A tifi i l I t lli A I t d tiA tifi i l I t lli A I t d tiArtificial Intelligence: An IntroductionArtificial Intelligence: An Introduction
Revised September 2008Revised September 2008
Byoung-Tak Zhang
School of Computer Science and EngineeringSchool of Computer Science and EngineeringGraduate Programs in Cognitive Science, Brain Science,
and BioinformaticsSeoul National UniversitySeoul National University
E-mail: [email protected]://bi snu ac kr/http://bi.snu.ac.kr/
Can Machines Think?Can Machines Think?Can Machines Think?Can Machines Think?
Th T i TThe Turing TestComputing Machinery and Intelligence [Turing, 1950]
2(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Chess PlayingChess PlayingChess PlayingChess Playing
G K d D Bl © 1997
3(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Garry Kasparov and Deep Blue. © 1997
Mars Rover Sojourner:Mars Rover Sojourner: M P hfi d Mi iM P hfi d Mi iMars Rover Sojourner: Mars Rover Sojourner: Mars Pathfinder MissionMars Pathfinder Mission
4(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Natural Language ProcessingNatural Language ProcessingNatural Language ProcessingNatural Language Processing
Polysemy4 I keep the money in the bank.
4 I walk along the bank of the river4 I walk along the bank of the river.
Ambiguity4Time flies like an arrow4Time flies like an arrow.
4 I saw a man with a telescope.
Diversityy4She sold him a book for five dollars.
4He bought a book for five dollars from her.
Related Knowledge4Lexical, Grammatical, Situational, Contextual
5(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Expert SystemsExpert SystemsExpert SystemsExpert Systems
6(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Web Information RetrievalWeb Information RetrievalText Data
Classification System
Cl ifi i
Preprocessing and Indexing
Text Classification
Information Extraction
I f ti Filt i S t
DB Template Filling& InformationExtraction System
Information Filtering
Information Filtering System
questionuser profile
LocationDate
DB Record
Extraction System
quest o
feedback
answerfiltered data
7(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
feedbackDB
What Is Artificial Intelligence (AI)?What Is Artificial Intelligence (AI)?What Is Artificial Intelligence (AI)?What Is Artificial Intelligence (AI)?
Branch of computer science that is concerned with the automation of intelligent behavior.Design and study of computer programs that behave intelligentlyes g a d study o co pute p og a s t at be ave te ge t yStudy of how to make computers do things at which, at the moment, people are better.Designing computer programs to make computers smarterDesigning computer programs to make computers smarter.Develop programs that respond flexibly in situations that were not specifically
) h l i beg.) - house-cleaning robots- perceive its surroundings- navigate on the floor- respond to events- decide what to do next- space exploration (Fig. 1.1)p p ( g )
Synonyms of AI: machine intelligence
What is Artificial Intelligence?What is Artificial Intelligence?What is Artificial Intelligence?What is Artificial Intelligence?
AI i ll i f h d bl hi h b l d bAI is a collection of hard problems which can be solved by humans and other living things, but for which we don’t have good algorithms for solving.g g g4e. g., understanding spoken natural language, medical diagnosis,
circuit design, learning, self-adaptation, reasoning, chess playing, proving math theories etcproving math theories, etc.
Definition from R & N book: a program that4Acts like human (Turing test)4Thinks like human (human-like patterns of thinking steps)4Acts or thinks rationally (logically, correctly)
Some problems used to be thought of as AI but are nowSome problems used to be thought of as AI but are now considered not4e. g., compiling Fortran in 1955, symbolic mathematics in 1965,
9(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
pattern recognition in 1970
Research GoalsResearch GoalsResearch GoalsResearch Goals
Making machines more usefulg
U d t di i t lliUnderstanding intelligence
History of AIHistory of AIHistory of AIHistory of AI
Early enthusiasm (1950’s & 1960’s)4 Turing test (1950)4 1956 Dartmouth conference4 Emphasize on intelligent general problem solvingp g g p g
Emphasis on knowledge (1970’s)4 Domain specific knowledge4 DENDRAL, MYCIN
AI became an industry (late 1970’s & 1980’s)AI became an industry (late 1970 s & 1980 s)4 Knowledge-based systems or expert systems4 Wide applications in various domains
Searching for alternative paradigms (late 1980’s - early 1990’s)44 AI’s Winter: limitations of symbolic/logical approaches4 New paradigms: neural networks, fuzzy logic, genetic algorithms, artificial life
Resurge of AI (mid 1990’s – present)4 Internet Information retrieval data mining bioinformaticsInternet, Information retrieval, data mining, bioinformatics4 Intelligent agents, autonomous robots
Recent trends:4 Probabilistic computation4 Biological basis of intelligence
11(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
4 Biological basis of intelligence4 Brain research, cognitive science
Artificial Intelligence (AI)Artificial Intelligence (AI)g ( )g ( )
S b li AISymbolic AI Rule-Based Systems
Connectionist AI Neural Networks
Evolutionary AI Genetic Algorithms
Molecular AI:
12(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Molecular AI: DNA Computing
Research Areas and ApproachesResearch Areas and ApproachesppppLearning AlgorithmsInference Mechanisms
ResearchInference MechanismsKnowledge RepresentationIntelligent System Architecture
Intelligent AgentsInformation RetrievalElectronic Commerce
ArtificialIntelligence Application
Electronic CommerceData MiningBioinformaticsN t l L P
R ti li (L i l)
Natural Language Proc.Expert Systems
Rationalism (Logical)Empiricism (Statistical)Connectionism (Neural)Paradigm
13(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Evolutionary (Genetic)Biological (Molecular)
14(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Intelligent AgentsIntelligent AgentsIntelligent AgentsIntelligent Agents
Autonomous Agents
Biological Agents Robotic Agents Computational Agents
Software Agents Artificial Life g
Agents
Entertainment
Agents
Task-specific
AgentsViruses
15(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Applications of Intelligent Agents (1)Applications of Intelligent Agents (1)Applications of Intelligent Agents (1)Applications of Intelligent Agents (1)
E il AE-mail Agents4Beyond Mail, Lotus Notes, Maxims
h d liScheduling Agents4ContactFinder
Desktop Agents4Office 2000 Help, Open Sesame
Web-Browsing Assistants4WebWatcher, Letizia
Information Filtering Agents4Amalthaea, Jester, InfoFinders, Remembrance agent,
PHOAKS Sit S
16(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
PHOAKS, SiteSeer
Applications of Intelligent Agents (2)Applications of Intelligent Agents (2)Applications of Intelligent Agents (2)Applications of Intelligent Agents (2)
N i ANews-service Agents4NewsHound, GroupLens, FireFly, Fab, ReferralWeb,
NewTNewT
Comparison Shopping Agents4Mysimon BargainFinder Bazzar Shopbor Fido4Mysimon, BargainFinder, Bazzar, Shopbor, Fido
Brokering Agents4P lL i B K b h J Y t4PersonalLogic, Barnes, Kasbah, Jango, Yenta
Auction Agents4A ti B t A ti W b4AuctionBot, AuctionWeb
Negotiation Agents4D t D t t T@T
17(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
4DataDetector, T@T
Computers Meet BiosciencesComputers Meet BiosciencesComputers Meet BiosciencesComputers Meet Biosciences
BT IT
Bioinformation Technology
(BIT)
18(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
( )
AI in Life SciencesAI in Life SciencesAI in Life SciencesAI in Life Sciences
S l iS l iSequence analysisSequence analysis4 Sequence alignment4 Structure and function prediction
Structure analysisStructure analysis4 Protein structure comparison
4 Gene finding
p4 Protein structure prediction 4 RNA structure modeling
Expression analysisExpression analysis4 Gene expression analysis4 Gene clustering
Pathway analysisPathway analysis
4 Gene clustering
19(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
4Metabolic pathway4 Regulatory networks
Bi l i l A li tiBi l i l A li tiBiological ApplicationBiological Application
&
120 l f
&
120 samples from60 leukemia patients
Gene expression data Class: ALL/AML
Di iTraining with
20(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Diagnosis
[Cheok et al., Nature Genetics, 2003]
6-fold validation
Evolutionary ComputationEvolutionary Computation: : y py pNature as ComputerNature as Computer
“Owing to this struggle for life, any variation, however slight and from whatever cause proceeding, if it be in any degree profitable to an individual of any species, in its infinitely complex relations to other organic beings and to external nature, will tend to the preservation of that individual, and will generally be inherited by its offspring.” p g
Origin of Species “Charles Darwin (1859)”
21(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Genetic AlgorithmsGenetic AlgorithmsGenetic AlgorithmsGenetic Algorithms110010 1010
crossovercrossover
solutions
1100101010
1011101110 110010 1110
101110 1110
encoding
chromosomes
solutions0011011001
1100110001
mutationmutation
101110 1010
00110 10011
00110 10010
evaluationevaluationselectionselectionnew
population 00110 100101100101110
1011101010
0011001001
roulette
solutions
fitness
roulettewheel
22(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
fitnesscomputation
Hot Water Flashing Nozzle (1)Hot Water Flashing Nozzle (1)Application Example 1Application Example 1
Hot Water Flashing Nozzle (1)Hot Water Flashing Nozzle (1)Hans-Paul Schwefel
Start
Hot water entering Steam and droplet at exit
performed the original experiments
Hot water entering Steam and droplet at exit
At throat: Mach 1 and onset of flashingg
23(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Concrete Shell RoofConcrete Shell RoofApplication Example 3Application Example 3
Concrete Shell RoofConcrete Shell Roof
under own and outer load (snow and wind)
Optimal shape
Height 1.34m
Spherical shapep p
Half span 5.00m
S i 36% h ll thi ki
24(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Savings : 36% shell thickness
27% armation
max min
Orthogonal bending strength
→ϕm
Cooperating Robots (3)Cooperating Robots (3)Application Example 13Application Example 13
Cooperating Robots (3)Cooperating Robots (3)
Cooperating Autonomous Robots
25(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
CC l i S S ftb t (1)l i S S ftb t (1)Application Example 14Application Example 14
CoCo--evolving Soccer Softbots (1)evolving Soccer Softbots (1)
CoCo evolvingevolvingCoCo--evolvingevolvingSoccer Softbots Soccer Softbots With Genetic With Genetic P iP i
At R b C th t "l " th " l" b t
ProgrammingProgramming
At RoboCup there are two "leagues": the "real" robot league and the "virtual" softbot league
How do you do this with GP?How do you do this with GP?4GP breeding strategies: homogeneous and heterogeneous
4Decision of the basic set of function with which to evolve players
26(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Decision of the basic set of function with which to evolve players
4Creation of an evaluation environment for our GP individuals
CoCo evolving Soccer Softbots (2)evolving Soccer Softbots (2)Application Example 14Application Example 14
CoCo--evolving Soccer Softbots (2)evolving Soccer Softbots (2)
Initial Random Population
27(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
CoCo evolving Soccer Softbots (3)evolving Soccer Softbots (3)Application Example 14Application Example 14
CoCo--evolving Soccer Softbots (3)evolving Soccer Softbots (3)
Kiddie Soccer
28(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
CoCo evolving Soccer Softbots (4)evolving Soccer Softbots (4)Application Example 14Application Example 14
CoCo--evolving Soccer Softbots (4)evolving Soccer Softbots (4)
Learning to Block the Goal
29(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
CoCo evolving Soccer Softbots (5)evolving Soccer Softbots (5)Application Example 14Application Example 14
CoCo--evolving Soccer Softbots (5)evolving Soccer Softbots (5)
Becoming Territorial
30(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
10001000--Pentium BeowulfPentium Beowulf--Style Style yyCluster Computer for Parallel GPCluster Computer for Parallel GP
31(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Computing Power and Memory Capacity of Computing Power and Memory Capacity of p g y p yp g y p yComputers and Organisms [Moravec, 1988]Computers and Organisms [Moravec, 1988]
32(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Von Neumann’s Von Neumann’s The Computer The Computer ppand the Brain (1958)and the Brain (1958)
33(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
John von Neumann (1903-1957)
The Computer and the BrainThe Computer and the BrainThe Computer and the BrainThe Computer and the Brain
- 10 billion neurons(100 trillion synapses)
- Less than 1 million processors (10 –9 sec each, neuron: 10-3 sec) ( y p )
- Distributed processing- Nonlinear processing- Parallel processing
(10 sec each, neuron: 10 sec)- Central processing- Arithmetic operation (linearity)
Sequential processing
34(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
- Parallel processing- Carbon-based (wet)
- Sequential processing- Silicon-based (dry)
From Biological Neurons to From Biological Neurons to ggArtificial NeuronsArtificial Neurons
35(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Multilayer Perceptron (MLP)Multilayer Perceptron (MLP)Error Backpropagation
Multilayer Perceptron (MLP)Multilayer Perceptron (MLP)
E∂Output Comparison
Information Propagation ∑ −≡ kkd otwE 2)(1
)(
iiiii w
Ewwww
∂∂
−=ΔΔ+← η ,
p g
Input x1Weights
∑∈
≡outputsk
kkd otwE )(2
)(
Input x2 Outputx )(xfo =
Input x3
Activation FunctionScaling Function
Input Layer Hidden Layer Output LayerActivation Function
36(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Application Example:Application Example:pp ppp pAutonomous Land Vehicle (ALV)Autonomous Land Vehicle (ALV)
NN learns to steer an autonomous vehicle.
960 input units, 4 hidden units, 30 output units p p
Driving at speeds up to 70 miles per hour
ALVINN System
Weight values
Image of aforward -mounted
for one of the hidden units
mountedcamera
37(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Neural Nets for Face RecognitionNeural Nets for Face Recognition
960 x 3 x 4 network is trained on gray-level images of faces to predict
38(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
960 x 3 x 4 network is trained on gray level images of faces to predict whether a person is looking to their left, right, ahead, or up.
New Directions: New Directions: BiointelligenceBiointelligence
Humans and ComputersHumans and ComputersHumans and ComputersHumans and Computers
The Entire Problem Space
What Kind ofHuman Computers What Kind of
Computers?
Human Computers
Current Computers
40(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Humans and MachinesHumans and MachinesHumans and MachinesHumans and Machines
Humans are 4creative
Humans are ♦ imprecise, creative,
4compliant,
4attentive to change
p ,
♦ sloppy,
♦ distractable, attentive to change,
4resourceful, and
4multipurpose
,
♦ emotional, and
♦ illogical4multipurpose
To achieve human-level intelligence these
♦ illogical
To achieve human-level intelligence these
properties should be taken into account.
© 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/41
Toward HumanToward Human Level IntelligenceLevel IntelligenceToward HumanToward Human--Level Intelligence Level Intelligence
Human intelligence develops situated in a multimodal environment [Gibbs, 2005].[ , ]The human mind makes use of multiple representations and problem-solving strategies [Fuster, 2003]. ]The brain consists of functional modules which are localized in subcortical areas but work togetheron the whole-brain scale [Grillner et o e w o e b sc e [G e eal., 2006]. Humans can integrate the multiple tasks into a coherent solution [Jones, 2004].2004]. Humans are versatile and come up with many new ideas and solutions to a given problem [Minsky, 2006].
© 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/42
What is What is the information the information i i i li i i l d l id l iprocessing principleprocessing principle underlying underlying
human intelligence?human intelligence?human intelligence?human intelligence?
43(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Mind Brain Cell MoleculeMind Brain Cell MoleculeMind, Brain, Cell, MoleculeMind, Brain, Cell, MoleculeMind
Mind
Mind
Brain
C llCell
M l l∞ Molecule
1011 cells
∞ memory
0 ce s
1010 mol.
44(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Molecular Basis of Learning and Memory Molecular Basis of Learning and Memory g yg yin the Brainin the Brain
45(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Principles of Learning: Modern ConceptsPrinciples of Learning: Modern ConceptsPrinciples of Learning: Modern ConceptsPrinciples of Learning: Modern Concepts
Types of learning: Accretion, tuning, restructuring (e grestructuring (e.g., Rumelhart & Norman, 1976)Encoding specificityEncoding specificity principle (Tulving, 1970’s)Cellular and molecular basis of learning andbasis of learning and memory (Kandel et al., 1990’s) Conceptual blend andConceptual blend and chemical scramble (e.g., Feldman, 2006)
© 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
46
Principles of Information Processing in thePrinciples of Information Processing in thePrinciples of Information Processing in the Principles of Information Processing in the BrainBrain
The Principle of Uncertainty4Precision vs. predictionp
The Principle of Nonseparability “UN-IBM”4Processor vs. memoryy
The Principle of Infinity4Limited matter vs. unbounded memory
The Principle of “Big Numbers Count”4Hyperinteraction of 1011 neurons (or > 1017 molecules)
The Principle of “Matter Matters”4Material basis of “consciousness” [Zhang, 2005]
47(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Unconventional ComputingUnconventional ComputingUnconventional ComputingUnconventional Computing
Quantum Computing4Atoms
4Superposition, quantum entanglements
Chemical ComputingChemical Computing 4Chemicals
4Reaction diffusion computing4Reaction-diffusion computing
Molecular Computing4M l l4Molecules
4“Self-organizing hardware”
48(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Molecular Computing: Molecular Computing: p gp gBioMolecules as ComputerBioMolecules as Computer
49(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
011001101010001 ATGCTCGAAGCT
DNA Molecules for Information DNA Molecules for Information Storage and ProcessingStorage and Processing
Writing: make DNA sequences
DNA
a t c g g t c a t ag c a c t
0 0 0DNAmemory
strands1 0 1
t a g c c c g t g a
t c a t a
t a g c c c g t g a
a t c g g t c a t a
Reading: hybridization and readout
50(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
g y
x1=1
x2=0
x3=0
x4=1
x5=0
x6=0
x7 =0
x8=0
x9=0
x10=1
x11=0
x12=1
x13=0
x14=0
x15=0
y
= 1
x1=0
x2=1
x3=1
x4=0
x5=0
x6=0
x7 =0
x8=0
x9=1
x10=0
x11=0
x12=0
x13=0
x14=1
x15=0
y
= 0
1
2
Learningx1=0
x2=0
x3=1
x4=0
x5=0
x6=1
x7 =0
x8=1
x9=0
x10=0
x11=0
x12=0
x13=1
x14=0
x15=0
y
=14 Data Items3
x1=0
x2=0
x3=0
x4=0
x5=0
x6=0
x7 =0
x8=1
x9=0
x10=0
x11=1
x12=0
x13=0
x14=0
x15=1
y
=14
x1x2 x15
x4 x10 y=1x1
Round 1Round 2Round 3x3 x14
4 10 y1
x4 x12 y=1x1
x10 x12 y=1x4
1
x12
x4 x13
x3 x9 y=0x2
x3 x14 y=0x22x12
x5x9 x14 y=0x3
x6 x8 y=1x3
x6 x11x6 x13 y=1x3
x8 x13 y=1x6
3
© 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
51x8 x9
x7x10x11 x15 y=0x84
Hypernetwork of DNA Molecules
[Zhang, DNA-2006]
““학습학습 추론”하는추론”하는 DNADNA 컴퓨터컴퓨터““학습학습 추론”하는추론”하는 DNA DNA 컴퓨터컴퓨터
MP4.avi
© 2007, SNU Biointelligence Lab, http://bi.snu.ac.kr/53
M l l C Sili CM l l C Sili CMolecular Computers vs. Silicon ComputersMolecular Computers vs. Silicon Computers
Molecular Computers Silicon Computers
Processing Ballistic Hardwired
Medium Liquid (wet) or Gaseous (dry) Solid (dry)
Communication 3D collision 2D switchingCommunication 3D collision 2D switching
Configuration Amorphous (asynchronous) Fixed (synchronous)
Parallelism Massively parallel Sequentialy p q
Speed Fast (millisec) Ultra-fast (nanosec)
Reliability Low High
Density Ultrahigh Very high
Reproducibility Probabilistic Deterministic
54(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
B h k P bl Di it IB h k P bl Di it IBenchmark Problem: Digit ImagesBenchmark Problem: Digit Images
• 8x8=64 bit images (made from 64x64 scanned gray images)• Training set: 3823 images• Test set: 1797 images
55(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
• Test set: 1797 images
Pattern Information ProcessingPattern Information ProcessingPattern Information ProcessingPattern Information Processing
0.7
0.8
0.9
1
ate
0 2
0.3
0.4
0.5
0.6
Cla
ssifi
catio
n ra
Order 1
Order 2
0
0.1
0.2
1 7 13 19 25 31 37 43 49 55 61 67 73 79 85 91 97
epoch
56(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Cognitive Learning and MemoryCognitive Learning and Memory
Toward HumanToward Human--Level Machine Learning: Level Machine Learning: M l i d l M G (MMG)M l i d l M G (MMG)Multimodal Memory Game (MMG)Multimodal Memory Game (MMG)
But, I'm getting married tomorrowBut, I'm getting married tomorrowBut, I'm getting married tomorrowBut, I'm getting married tomorrowBut, I'm getting married tomorrowBut, I'm getting married tomorrow
Well, maybe I am...
I keep thinking about you.
And I'm wondering if we made a mistake giving up so fast.
Well, maybe I am...
I keep thinking about you.
And I'm wondering if we made a mistake giving up so fast.But, I'm getting married tomorrow
W ll b I
But, I'm getting married tomorrow
W ll b I
But, I'm getting married tomorrow
W ll b I
But, I'm getting married tomorrow
W ll b I
Well, maybe I am...
I keep thinking about you.
And I'm wondering if we made a mistake giving up so fast.
Well, maybe I am...
I keep thinking about you.
And I'm wondering if we made a mistake giving up so fast.
Well, maybe I am...
I keep thinking about you.
And I'm wondering if we made a mistake giving up so fast.
Well, maybe I am...
I keep thinking about you.
And I'm wondering if we made a mistake giving up so fast.g g g p
Are you thinking about me?
But if you are, call me tonight.
g g g p
Are you thinking about me?
But if you are, call me tonight.
Well, maybe I am...
I keep thinking about you.
And I'm wondering if we made a mistake giving up so fast.
A thi ki b t ?
Well, maybe I am...
I keep thinking about you.
And I'm wondering if we made a mistake giving up so fast.
A thi ki b t ?
Well, maybe I am...
I keep thinking about you.
And I'm wondering if we made a mistake giving up so fast.
A thi ki b t ?
Well, maybe I am...
I keep thinking about you.
And I'm wondering if we made a mistake giving up so fast.
A thi ki b t ?
g g g p
Are you thinking about me?
But if you are, call me tonight.
g g g p
Are you thinking about me?
But if you are, call me tonight.
g g g p
Are you thinking about me?
But if you are, call me tonight.
g g g p
Are you thinking about me?
But if you are, call me tonight.
Are you thinking about me?
But if you are, call me tonight.
Are you thinking about me?
But if you are, call me tonight.
Are you thinking about me?
But if you are, call me tonight.
Are you thinking about me?
But if you are, call me tonight.
Image Sound Text
Text HintHint Image
© 2007, SNU Biointelligence Lab, http://bi.snu.ac.kr/Image-to-Text Generator Text-to-Image GeneratorMachine Learner
Text Generation Game (from Image)Text Generation Game (from Image)Text Generation Game (from Image)Text Generation Game (from Image)
Image SoundSound Text
LearningLearning
by ViewingT
I2T T2IGame
Manager
Text HintT
© 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
59
g
Image Generation Game (from Text)Image Generation Game (from Text)Image Generation Game (from Text)Image Generation Game (from Text)
TextImage SoundSound
LearningLearning
by ViewingI
I2T T2IGame
Manager
Hint ImageI
© 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
60
g
Three ExperimentsThree ExperimentsThree ExperimentsThree Experiments
S G iSentence Generation4Learn: a linguistic recall memory from a sentence corpus4Given: a partial or corrupt sentencep p4Generate: a complete sentence
Image-to-Text Translation4 i j i d l f i i4Learn: an image-text joint model from an image-text pair corpus4Given: an image (scene)4Generate: a text (dialogue of the scene)( g )
Text-to-Image Translation4Learn: an image-text joint model from an image-text pair corpus4Gi (di l )4Given: a text (dialogue)4Generate: an image (scene of the dialogue)
© 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
61
E i 1 L i Li i i ME i 1 L i Li i i MExperiment 1: Learning Linguistic MemoryExperiment 1: Learning Linguistic Memory
Dataset: scripts from dramas4Friends44House4244Grey AnatomyGrey Anatomy 4Gilmore Girls 4Sex and the City
Training data: 289,468 sentences Test data: 700 sentences with bl kblanksVocabulary size: 34,219 words
© 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
62
S t C l ti R ltS t C l ti R ltSentence Completion ResultsSentence Completion Results
? gonna ? upstairs ? ? a shower I'm gonna go upstairs and take a shower
We ? ? a lot ? giftsWe don't have a lot of gifts
? have ? visit the ? roomI have to visit the ladies' room
? ? don't need your ?if I don't need your help
? still ? believe ? did thisI still can't believe you did this
? ? a dream about ? In ?I had a dream about you in Copenhagen
? i t it if ? ll h b ? ?
? ? ? decisionto make a decision
I still can't believe you did this
What ? ? ? hereWhat are you doing here
? ? fi ? f di l h l
I had a dream about you in Copenhagen
? appreciate it if ? call her by ? ? I appreciate it if you call her by the way
I'm standing ? the ? ? ? cafeteria I' t di i th f th f t i
Would you ? to meet ? ? Tuesday ? W ld i t t i T d d
? you ? first ? of medical schoolAre you go first day of medical school
Why ? you ? come ? down ? Why are you go come on down here
? think ? I ? met ? somewhere beforeI think but I am met him somewhere before
I'm standing in the one of the cafeteriaWould you nice to meet you in Tuesday and
© 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
63
Experiments 2 & 3: CrossmodalExperiments 2 & 3: CrossmodalExperiments 2 & 3: Crossmodal Experiments 2 & 3: Crossmodal TranslationTranslation
D t t d di
The order (k) of hyperedge♦ Text: Order 2~4
Dataset: scenes and corresponding scripts from two dramas4 Friends
♦ Image: Order 10~340
The method of creating hyperedges from training data
4 Prison Break
Training data: 2,808 scenes and scripts
yp g g♦ Text: Sequential sampling from a
randomly selected position♦ Image: Random sampling in 4,800 p
Scene (image) size: 80 x 60 = 4800 binary pixels
V b l i 2 579 d
g p gpixel positions
Number of samples from an image-text pair
Vocabulary size: 2,579 words
Where am I giving birth
g p♦ From 150 to 300
Where am I giving birth
I know it's been really hard for you
So when you guys get in there
© 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
64
So when you guys get in there
ImageImage--toto--Text Translation ResultsText Translation Results
Matching &
ImageImage toto Text Translation ResultsText Translation Results
AnswerQuery
I don't know
g
Completion
I don't know what happeneddon't know what
know what happened
There's a kitty in my guitar case
There's a
a kitty in case…
in my guitar case
Maybe there's something I can do to make sure I get pregnant
Maybe there's something
there's something I
… pregnantI get pregnant
© 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
TextText--toto--Image Translation ResultsImage Translation Results
Matching &
TextText toto Image Translation ResultsImage Translation Results
Query Matching &
CompletionAnswer
I don't know what happened
Take a look at this
There's a kitty in my guitar case
Maybe there's something I can d t k I t tdo to make sure I get pregnant
© 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
Toward HumanToward Human--Level IntelligenceLevel Intelligence
From Mind to Molecules and BackFrom Mind to Molecules and BackFrom Mind to Molecules and BackFrom Mind to Molecules and BackMind
BrainBrain
Cell
Molecule∞ memory
Molecule
1011 cells
>103 molecules
© 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
68
P di f C t ti l I t lliP di f C t ti l I t lliParadigms for Computational IntelligenceParadigms for Computational Intelligence
Symbolism Connectionism DynamicismHyperinter-actionism
Metaphorsymbol
tneural
tdynamical system
biomoleculart
psystem system
y ysystem
Mechanism logical electrical mechanical chemical
Description syntactic functional behavioral relational
Representation localist distributed continuous collective
Organization structural connectionist differential combinatorial
Adaptation substitution tuning rate change self-assembly
Processing sequential parallel dynamical massively parallel
Structure procedure network equation hypergraphStructure procedure network equation hypergraph
Mathematicslogic, formal
languagelinear algebra,
statisticsgeometry, calculus
graph theory,probabilistic logic
S /ti f l ti l t l ti t l
© 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
69
Space/time formal spatial temporal spatiotemporal
[Zhang, IEEE Comp. Intel. Mag., August 2008]
Da Vinci’s Dream of Flying MachinesDa Vinci’s Dream of Flying MachinesDa Vinci s Dream of Flying MachinesDa Vinci s Dream of Flying Machines
70(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Engines of FlightEngines of FlightEngines of FlightEngines of Flight
R k E i
71(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Piston Engine Jet Engine Rocket Engine
T i ’ D f I t lli t M hiT i ’ D f I t lli t M hiTuring’s Dream of Intelligent MachinesTuring’s Dream of Intelligent Machines
Alan Turing
(1912-1954)
72(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Computing Machinery and Intelligence (1950)
C t d I t lliC t d I t lliComputers and IntelligenceComputers and Intelligence
73(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Future Technology EnablersFuture Technology EnablersFuture Technology EnablersFuture Technology Enablers
Bio-electric computers
True neural computing
Quantum computer, molecular electronics
computers1e6-1e7 x lower power for lifetime batteries
Full motion
Smart lab-on-chip, plastic/printed ICs, self-assembly
mobile video/office
Metal gates Pervasive voice
Vertical/3D CMOS, Micro-wireless nets, Integrated optics
yWearable communications, wireless remote medicine, ‘hardware over internet’ !
Metal gates, Hi-k/metal oxides, Lo-k with Cu, SOI
recognition, “smart” transportation
Integrated optics
74(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/Source: Motorola, Inc, 2000
Now +2 +4 +6 +8 +10 +12
Artificial Intelligence (AI)Artificial Intelligence (AI)g ( )g ( )
S b li AISymbolic AI Rule-Based Systems
Connectionist AI Neural Networks
Evolutionary AI Genetic Algorithms
Molecular AI:
75(c) 2000-2007 SNU CSE Biointelligence Lab, http://bi.snu.ac.kr/
Molecular AI: DNA Computing