower of Generate-and-Test Perception Model/Learning Unspecialized Sensors Pattern-based Hierarchical Memory / Pattern Classifica ( High entropy, tion or Clustering complex data ) 9
Gregorian Creature • Mind tools: words • Benefiting from the experience of others Richard Langton Gregory with the mind tools(words) (1923~) • British psychologist • Sharable and reusable Knowledge • Emeritus professor of neuropsychol ogy at the university of Bristol • Eye and Brain, Mind in Science • the modern founder of the science o f perception Semantics, lang Gregorian creature Knowledge representation (sharing) uage Imitation Popperian creature Mind rehearsal (image based) Skinnerian creature Learning Darwinian creature (reinforcement) Tower of Generate-and-Test 10
Gregorian Creature • Imitation & Symbol-based Prediction Intention Prediction Fully Altruism, Lud Inference-based Predicti on from many knowledg ens (Play) e resources ( Complex bottom-up & top-down processing ) Knowledge representation (sharing) Gregorian creature Imitation Popperian creature Mind rehearsal (image based) Skinnerian creature Learning (reinforcement) Darwinian creature Tower of Generate-and-Test Perception Model/Learning Unspecialized Senso Symbolic Model rs / Pattern and Rule ( very high entropy, Learning Very complex data ) 11
Gregorian Creature : Semantic Representation · Important resource for language processing · Reducing the amount of data to be stored in memory · Strongly invariant to scene variations · Logical inference using relations between concepts … 12
Characteristics of Four Kinds of Mind (Intelligence) Darwinian Le Skinnerian Le Popperian Le Gregorian Le Level vel vel vel vel 1 or 2 High performanc Some High performanc e Specialized Sensors Unspecialized Sensors Unspecialized Sensors e Specialized Sensors Perception very simple data, complex data, very complex data, simple data, very low entropy high entropy very high entropy low entropy Inference-based Pr Specialized sensor- Stimulus-Response ediction from many Prediction by Patter Prediction based Simple Predi learning-based Pred n-based Simulation knowledge resource ction iction s Plastic & loosely couple Pattern-based Hierarch Model / Learni Fixed & tightly coupled Symbolic Model d Sensory-motor Coord ical Memory / Pattern C Sensory-motor Coordin / Pattern and Rule Lea ng ination / Reinforcement lassification or Clusteri ation / Natural selection rning Learning ng 13
Subsumption of Four Kinds of Mind (Intelligence) Gregorian Intelligence Popperian Intelligence Learning Learning Skinnerian Symbolic Model Intelligence Pattern-based Hiera / Pattern and Rule Le Learning rchical Memory / Pat Darwinian arning Reinforcement le Intelligence tern Classification or Prediction arning Clustering Learning Natural selection Prediction Prediction Prediction Inference-based Predi Stimulus-Respo Specialized sensor ction from many know nse learning-bas -based Prediction by Pattern- ed ledge resources based Simulation 14
Fundamental Mind Functions Manipulation Planning Recognition Navigation 15
Can We Develop Mind Functions in the Brain of Gregorian Creature? Manipulation Planning Recognition Navigation 16
Three Fundamental Information Processes in Human Brain Red : feedback Green :feed-forward Thalamocortical system Prediction is processed by top-down infor (Learning is processed by feedback ) mation Category learning Sequential learning Language, grammar learning ! Granger R, Engines of the Brain: The computational D. George amd J. Hawkins, Towards a mathematical theory of cortic instruction set of human cognition, AI Magazine (2006) al micro-circuits, PLoS Computational Biology, 2009. 17
What Is Key Property in Information Processes of Gregorian Brain RBU(Risky But Useful) Process Top-down information Pencil Pencil Prior Christian Dior Feedback signal Christmas Doll Attention … (M. Butz et al. 2003.) 18
RESEARCHES ON FOUR MIND FUNCTIONS IN INCORL Planning · Reactive planning · Improvisational planning · Proactive planning Navigation · Semantic SLAM and navigation · L-SLAM Recognition · Oriented edge-selective band-pass filtering Manipulation · Skill Learning
List of Contributors • Post Doctor – Sang Hoon Lee • Recognition – Gi Hyun Lim • Planning • PhD Student – Guoxuan Zhang, Jin Han Lee, Dong Wook Ko • Navigation – Woo Young Kwon, Sang Hyoung Lee • Manipulation – Young Bin Park , Gwang Geun Ryu, Deok Hyeon Cho, Se Hyung Lee • Recognition – Seung Woo Hong • Planning 20
PLANNING
Four Kinds of Planning Gregorian Intelligence Popperian Intelligence Skinnerian Proactive planning Motion planning Intelligence Improvisational plann Reactive planning ( Darwinian Sensorimotor casca ing Intelligence des) Reflexes Skinnerian Conditi phototaxis oning 22
Future plan - Integrated framework for Gregorian-level planning including Proactive Planning Proactive be havior Improvisational Planning Alternative b ehavior Unexpected situation Reactive Planning Situation ade quate behavi Undefined response or Sensors Reflex Control Actuators Instinct behavior 23
REACTIVE PLANNING
Goal-Oriented Control vs. Reactive Control Based on Reactive Planning Every possible state sequence of a task (fully connected finite state machine) Reactive but not goal-oriented Goal-oriented as well as Reactive 25
Recommend
More recommend
