Download - Richard W. Hamming
Richard W. HammingRichard W. HammingRichard W. HammingRichard W. Hamming
Learning to LearnLearning to LearnThe Art of Doing Science and EngineeringThe Art of Doing Science and Engineering
Session 7: Session 7: Artificial Intelligence IIArtificial Intelligence II
Topic OutlineTopic OutlineTopic OutlineTopic Outline
Machines Complementing MankindMachines Complementing Mankind
What is Thinking?What is Thinking?
Machines and OriginalityMachines and Originality
Machines Competing with MankindMachines Competing with Mankind
AI ApplicationsAI Applications
Fundamental QuestionsFundamental Questions
Machines Complementing MankindMachines Complementing Mankind
What is Thinking?What is Thinking?
Machines and OriginalityMachines and Originality
Machines Competing with MankindMachines Competing with Mankind
AI ApplicationsAI Applications
Fundamental QuestionsFundamental Questions
FocusFocusFocusFocus
Hamming is more concerned with the aid Hamming is more concerned with the aid computers can give us in computers can give us in intellectualintellectual areas areas rather than rather than mechanicalmechanical areas areas
• Manufacturing: better, preferable, cheaper productsManufacturing: better, preferable, cheaper products
• Essential in space flights, aircraft controlEssential in space flights, aircraft control
AI complementary to AI complementary to roboticsrobotics: intellect : intellect versus physicsversus physics
Hamming is more concerned with the aid Hamming is more concerned with the aid computers can give us in computers can give us in intellectualintellectual areas areas rather than rather than mechanicalmechanical areas areas
• Manufacturing: better, preferable, cheaper productsManufacturing: better, preferable, cheaper products
• Essential in space flights, aircraft controlEssential in space flights, aircraft control
AI complementary to AI complementary to roboticsrobotics: intellect : intellect versus physicsversus physics
CommonalitiesCommonalitiesCommonalitiesCommonalities
ComponentsComponents MachinesMachines HumansHumans
Built from Atoms Built from Atoms and Moleculesand Molecules
YesYes YesYes
Basic PartsBasic Parts State devices for State devices for storage and gatesstorage and gates
CellsCells
Larger StructuresLarger Structures Arithmetic unitsArithmetic units
StorageStorage
ControlControl
I/OI/O
Wiring, sensorsWiring, sensors
BonesBones
MusclesMuscles
OrgansOrgans
Blood vesselsBlood vessels
Nervous SystemNervous System
Key IdeasKey IdeasKey IdeasKey Ideas
From large organizations new effects can From large organizations new effects can arisearise
When we engineer some device to do some When we engineer some device to do some thing Nature does, we do it differentlything Nature does, we do it differently
• AircraftAircraft
• WheelsWheels
• Nervous system versus computer signalsNervous system versus computer signals
From large organizations new effects can From large organizations new effects can arisearise
When we engineer some device to do some When we engineer some device to do some thing Nature does, we do it differentlything Nature does, we do it differently
• AircraftAircraft
• WheelsWheels
• Nervous system versus computer signalsNervous system versus computer signals
ThinkingThinkingThinkingThinking
Something the human brain can do.Something the human brain can do.
Is the failure of computers to think based on Is the failure of computers to think based on size, speed, etc?size, speed, etc?
Is thinking a new effect from enough small Is thinking a new effect from enough small parts – thinking from non-thinking parts?parts – thinking from non-thinking parts?
We do not know what thinking really is!We do not know what thinking really is!
Something the human brain can do.Something the human brain can do.
Is the failure of computers to think based on Is the failure of computers to think based on size, speed, etc?size, speed, etc?
Is thinking a new effect from enough small Is thinking a new effect from enough small parts – thinking from non-thinking parts?parts – thinking from non-thinking parts?
We do not know what thinking really is!We do not know what thinking really is!
Geometry Theorem ProvingGeometry Theorem ProvingGeometry Theorem ProvingGeometry Theorem Proving
Problem: If two sides of a triangle are equal Problem: If two sides of a triangle are equal then the base angles are also equal.then the base angles are also equal.
Problem: If two sides of a triangle are equal Problem: If two sides of a triangle are equal then the base angles are also equal.then the base angles are also equal.
A CC
BB
A
Obtain two congruent triangles, hencecorresponding angles are equal.
Machine ProofMachine ProofMachine ProofMachine Proof
Compared triangle ABC with triangle CBA to Compared triangle ABC with triangle CBA to show selfcongruence, hence equal show selfcongruence, hence equal angles!angles!
Compared triangle ABC with triangle CBA to Compared triangle ABC with triangle CBA to show selfcongruence, hence equal show selfcongruence, hence equal angles!angles!
A C
B
Did the program show “originality”??
AC
B
Separating Us from MachinesSeparating Us from MachinesSeparating Us from MachinesSeparating Us from Machines
Programmers gave the machine instructions Programmers gave the machine instructions so the ability was programmed inso the ability was programmed in
Any different than when we were taught Any different than when we were taught geometry by a teacher?geometry by a teacher?
Did Samuel’s checkers playing program Did Samuel’s checkers playing program show originality when it made surprising show originality when it made surprising moves and defeated the State Checkers moves and defeated the State Checkers Champion?Champion?
What is the test we will use to separate us What is the test we will use to separate us from a computer program?from a computer program?
Programmers gave the machine instructions Programmers gave the machine instructions so the ability was programmed inso the ability was programmed in
Any different than when we were taught Any different than when we were taught geometry by a teacher?geometry by a teacher?
Did Samuel’s checkers playing program Did Samuel’s checkers playing program show originality when it made surprising show originality when it made surprising moves and defeated the State Checkers moves and defeated the State Checkers Champion?Champion?
What is the test we will use to separate us What is the test we will use to separate us from a computer program?from a computer program?
Our BiasOur BiasOur BiasOur Bias
Could say…Could say…
• Checkers playing program “learned”Checkers playing program “learned”
• Geometry theorem proving program showed Geometry theorem proving program showed “originality” or “creativity”“originality” or “creativity”
However, once the program exists we regard However, once the program exists we regard it as nothing but a rote routine, not it as nothing but a rote routine, not exhibiting creativity or originality.exhibiting creativity or originality.
Could say…Could say…
• Checkers playing program “learned”Checkers playing program “learned”
• Geometry theorem proving program showed Geometry theorem proving program showed “originality” or “creativity”“originality” or “creativity”
However, once the program exists we regard However, once the program exists we regard it as nothing but a rote routine, not it as nothing but a rote routine, not exhibiting creativity or originality.exhibiting creativity or originality.
From this perspective, there is no way the machine can demonstrate it can “think”!
DilemmaDilemmaDilemmaDilemma
Hard AI people claim man is only a machine, Hard AI people claim man is only a machine, so anything people can do intellectually so anything people can do intellectually can be copied by a machine.can be copied by a machine.
We often believe that because a machine We often believe that because a machine was previously programmed, it cannot was previously programmed, it cannot exhibit human intellect.exhibit human intellect.• Is this fair?Is this fair?
• Perhaps the whole world is merely molecule Perhaps the whole world is merely molecule bouncing off molecule? bouncing off molecule?
Hard AI people claim man is only a machine, Hard AI people claim man is only a machine, so anything people can do intellectually so anything people can do intellectually can be copied by a machine.can be copied by a machine.
We often believe that because a machine We often believe that because a machine was previously programmed, it cannot was previously programmed, it cannot exhibit human intellect.exhibit human intellect.• Is this fair?Is this fair?
• Perhaps the whole world is merely molecule Perhaps the whole world is merely molecule bouncing off molecule? bouncing off molecule?
MusicMusicMusicMusic
Digital sampling and instrument replicationDigital sampling and instrument replication
• Frequency Frequency
• AttackAttack
• SynthesisSynthesis
CompositionComposition
• Composers now have available any sound which can exist, at Composers now have available any sound which can exist, at any rates, in any combinations, tempos, and intensitiesany rates, in any combinations, tempos, and intensities
• Immediate feedback to human composersImmediate feedback to human composers
Recording: mixing and correctionRecording: mixing and correction
Digital sampling and instrument replicationDigital sampling and instrument replication
• Frequency Frequency
• AttackAttack
• SynthesisSynthesis
CompositionComposition
• Composers now have available any sound which can exist, at Composers now have available any sound which can exist, at any rates, in any combinations, tempos, and intensitiesany rates, in any combinations, tempos, and intensities
• Immediate feedback to human composersImmediate feedback to human composers
Recording: mixing and correctionRecording: mixing and correction
The PointThe PointThe PointThe Point
Computers push us from the world of things Computers push us from the world of things to the world of ideas.to the world of ideas.
Computers supplement and extend what Computers supplement and extend what humans can do.humans can do.
Interested in what man and computer can do Interested in what man and computer can do together.together.
Computers push us from the world of things Computers push us from the world of things to the world of ideas.to the world of ideas.
Computers supplement and extend what Computers supplement and extend what humans can do.humans can do.
Interested in what man and computer can do Interested in what man and computer can do together.together.
Computers Versus HumansComputers Versus HumansComputers Versus HumansComputers Versus Humans
Computers can do many jobs better than Computers can do many jobs better than humans and are replacing humans.humans and are replacing humans.
• e.g. robotics in manufacturinge.g. robotics in manufacturing
Many humans are not equipped to compete Many humans are not equipped to compete with machines in these areas.with machines in these areas.
Very few people in the population can be Very few people in the population can be trained to develop software.trained to develop software.
Computers can do many jobs better than Computers can do many jobs better than humans and are replacing humans.humans and are replacing humans.
• e.g. robotics in manufacturinge.g. robotics in manufacturing
Many humans are not equipped to compete Many humans are not equipped to compete with machines in these areas.with machines in these areas.
Very few people in the population can be Very few people in the population can be trained to develop software.trained to develop software.
Computers Versus HumansComputers Versus HumansComputers Versus HumansComputers Versus Humans
No way to compare the number of people No way to compare the number of people whose jobs are displaced (and the number whose jobs are displaced (and the number provided new jobs) through computers.provided new jobs) through computers.
• On average, lower-level jobs are disappearing and On average, lower-level jobs are disappearing and higher-level jobs are being created.higher-level jobs are being created.
• Unclear that enough people can be trained to meet Unclear that enough people can be trained to meet the higher-level demand.the higher-level demand.
No way to compare the number of people No way to compare the number of people whose jobs are displaced (and the number whose jobs are displaced (and the number provided new jobs) through computers.provided new jobs) through computers.
• On average, lower-level jobs are disappearing and On average, lower-level jobs are disappearing and higher-level jobs are being created.higher-level jobs are being created.
• Unclear that enough people can be trained to meet Unclear that enough people can be trained to meet the higher-level demand.the higher-level demand.
Other AI ApplicationsOther AI ApplicationsOther AI ApplicationsOther AI Applications
Algebra-manipulating programsAlgebra-manipulating programs
• Depend on humans for guidance at various stages of the Depend on humans for guidance at various stages of the manipulationmanipulation
• Difficulties in areas such as Difficulties in areas such as simplificationsimplification
Algebra-manipulating programsAlgebra-manipulating programs
• Depend on humans for guidance at various stages of the Depend on humans for guidance at various stages of the manipulationmanipulation
• Difficulties in areas such as Difficulties in areas such as simplificationsimplification
yx
11 versus
xy
yxxy )(
Which expression is in simplified form?
Synthesis of Chemical CompoundsSynthesis of Chemical CompoundsSynthesis of Chemical CompoundsSynthesis of Chemical Compounds
Program providesProgram provides
• Possible routes to the synthesisPossible routes to the synthesis
• CostsCosts
• Times of the reactions in the processTimes of the reactions in the process
• Effective yieldsEffective yields
Programmer can explore various ways of Programmer can explore various ways of synthesizing a new compound, or else synthesizing a new compound, or else re-exploring old ones to find new methodsre-exploring old ones to find new methods
Program providesProgram provides
• Possible routes to the synthesisPossible routes to the synthesis
• CostsCosts
• Times of the reactions in the processTimes of the reactions in the process
• Effective yieldsEffective yields
Programmer can explore various ways of Programmer can explore various ways of synthesizing a new compound, or else synthesizing a new compound, or else re-exploring old ones to find new methodsre-exploring old ones to find new methods
Medical ApplicationsMedical ApplicationsMedical ApplicationsMedical Applications
Measurement of blood samplesMeasurement of blood samples
Medical diagnosisMedical diagnosis
Over the long run, machines can probably Over the long run, machines can probably do better than the average doctor, and it is do better than the average doctor, and it is average doctors that treat the majority of average doctors that treat the majority of the people!the people!
Measurement of blood samplesMeasurement of blood samples
Medical diagnosisMedical diagnosis
Over the long run, machines can probably Over the long run, machines can probably do better than the average doctor, and it is do better than the average doctor, and it is average doctors that treat the majority of average doctors that treat the majority of the people!the people!
Legal Issues -- CulpabilityLegal Issues -- CulpabilityLegal Issues -- CulpabilityLegal Issues -- Culpability
Human doctors are protected by “due prudence” if Human doctors are protected by “due prudence” if they make a mistake in diagnosis.they make a mistake in diagnosis.
If the machine errs, who is responsible (and If the machine errs, who is responsible (and therefore who can be sued)?therefore who can be sued)?
• Machine? Programmers? Experts who provided the rules?Machine? Programmers? Experts who provided the rules?
Often the legal problems of new applications are Often the legal problems of new applications are the main difficulty, not the engineering!the main difficulty, not the engineering!
Human doctors are protected by “due prudence” if Human doctors are protected by “due prudence” if they make a mistake in diagnosis.they make a mistake in diagnosis.
If the machine errs, who is responsible (and If the machine errs, who is responsible (and therefore who can be sued)?therefore who can be sued)?
• Machine? Programmers? Experts who provided the rules?Machine? Programmers? Experts who provided the rules?
Often the legal problems of new applications are Often the legal problems of new applications are the main difficulty, not the engineering!the main difficulty, not the engineering!
Medical ProfessionMedical ProfessionMedical ProfessionMedical Profession
Computers do billing, scheduling and record Computers do billing, scheduling and record keepingkeeping
Computers monitor patientsComputers monitor patients
Doubtful that full-time nurses could equal Doubtful that full-time nurses could equal the combination of computer and nursethe combination of computer and nurse
Computers do billing, scheduling and record Computers do billing, scheduling and record keepingkeeping
Computers monitor patientsComputers monitor patients
Doubtful that full-time nurses could equal Doubtful that full-time nurses could equal the combination of computer and nursethe combination of computer and nurse
Other Abstract Symbol Other Abstract Symbol Manipulation ProgramsManipulation ProgramsOther Abstract Symbol Other Abstract Symbol Manipulation ProgramsManipulation Programs
DifferentiationDifferentiation
Coordinate ConversionCoordinate Conversion• Extra degree of freedom in all radars so the target Extra degree of freedom in all radars so the target
cannot fly over the end of an axis of rotation and cannot fly over the end of an axis of rotation and force the radar to slew 180force the radar to slew 180° to track it° to track it
Analytical Integration (Slagle, 1961, MIT)Analytical Integration (Slagle, 1961, MIT)• Improved versions able to find any integral that can Improved versions able to find any integral that can
be done in closed form, or prove it cannot existbe done in closed form, or prove it cannot exist
DifferentiationDifferentiation
Coordinate ConversionCoordinate Conversion• Extra degree of freedom in all radars so the target Extra degree of freedom in all radars so the target
cannot fly over the end of an axis of rotation and cannot fly over the end of an axis of rotation and force the radar to slew 180force the radar to slew 180° to track it° to track it
Analytical Integration (Slagle, 1961, MIT)Analytical Integration (Slagle, 1961, MIT)• Improved versions able to find any integral that can Improved versions able to find any integral that can
be done in closed form, or prove it cannot existbe done in closed form, or prove it cannot exist
Computers Building ComputersComputers Building ComputersComputers Building ComputersComputers Building Computers
Robot assembly of computer components Robot assembly of computer components and integrated packagesand integrated packages
Integrated circuitsIntegrated circuits
Robot assembly of computer components Robot assembly of computer components and integrated packagesand integrated packages
Integrated circuitsIntegrated circuitsIn restricted areas where there can be no surprises, robots are fairly effective, but where unexpected things can happen then simple robots are often in serious trouble
Navy ApplicationsNavy ApplicationsNavy ApplicationsNavy Applications
Robots onboard ship (running on rails to Robots onboard ship (running on rails to avoid problems with obstacles?)avoid problems with obstacles?)
Damage ControlDamage Control
Remote Controlled Mine SweepersRemote Controlled Mine Sweepers
Current NPS Research – Autonomous Current NPS Research – Autonomous Underwater Vehicles (AUVs) and Underwater Vehicles (AUVs) and Unmanned Aerial Vehicles (UAVs)Unmanned Aerial Vehicles (UAVs)
Robots onboard ship (running on rails to Robots onboard ship (running on rails to avoid problems with obstacles?)avoid problems with obstacles?)
Damage ControlDamage Control
Remote Controlled Mine SweepersRemote Controlled Mine Sweepers
Current NPS Research – Autonomous Current NPS Research – Autonomous Underwater Vehicles (AUVs) and Underwater Vehicles (AUVs) and Unmanned Aerial Vehicles (UAVs)Unmanned Aerial Vehicles (UAVs)
Game PlayingGame PlayingGame PlayingGame Playing
Chess – Chess – IBM’s Big Blue defeated World IBM’s Big Blue defeated World Chess Champion Garry Kasparov May 11, Chess Champion Garry Kasparov May 11, 1997!1997!• Program examines millions of board positions per Program examines millions of board positions per
second compared to humans examining 50 to 100 second compared to humans examining 50 to 100 before making a move (based on what before making a move (based on what psychologists think chess masters think!)psychologists think chess masters think!)
Generally, machines “solve the problem by Generally, machines “solve the problem by volume of computations” rather than by volume of computations” rather than by “insight” (whatever “insight” is!)“insight” (whatever “insight” is!)
Chess – Chess – IBM’s Big Blue defeated World IBM’s Big Blue defeated World Chess Champion Garry Kasparov May 11, Chess Champion Garry Kasparov May 11, 1997!1997!• Program examines millions of board positions per Program examines millions of board positions per
second compared to humans examining 50 to 100 second compared to humans examining 50 to 100 before making a move (based on what before making a move (based on what psychologists think chess masters think!)psychologists think chess masters think!)
Generally, machines “solve the problem by Generally, machines “solve the problem by volume of computations” rather than by volume of computations” rather than by “insight” (whatever “insight” is!)“insight” (whatever “insight” is!)
Logical Versus Psychological Logical Versus Psychological NoveltyNoveltyLogical Versus Psychological Logical Versus Psychological NoveltyNovelty
Discovery led by past experiences: Discovery led by past experiences: Psychological NoveltyPsychological Novelty
Discovery from independent insight: Discovery from independent insight: Logical NoveltyLogical Novelty
• Machines do not produce logical novelty when Machines do not produce logical novelty when working properlyworking properly
• Can a human produce logical novelty?Can a human produce logical novelty?
• Is logical novelty actually possible?Is logical novelty actually possible?
Discovery led by past experiences: Discovery led by past experiences: Psychological NoveltyPsychological Novelty
Discovery from independent insight: Discovery from independent insight: Logical NoveltyLogical Novelty
• Machines do not produce logical novelty when Machines do not produce logical novelty when working properlyworking properly
• Can a human produce logical novelty?Can a human produce logical novelty?
• Is logical novelty actually possible?Is logical novelty actually possible?
Random SourceRandom SourceRandom SourceRandom Source
Claim: A truly random source contains all Claim: A truly random source contains all knowledgeknowledge
• Monkeys on typewriters!Monkeys on typewriters!
• Can be obtained if you write a program to recognize Can be obtained if you write a program to recognize “information”“information”
But, the wait time is too long and you cannot But, the wait time is too long and you cannot always recognize “information” when you always recognize “information” when you see it!see it!
Claim: A truly random source contains all Claim: A truly random source contains all knowledgeknowledge
• Monkeys on typewriters!Monkeys on typewriters!
• Can be obtained if you write a program to recognize Can be obtained if you write a program to recognize “information”“information”
But, the wait time is too long and you cannot But, the wait time is too long and you cannot always recognize “information” when you always recognize “information” when you see it!see it!
Free WillFree WillFree WillFree Will
Claim: Free will is a myth, in a given Claim: Free will is a myth, in a given circumstance you being you as you are at circumstance you being you as you are at the moment you can only do as you do.the moment you can only do as you do.
• What experiment would you perform to prove or What experiment would you perform to prove or disprove this?disprove this?
Often accept belief in free will in ourselves, Often accept belief in free will in ourselves, but deny it for others!but deny it for others!
Claim: Free will is a myth, in a given Claim: Free will is a myth, in a given circumstance you being you as you are at circumstance you being you as you are at the moment you can only do as you do.the moment you can only do as you do.
• What experiment would you perform to prove or What experiment would you perform to prove or disprove this?disprove this?
Often accept belief in free will in ourselves, Often accept belief in free will in ourselves, but deny it for others!but deny it for others!
Can Machines Think?Can Machines Think?Can Machines Think?Can Machines Think?
Perhaps thinking should be measured not Perhaps thinking should be measured not by what you do but how you do it.by what you do but how you do it.
• Child learning how to multiply versus adult Child learning how to multiply versus adult performing multiplicationperforming multiplication
““Hard AI” people accept only what is done Hard AI” people accept only what is done as a measure of success, and computers as a measure of success, and computers have not measured up by the resultshave not measured up by the results
Perhaps thinking should be measured not Perhaps thinking should be measured not by what you do but how you do it.by what you do but how you do it.
• Child learning how to multiply versus adult Child learning how to multiply versus adult performing multiplicationperforming multiplication
““Hard AI” people accept only what is done Hard AI” people accept only what is done as a measure of success, and computers as a measure of success, and computers have not measured up by the resultshave not measured up by the results
Can Computers Think?Can Computers Think?Can Computers Think?Can Computers Think?
We want to believe so machines could help us in We want to believe so machines could help us in our mental worldour mental world
We want to disbelieve to preserve our feeling of We want to disbelieve to preserve our feeling of self-importanceself-importance
Main sticky points:Main sticky points:
• If a machine does it then it must be an algorithm and cannot If a machine does it then it must be an algorithm and cannot be thinking!be thinking!
• By what forces do our thinking, self-awareness, and self-By what forces do our thinking, self-awareness, and self-consciousness affect the paths of molecules if everything is consciousness affect the paths of molecules if everything is just molecule banging against molecule?just molecule banging against molecule?
We want to believe so machines could help us in We want to believe so machines could help us in our mental worldour mental world
We want to disbelieve to preserve our feeling of We want to disbelieve to preserve our feeling of self-importanceself-importance
Main sticky points:Main sticky points:
• If a machine does it then it must be an algorithm and cannot If a machine does it then it must be an algorithm and cannot be thinking!be thinking!
• By what forces do our thinking, self-awareness, and self-By what forces do our thinking, self-awareness, and self-consciousness affect the paths of molecules if everything is consciousness affect the paths of molecules if everything is just molecule banging against molecule?just molecule banging against molecule?
Related topicsRelated topicsRelated topicsRelated topics
David B. Fogel, David B. Fogel, Blondie24: Playing at Edge of AIBlondie24: Playing at Edge of AI, , Morgan Kaufman Publishers, San Francisco, Morgan Kaufman Publishers, San Francisco, 2002.2002.
• Excellent little book on the topic of machine intelligence through Excellent little book on the topic of machine intelligence through game-playing examples (great discussion of Chess and game-playing examples (great discussion of Chess and Checkers)Checkers)
http://www.digenetics.com/company/blondie24.htmhttp://www.digenetics.com/company/blondie24.htm
• See also the following interview with current World Chess See also the following interview with current World Chess Champion Vishwanathan Anand of IndiaChampion Vishwanathan Anand of India
http://www.time.com/time/asia/features/interviews/2001/01/09/http://www.time.com/time/asia/features/interviews/2001/01/09/int.kasparov.htmlint.kasparov.html
David B. Fogel, David B. Fogel, Blondie24: Playing at Edge of AIBlondie24: Playing at Edge of AI, , Morgan Kaufman Publishers, San Francisco, Morgan Kaufman Publishers, San Francisco, 2002.2002.
• Excellent little book on the topic of machine intelligence through Excellent little book on the topic of machine intelligence through game-playing examples (great discussion of Chess and game-playing examples (great discussion of Chess and Checkers)Checkers)
http://www.digenetics.com/company/blondie24.htmhttp://www.digenetics.com/company/blondie24.htm
• See also the following interview with current World Chess See also the following interview with current World Chess Champion Vishwanathan Anand of IndiaChampion Vishwanathan Anand of India
http://www.time.com/time/asia/features/interviews/2001/01/09/http://www.time.com/time/asia/features/interviews/2001/01/09/int.kasparov.htmlint.kasparov.html