CS 188: Artificial Intelligence Spring 2007AnnouncementsSci-Fi AI?What is AI?Thinking Like Humans?Basic IdeasBRAINSlide 9Imaging the BrainSensory SystemsMotor SystemsNEURONBrains ~ ComputersSlide 16Slide 17Observation of action activates premotor cortex in topographic manner, consistent with motor topography.Slide 19Significance of Mirror NeuronsModels of LearningSlide 22Synaptic PlasticityThe Hebb rule is found with long term potentiation (LTP) in the hippocampusNeural Correlates of RLSlide 26Dopamine Levels track RL signalsCurrent HypothesisNeural Basis of IntelligenceSlide 30Brain Machine InterfacesSensory ProsthesisVisual ProsthesisCochlear ImplantsBCI using EEGEEG Control of a robot in a labyrinthDecoding Cognitive SignalsSlide 38Slide 39Slide 40Slide 42CS 188: Artificial IntelligenceSpring 2007Lecture 27: Neural Computation5/1/2007Srini Narayanan– ICSI and UC BerkeleyAnnouncementsReinforcement Learning Q/AWednesday 6 PM, 306 SODAHelicopter control talkThursday (4 -5:30 PM) 310 SODASci-Fi AI?What is AI?Think like humans Think rationallyAct like humans Act rationallyThe science of making machines that:Thinking Like Humans?The Cognitive Science approach:1960s ``cognitive revolution'': information-processing psychology replaced prevailing orthodoxy of behaviorismScientific theories of internal activities of the brainWhat level of abstraction? “Knowledge'' or “circuits”?Cognitive science: Predicting and testing behavior of human subjects (top-down)Cognitive neuroscience: Direct identification from neurological data (bottom-up)Both approaches now distinct from AIBoth share with AI the following characteristic:The available theories do not explain (or engender) anything resembling human-level general intelligence}Hence, all three fields share one principal direction!Images from Oxford fMRI centerBasic IdeasNeural ComputationThe brain as a computing deviceLearning in the brain.Brain-based ComputingBrain Machine InterfacesBRAINBroca’sareaMotor cortex Somatosensory cortexSensory associativecortexPrimaryAuditory cortexWernicke’sareaVisual associativecortexVisualcortexImaging the BrainSensory SystemsVision (nearly 30-50% )Audition (nearly 10%)Somatic ChemicalTasteOlfactionMotor SystemsLocomotionManipulationSpeechNEURONBrains ~ Computers1000 operations/sec100,000,000,000 units10,000 connections/graded, stochasticembodiedfault tolerantevolves, learns1,000,000,000 ops/sec1-100 processors~ 4 connectionsbinary, deterministicabstractcrashesdesigned, programmed (usually)Approx 2 s: Experimenter picks up foodApprox 4 s: Monkey picks up foodComputing in the Brain: Mirror Neuron in F5 (premotor cortex)Many human imaging studies showing activation of motor regions (primary and secondary) during action perception.Regions overlap with those engaged during action production.Distributed frontal/parietal activation during viewing of actions performed with mouth, hand, or foot. All activations compared to rest baseline. Mouth Hand Foot (chewing, biting) (grasping, pinching) (kicking, jumping)Buccino et al. 2001)Observation of action activates premotor cortex in topographic manner, consistent with motor topography. Foot ActionFoot ActionHand ActionHand ActionMouth ActionMouth Actiona) no-objectb) w/ objectMEG (magnetoencephalography) study comparing pianists and non-pianists. Pianists show activation in primary motor cortex when listening to piano.Activation is specific to fingers used to play the notes. Colored region: MEG signal for pianists minus non-pianists.Significance of Mirror NeuronsAction, Perception, Imagination, and Understanding share a lot of the same brain circuits. Question:How are these circuits learnt?Models of LearningHebbian ~ coincidence Reinforcement ~ delayed rewardSupervised ~ correction (perceptron, mlp)Unsupervised-similarityLong-term Potentiation (LTP)Rapid and long-term increase in synaptic strength resulting from the pairing of presynaptic activity with postsynaptic depolarizationSynaptic PlasticityHebb’s Postulate: When an axon of cell A... excites cell B and repeatedly or persistently takes part in firing it, some growth process or metabolic change takes place in one or both cells so that A's efficiency as one of the cells firing B is increased.Slices of the hippocampus can be removed and its CA1 neurons studied in vitro with recording electrodes. Rapid, intense stimulation of presynaptic neurons evokes action potentials in the postsynaptic neuron. This is just what we would expect from the properties of synapses.The Hebb rule is found with long term potentiation (LTP) in the hippocampus1 sec. stimuliAt 100 hzSchafer collateral pathwayPyramidal cellsNeural Correlates of RLParkinson’s Disease Motor control + initialtion?Intracranial self-stimulation;Drug addiction;Natural rewards Reward pathway? Learning?Also involved in:Working memoryNovel situationsOCDSchizophrenia…ConditioningIvan Pavlov= Conditional stimulus = Unconditional stimulusResponse = Unconditional response (reflex); conditional response (reflex)Dopamine Levels track RL signalsUnpredicted reward(unlearned/no stimulus)Predicted reward(learned task)Omitted reward(probe trial)(Montague et al. 1996)Current Hypothesis Phasic dopamine encodes a reward prediction errorEvidenceMonkey single cell recordingsHuman fMRI studiesCurrent ResearchBetter information processing model Other reward/punishment circuits including Amygdala (for visual perception)Overall circuit (PFC-Basal Ganglia interaction)Neural Basis of IntelligenceHow does a system of neurons with specific processes, connectivity, and functions support the ability to think, reason, and communicate?Take CS 182 in Spring 2008.Basic IdeasNeural ComputationThe brain as a computing deviceLearning in the brain.Brain-based ComputingDetermining cognitive states from imaging data.Brain Machine InterfacesBrain Machine InterfacesSensory ProsthesisBrain Computer Interfaces from Brain signalsSensory ProsthesisVisual ProsthesisCochlear ImplantsBCI using EEGEEG Control of a robot in a labyrinthDecoding Cognitive
View Full Document