Lecture1 AI1 Introduction to artificial intelligence

Introduction ArtificialIntroduction Artificial IntelligenceIntelligence

Lecture 1

Albert Orriols i Puightt // lb t i l thttp://www.albertorriols.net

[email protected]

Artificial Intelligence – Machine Learningg gEnginyeria i Arquitectura La Salle

Universitat Ramon Llull

Today’s Agenda

Brainstorming from your “postits”g y pSome DefinitionsPrehistory and History of AIWhere are we headed?

Slide 2Artificial Intelligence Machine Learning

BrainstormingWhat’s AI?

A

AA

…

Do you know of some real-world applications?Do you know of some real-world applications?A

A

…


What’s Intelligence?Intelligence (dictionary)g ( y)

capacity for learning, reasoning, understanding, and similar forms of mental activity; aptitude in grasping truths, o s o e a ac y; ap ude g asp g u s,relationships, facts, meanings, etc.

In particular, we could say:pa cu a , e cou d sayAbility to act as human beings

Solve problemsThink rationally

Artificial intelligence … Building a machine that is (or seems to be at the eyes of the beholder) intelligent


Can You Be More Formal?What is artificial intelligence? g

It is the science and engineering of making intelligent machines, especially intelligent computer programs. It is related to the similar task of using computers to understand human intelligence, but AI does not have to confine itself to methods that are biologically observable.o co e se o e ods a a e b o og ca y obse ab e

Yes, but what is intelligence? I t lli i th t ti l t f th bilit t hi l iIntelligence is the computational part of the ability to achieve goals in the world. Varying kinds and degrees of intelligence occur in people, many animals and some machines.

Isn't there a solid definition of intelligence that doesn't depend on relating it to human intelligence?

Not yet. The problem is that we cannot yet characterize in general what kinds of computational procedures we want to call intelligent. We

d t d f th h i f i t lli d t th

Slide 5

understand some of the mechanisms of intelligence and not others. See the complete interview at: http://www-formal.stanford.edu/jmc/whatisai/node1.html

Artificial Intelligence Machine Learning

What’s Involved in Intelligence?Ability to interact with the real world

to perceive, understand, and acte.g., speech recognition and understanding

Searching the best solution

Reasoning and PlanningReasoning and Planningmodeling the external world, given input

solving new problems, planning, and making decisions

ability to deal with unexpected problems, uncertainties

Learning and Adaptationwe are continuously learning and adaptingy g p g

our internal models are always being “updated”e.g., a baby learning to categorize and recognize animals

Slide 6

g , y g g g


AI Is Not Alone at HomeCrossbreeding of a lot of fieldsg

Philosophy Logic, methods of reasoning, mind as physical system, foundations of learning language rationalityfoundations of learning, language, rationality.

Mathematics Formal representation and proof, algorithms, computation, (un)decidability, (in)tractability

Statistics Modeling uncertainty, learning from data

Economics Utility, decision theory, rational economic agents

Neuroscience Neurons as information processing units

Psychology / NeuroScience

How do people behave, perceive, process cognitive information represent knowledgeScience information, represent knowledge

Computer Engineering Building fast computers

Control Theory Design systems that maximize an objective function y g y jover time

Linguistics Knowledge representation, grammars


Prehistory of AIThrough history, people though of mythic “artificial” g y, p p g yrobots

golden robots of Hephaestus and Pygmalion's Galateagolden robots of Hephaestus and Pygmalion s Galatea

alchemical means of placing mind into matter

More specific, tangible advances5th century B.C.

Aristotle invented syllogistic logic, the first formal deductive reasoning system.

13th century.Talking heads were said to have been created (Roger Bacon

d Alb t th G t)and Albert the Great).Ramon Lull, Spanish theologian, invented machines for discovering nonmathematical truths through combinatory

Slide 8

discovering nonmathematical truths through combinatory.


Prehistory of AIMore specific, tangible advances (cont.)p , g ( )

15th centuryInvention of printing using moveable type Gutenberg BibleInvention of printing using moveable type. Gutenberg Bible printed (1456).

15th-16th century15th 16th centuryClocks, the first modern measuring machines, were first produced using lathes.

16th centuryClockmakers extended their craft to creating mechanicalClockmakers extended their craft to creating mechanical animals and other novelties.



17th century - The revolution of thinking about thinkingDescartes proposed that bodies of animals are nothingDescartes proposed that bodies of animals are nothing more than complex machines (strong AI). Variations and elaborations of Cartesian mechanism.

Hobbes published The Leviathan, containing a material and combinatorial theory of thinking.Pascal created the first mechanical digital calculating machine (1642).

Leibniz improved Pascal's machine to do multiplication & division (1673) and envisioned a universal calculus of reasoning by which

Slide 10

( ) g yarguments could be decided mechanically.



18th century – Mechanical toys

Vaucanson’s Duck Von Kempelen’s phony mechanical chess player



19th century – Frankenstein’s birthGeorge Boole developed a binary algebra representing (some)George Boole developed a binary algebra representing (some) "laws of thought," published in The Laws of Thought.Charles Babbage and Ada Byron (Lady Lovelace) worked on g y ( y )programmable mechanical calculating machines.

Mary Shelley published the story of Frankenstein's monster (1818).Crossing the century bridgeCrossing the century bridge

Behaviorism was expounded by psychologist Edward Lee Thorndike in

Slide 12

"Animal Intelligence."


Pre-birth of AIBeginning of the 20th centuryg g yRussell and Whitehead published Principia Mathematica.

Capek's play “Rossum's Universal Robots” produced in 1921 (LondonCapek s play Rossum s Universal Robots produced in 1921 (London opening, 1923). First use of the word 'robot' in English.

McCulloch and Pitts publish "A Logical Calculus of the Ideas Immanent inMcCulloch and Pitts publish A Logical Calculus of the Ideas Immanent in Nervous Activity" (1943), laying foundations for neural networks.

Rosenblueth, Wiener and Bigelow coin the term cybernetics (1943).

Bush published As We May Think (1945) a prescient vision of the future in which computers assist humans in many activities.


The 3 Key IngredientsThe first key ingredient: The computer and the programy g p p g

ENIAC (1945). The first electronic digital computer

EDVAC (1949) Th fi t t d tEDVAC (1949). The first stored program computer


The 3 Key IngredientsThe second key ingredient: The TURING TEST.y g

(Human) judge communicates with a human and a machine over text-only channel.o e e o y c a e

Both human and machine try to act like a human

J d t i t t ll hi h i hi hJudge tries to tell which is which.

Numerous variants.

Loebner prize.

Current programs nowhere close Cu e t p og a s o e e c oseto passing this

http://www.jabberwacky.com/http://turingtrade.org/


The Turing TestMore on Turing testg

Objective: The machine needs to fool the machine[INT] I heard that a striped rhinoceros flow on the[INT] I heard that a striped rhinoceros flow on the Mississippi in a pink balloon this morning. What do you think about?[COMP] That sound rather ridiculous to me[COMP] That sound rather ridiculous to me[INT] Really? My uncle did this one... Why this sound ridiculous?[COMP] Option 1: Rhinoceros don't have stripes[COMP] Option 1: Rhinoceros don t have stripes[COMP] Option 2: Rhinoceros can't fly

Tr to change ON for UNDER the Mississipi

Is this unfair for the computer?

Try to change ON for UNDER the Mississipi

[INT] What’s the result of 324 x 678?[COMP] This is too difficult. I’m not a calculator!

Slide 16

Needs to seem more foolish than it actually is (has to lie!)


The 3 Key IngredientsThe third key ingredient: THE DARMONT CONFERENCE. y gPeople working on building intelligent machines.

J. McCarthy, M. L. Minsky, N. Rochester, and C.E. Shannon August 31 1955 "We propose that a 2 monthShannon. August 31, 1955. We propose that a 2 month, 10 man study of artificial intelligence be carried out during the summer of 1956 at Dartmouth College induring the summer of 1956 at Dartmouth College in Hanover, New Hampshire. The study is to proceed on the basis of the conjecture that every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to

i l t it "simulate it."


Brief History of AIThe Golden years (1956 – 1974)y ( )

‘1960s Strong funding of AI centersStrong funding of AI centersBuilding intelligent automataSearching in complex search spacesSearching in complex search spaces

First AI programs that workS l’ h k ( hi h l )Samuel’s checker program (which learns)Newell and Simon’s Logic TheoristG l t ’ t iGelernter’s geometry engineRobinson’s complete algorithm for logical reasoning

First programming languages for AIMcCarthy - Lisp (1958)


Brief History of AIThe Golden years (1956 – 1974)y ( )

And the first chatterbots:ELIZA (1966)ELIZA (1966).

It carried out very realistic conversations. It searched for key words in the conversation and asked yinformation about that


Brief History of AIThe Winter: After expansion, there’s always a contractionp , y

First doubts on the feasibility of all the approach

P blProblems:Limited computer power C bi t i l l i ( ti l ti )Combinatorial explosion (exponential time)Commonsense knowledge and reasoningM ’ dMoravec’s paradoxThe Chinese room argument undermined the goal of building intelligent machinesintelligent machinesEND OF FUNDING


Brief History of AIThe Chinese room argument (Searle, 1980)g ( , )

Person who knows English but not Chinese sits in roomC ese s s oo

Receives notes in Chinese

H t ti E li h l b k fHas systematic English rule book for how to write new Chinese characters based on input Chinese characters, returns his notesbased o put C ese c a acte s, etu s s otes

Person=CPU, rule book=AI program, really also need lots of paper (storage)

Has no understanding of what they meanBut from the outside, the room gives perfectly reasonable

i Chi !answers in Chinese!

Searle’s argument: the room has no intelligence in it!


Brief History of AIBut in parallel… expert systems rise and growp p y g

MYCIN(1972): Diagnosed infection blood diseasesDiagnosed infection blood diseases. It had a set of about 600 rules and started to ask questions.In some cases better than human expertsIn some cases, better than human experts.

XCON (1980): P d ti l b d t th t i t d th d i fProduction-rule-based system that assisted the ordering of a type of computers systems by automatically selecting the computer systems components based on the customers requirements.Saving $40 million dollars to the company. 2500 rules and processed 80000 orders with 95%-98% accuracy. The gain in money was because it reduced the need to give free components when the technicians made errors by speeding

Slide 22

components when the technicians made errors, by speeding the assembly process and by increasing customer satisfaction


Brief History of AIBut in parallel… expert systems rise and growp p y g

PROSPECTOR (1981) A computer-based consultation system for mineralA computer-based consultation system for mineral exploration. Recommending exploratory drillingg p y g

And many others. Search the web for more!

New funding due to this successNew funding due to this successAI groups were formed in many large companies to develop

t texpert systems.

1986 sales of AI-based hardware and software were $425 illimillion.


Brief History of AIQuick pace in the ‘90sQ p

NCSA releases the first web browser, Mosaic

D Bl b t G KDeep Blue beats Gary Kasparov

Robotic soccer players in RoboCup

Sony corporation introduced the robotic dog AIBO

Remote agent autonomously drive deep space 1e ote age t auto o ous y d e deep space

Even moving faster in the 00’siRobot introduces the vacuum cleaning robot Roomba

DARPA grand challenge (we’ll see it in a minute)A Touareg R5 won the challenge


Some Cool ApplicationsThree cool applications among hundredspp g

Deep Blue

DARPA G d Ch llDARPA Grand Challenge

Robotics Cog

Loebner Prize

Roombaoo ba

Rob-Cup

ASIMOASIMO

Data miningStock MarketMedical Diagnosis


Deep Blue

Origins at CMU

It was a massively parallel, RS/6000 SP Thin P2SC-based system with 30-nodes

Deep Blue took Gary Kasparovto the cleaners


DARPA Grand ChallengeGrand Challenge

Cash prizes ($1 to $2 million) offered to first robots to complete a long course completely unassistedStimulates research in vision robotics planning machineStimulates research in vision, robotics, planning, machine learning, reasoning, etc

2004 Grand Challenge:2004 Grand Challenge: 150 mile route in Nevada desertFurthest any robot went was about 7 milesFurthest any robot went was about 7 miles … but hardest terrain was at the beginning of the course

2005 G d Ch ll2005 Grand Challenge:132 mile raceN t l i di t i tNarrow tunnels, winding mountain passes, etcStanford 1st, CMU 2nd, both finished in about 6 hours


DARPA Grand Challengehttp://cs.stanford.edu/group/roadrunner/p g p


DARPA Grand ChallengeThe challenge: a driverless car competes for wining the g p grace

150 mile off-road robot race across the Mojave desertNatural and manmade hazardsN d i t t l

150 mile off-road robot race across the Mojave desertNatural and manmade hazardsN d i t t lNo driver, no remote controlNo dynamic passingFastest vehicle wins the race (and 2 million dollar prize)

No driver, no remote controlNo dynamic passingFastest vehicle wins the race (and 2 million dollar prize)


(and 2 million dollar prize)(and 2 million dollar prize)

DARPA Grand ChallengeThe architecture


Robotics - CogHumanoid intelligence requires humanoid interactions g qwith the world


Loebner PrizePrizes the chatterbots considered to be the most human-like

Th t t b i 1990The contest begun in 1990

$25,000 is offered for the first chatterbot that judges cannot j gdistinguish from a real human and that can convince judges that the human is the computer program

$100,000 is the reward for the first chatterbot thatfor the first chatterbot that judges cannot distinguish from a real human in a Turing test that includesTuring test that includes deciphering and understanding text, visual, and auditory inputand auditory input


RoombaGo around “smartly” to clean up a housey p


RobCupFirst official Rob-Cup soccer match (1997)p ( )


ASIMO

Advanced Step in InnovativeMobilityy

Able ofMovingMovingInteracting with human beingsHelp peopleHelp people


Data Mining ExplosionData mining: Extract novel, useful, and interesting g , , ginformation from data

Why so a big deal?Companies are generating lots of data about the business

They want to process these data and obtain useful information

Wh no not before?Why now, not before?Computers have a lot of power nowadays


Modeling the Stock MarketModeling market tradersg

LETS project: Evolving artificial traders for successful market trading (Sonia Schulenburg et al, 2007)ad g (So a Sc u e bu g et a , 00 )

Evolutionary economics:Evolutionary economics:Create trend followersand value investors

Let them interact

Evolve a population ofstrategies


Medical DiagnosisData mining

An important application domain of artificial intelligence

John H. HolmesEpidemiologic study by means of LCSsHidden relationships among variables discovered by LCSs

Xavier Llorà et al.Better than Human Capability in Diagnosing Prostate Cancer Using Infrared SpectroscopicProstate Cancer Using Infrared Spectroscopic imaging


But… Slow it down!

There are no castles in the sky

All these applications rely on:Search & Optimization

Knowledge representation

LearningLearning

Planning

These are the four topics that we’ll see in this course. And we will start for the beginning


Detailed Outline AI12. Solving search problems

1. Introduction to search problems

2. Blind search

3. Informed/heuristic search

4. Adversary search (first project)4. Adversary search (first project)

5. Constraint satisfaction problems

3 Knowledge representation3. Knowledge representation

1. Introduction to knowledge representation

2 Knowledge representation based on logics2. Knowledge representation based on logics

3. Knowledge and uncertainty

F L i4. Fuzzy Logics

4. Lisp


Detailed Outline AI25. Machine learning

1. Introduction to machine learning

2. Supervised learning

1. Decision trees, Instance-based learning, Bayesian decision theory, Support vector machines and Neural networks

3 Unsupervised learning – association rules3. Unsupervised learning association rules

4. Unsupervised learning – clustering

5. Reinforcement learning g

6. New challenges in data mining

6. Planning

1. Introduction to planningp g

2. STRIPS

3. Search through the state world

Slide 41

g

4. Search through the plan world


Introduction ArtificialIntroduction Artificial IntelligenceIntelligence

Lecture 1

Albert Orriols i Puightt // lb t i l thttp://www.albertorriols.net

[email protected]

Artificial Intelligence – Machine Learningg gEnginyeria i Arquitectura La Salle

Universitat Ramon Llull

Lecture1 AI1 Introduction to artificial intelligence

Education