Transcript Dartmouth Conference: The Founding Fathers of AI Herbert Simon

Course Overview
 What is AI?
 What are the Major Challenges?
 What are the Main Techniques?
 Where are we failing, and why?
Part I:
Introduce you to
what’s happening in
Artificial Intelligence

Done
 Step back and look at the Science
Part II:
Step back and look at the History of AI Give you an
appreciation for
the big picture
 What are the Major Schools of Thought?
 What of the Future?
 Why it is a
grand challenge
Build a Real Universal Turing Machine
 By now had all necessary ideas… 1946 Turing’s plans got approval




Automatic Computing Engine (ACE)
Progress was slow – lack of cooperation
Turing without influence, disillusioned
(…full ACE was not actually complete until 1957 (obsolete))
 1947 Turing back to Cambridge
 Interest in Neurology
 Wrote early paper on Neural Nets
 Believed complex mechanical system could exhibit learning ability
 1948 Turing and Champernowne wrote a chess program
 (for a computer that did not yet exist.)
 1948 Manchester Computer completed




Turing accepted post as deputy director
Worked on software for Manchester Mark I
1950 “Computing Machinery and Intelligence” published
… but became more interested in biology - morphogenesis
“The 'skin of an onion' analogy is also helpful.
In considering the functions of the mind or the brain we
find certain operations which we can explain in purely
mechanical terms. This we say does not correspond to the
real mind: it is a sort of skin which we must strip off if we
are to find the real mind. But then in what remains we find
a further skin to be stripped off, and so on.
Proceeding in this way do we ever come to the 'real' mind, or
do we eventually come to the skin which has nothing in it?
In the latter case the whole mind is mechanical.”
Alan Turing
“The 'skin of an onion' analogy is also helpful.
In considering the functions of the mind or the brain we
find certain operations which we can explain in purely
mechanical terms. This we say does not correspond to the
real mind: it is a sort of skin which we must strip off if we
are to find the real mind. But then in what remains we find
a further skin to be stripped off, and so on.
Proceeding in this way do we ever come to the 'real' mind, or
do we eventually come to the skin which has nothing in it?
In the latter case the whole mind is mechanical.”
Alan Turing
“The 'skin of an onion' analogy is also helpful.
In considering the functions of the mind or the brain we
find certain operations which we can explain in purely
mechanical terms. This we say does not correspond to the
real mind: it is a sort of skin which we must strip off if we
are to find the real mind. But then in what remains we find
a further skin to be stripped off, and so on.
Proceeding in this way do we ever come to the 'real' mind, or
do we eventually come to the skin which has nothing in it?
In the latter case the whole mind is mechanical.”
Alan Turing
“The 'skin of an onion' analogy is also helpful.
In considering the functions of the mind or the brain we
find certain operations which we can explain in purely
mechanical terms. This we say does not correspond to the
real mind: it is a sort of skin which we must strip off if we
are to find the real mind. But then in what remains we find
a further skin to be stripped off, and so on.
Proceeding in this way do we ever come to the 'real' mind, or
do we eventually come to the skin which has nothing in it?
In the latter case the whole mind is mechanical.”
Alan Turing
Turing’s End
 March 1952 Arrested for “Gross Indecency”
 No denial - Saw no wrong with his actions
 Convicted – given choice
 Prison
 Oestrogen injections
 Lost security clearance for GCHQ
 June 1954
 Why apple?
 Conspiracy theories…
 Security risk
 Recognition: Turing Award established (ACM, 1966)
1956 Dartmouth Conference:
The Founding Fathers of AI
John McCarthy
Alan Newell
Marvin Minsky
Herbert Simon
Claude Shannon
Ray Solomonoff
Arthur Samuel
And three others…
Oliver Selfridge
(Pandemonium theory)
Nathaniel Rochester
(IBM, designed 701)
Trenchard More
(Natural Deduction)
Dartmouth Conference: The Founding Fathers of AI
 First degree in mathematics
 Graduate work on finite automata
 Got interested in digital computers after
Summer working at IBM
 Was teaching at Dartmouth
John McCarthy
 Brought together the researchers
 Labelled the field “Artificial Intelligence”
Later…
 Worked on Formal Logic side of AI
 Invented LISP programming language
 Won Turing Award in 1971
Dartmouth Conference: The Founding Fathers of AI
John McCarthy
Alan Newell
Marvin Minsky
Herbert Simon
Claude Shannon
Ray Solomonoff
Arthur Samuel
And three others…
Oliver Selfridge
(Pandemonium theory)
Nathaniel Rochester
(IBM, designed 701)
Trenchard More
(Natural Deduction)
Dartmouth Conference: The Founding Fathers of AI
 1951 built a neural net out of
vacuum tubes,
to train a simulated rat to get
out of a maze
 Combined learning with
planning ahead in his Ph.D.
thesis
Marvin Minsky
Later…
 Society of Mind idea
 Work on artificial neural networks:
proved perceptrons can’t solve some problems
 Work in theoretical Computer Science:
2-pushdown-stack automaton = Turing Machine
 Won Turing Award in 1969
 Recent book: The Emotion Machine
Dartmouth Conference: The Founding Fathers of AI
John McCarthy
Alan Newell
Marvin Minsky
Herbert Simon
Claude Shannon
Ray Solomonoff
Arthur Samuel
And three others…
Oliver Selfridge
(Pandemonium theory)
Nathaniel Rochester
(IBM, designed 701)
Trenchard More
(Natural Deduction)
Dartmouth Conference: The Founding Fathers of AI
Most famous of all participants,
but not for AI….
Worked on analogue computer
with cogs and wheels
Showed that electromechanical
relay switches could solve
boolean algebra problems
Claude Shannon
 digital instead of analogue
 Lead to digital calculators
1948 “A Mathematical Theory of Communication”
1950 created mechanical mouse
 Could find its way out of a maze
 Learnt from experience
1950 wrote about chess playing computer program
Made a fortune in Las Vegas applying his maths to roulette etc.
Dartmouth Conference: The Founding Fathers of AI
John McCarthy
Alan Newell
Marvin Minsky
Herbert Simon
Claude Shannon
Ray Solomonoff
Arthur Samuel
And three others…
Oliver Selfridge
(Pandemonium theory)
Nathaniel Rochester
(IBM, designed 701)
Trenchard More
(Natural Deduction)
Dartmouth Conference: The Founding Fathers of AI
Algorithmic Probability
 probability of some string having been generated
by an algorithm
Applied to Induction
 Optimal Machine Learner
Theoretical idea…
 Not computable
 But can be approximated
Ray Solomonoff
Dartmouth Conference: The Founding Fathers of AI
John McCarthy
Alan Newell
Marvin Minsky
Herbert Simon
Claude Shannon
Ray Solomonoff
Arthur Samuel
And three others…
Oliver Selfridge
(Pandemonium theory)
Nathaniel Rochester
(IBM, designed 701)
Trenchard More
(Natural Deduction)
Dartmouth Conference: The Founding Fathers of AI
 Originally a political scientist – how bureaucracies function
 Became interested in organisational decision making
 Around 1954 he decided…
best way to study problem-solving is to simulate on computer
 Developed experimental technique of verbal protocol analysis
 Interested in role of knowledge in expertise
 1978 won Nobel Prize in Economics
Herbert Simon
“Over Christmas,
Allen Newell and I created
a thinking machine.”
January 1956
Dartmouth Conference: The Founding Fathers of AI
Alan Newell
 1955 designed a chess playing program
Later…
 1983 Developed SOAR architecture
 Attempting a unified theory of cognition
Dartmouth Conference: The Founding Fathers of AI
 1956 Logic Theory Machine
 Saw that theorem proving can be reduced to search
 Search tree to find a proof for a theorem
 Considered to be first AI program
 1957 General Problem Solver
 Heuristics
 Means-ends analysis
 1975 won Turing Award
Alan Newell
Herbert Simon
Dartmouth Conference: The Founding Fathers of AI
John McCarthy
Alan Newell
Marvin Minsky
Herbert Simon
Claude Shannon
Ray Solomonoff
Arthur Samuel
And three others…
Oliver Selfridge
(Pandemonium theory)
Nathaniel Rochester
(IBM, designed 701)
Trenchard More
(Natural Deduction)
Dartmouth Conference: The Founding Fathers of AI




Developed a checkers playing program
Developed alpha-beta tree idea
Made his program learn to improve itself
1962 his program beat a state champion
Arthur Samuel
Dartmouth Conference: The Founding Fathers of AI
John McCarthy
Alan Newell
Marvin Minsky
Herbert Simon
Claude Shannon
Ray Solomonoff
Arthur Samuel
And three others…
Oliver Selfridge
(Pandemonium theory)
Nathaniel Rochester
(IBM, designed 701)
Trenchard More
(Natural Deduction)
“We propose that a 2 month, 10 man study of artificial
intelligence be carried out during 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 simulate it.
An attempt will be made to find how to make machines use
language, form abstractions and concepts, solve kinds of
problems now reserved for humans, and improve themselves.
We think that a significant advance can be made in one or
more of these problems if a carefully selected group of
scientists work on it together for a summer.”
McCarthy et al 1955
“We propose that a 2 month, 10 man study of artificial
intelligence be carried out during 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 simulate it.
An attempt will be made to find how to make machines use
language, form abstractions and concepts, solve kinds of
problems now reserved for humans, and improve themselves.
We think that a significant advance can be made in one or
more of these problems if a carefully selected group of
scientists work on it together for a summer.”
McCarthy et al 1955
“We propose that a 2 month, 10 man study of artificial
intelligence be carried out during 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 simulate it.
An attempt will be made to find how to make machines use
language, form abstractions and concepts, solve kinds of
problems now reserved for humans, and improve themselves.
We think that a significant advance can be made in one or
more of these problems if a carefully selected group of
scientists work on it together for a summer.”
McCarthy et al 1955
“We propose that a 2 month, 10 man study of artificial
intelligence be carried out during 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 simulate it.
An attempt will be made to find how to make machines use
language, form abstractions and concepts, solve kinds of
problems now reserved for humans, and improve themselves.
We think that a significant advance can be made in one or
more of these problems if a carefully selected group of
scientists work on it together for a summer.”
McCarthy et al 1955
“We propose that a 2 month, 10 man study of artificial
intelligence be carried out during 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 simulate it.
An attempt will be made to find how to make machines use
language, form abstractions and concepts, solve kinds of
problems now reserved for humans, and improve themselves.
We think that a significant advance can be made in one or
more of these problems if a carefully selected group of
scientists work on it together for a summer.”
McCarthy et al 1955
Dartmouth Conference: The Founding Fathers of AI
John McCarthy
Alan Newell
Marvin Minsky
Herbert Simon
Claude Shannon
Ray Solomonoff
Arthur Samuel
And three others…
Oliver Selfridge
(Pandemonium theory)
Nathaniel Rochester
(IBM, designed 701)
Trenchard More
(Natural Deduction)
1956 Dartmouth Conference: What was achieved?
 Not much
 People didn’t agree on the format and weren’t all there together
 Newell and Simon didn’t spend much time…
 Too busy working on their logic theorist
 McCarthy was disappointed
“The main reason the 1956 Dartmouth workshop didn't live
up to my expectations is that AI is harder than we thought.”
(McCarthy in 2006)
 But got people to know each other…
AI Developments from 1956 - 1963
Main Thrusts of Work in Early Days…
1.
Reduce the search tree for search programs




2.
For example, search programs for:
Logic Theorems
Geometry theorems
Algebra
Make computers learn for themselves




For example:
Chess playing machines
Checkers playing machines
Pattern recognition
Newell and Simon’s progress…
“It is not my aim to surprise or shock you –
but the simplest way I can summarize is to say that there
are now in the world machines that can think, that can
learn and that can create.
Moreover, their ability to do these things is going to increase
rapidly until – in a visible future – the range of problems
they can handle will be coextensive with the range to which
the human mind has been applied.”
Herbert Simon, 1957
Newell and Simon’s progress…




Discovered that humans don’t really act like Logic Theorist
Psychologists Moore and Anderson had pioneered “think aloud”
experiments
Other AI researchers were merely concerned with programs that
performed well
Newell and Simon wanted programs that solved problems in the
same ways as humans



They branched off…
More Cognitive Science than core AI
Developed the general problem solver (GPS)



Using heuristics
Using means-end analysis
Solved monkey-chair-banana type problems
Work at IBM…





Minsky hired Herbert Gelernter to work on new IBM 704
Geometry Theorem Prover
Gave visual input of geometry problem by coding it in (not camera)
This input reduced branching factor from 1000 to 5
Took Gelernter 3 yrs to program it (much longer than expected)

Also at IBM


Samuel working on his checkers program
Alex Bernstein working on chess program
Trouble at IBM…



AI work noticed by popular press
Publicity attracted attention of IBM shareholders
Asked T. J. Watson (president of IBM)




IBM noticed that customers were frightened of idea of
"electronic brains" and "thinking machines“
1960 Internal report prepared recommended IBM stop AI
IBM told customers




explain why research dollars were being used for "frivolous matters"
computers will only do what they were told
Bernstein became psychiatrist
Gelernter became physicist
Samuel went to Europe
McCarthy’s progress…





Developed LISP programming language
List Processing
Makes it easy to program AI ideas
Makes it easy for a program to alter its own instructions
McCarthy wanted programs to add to their own
commonsense

To deduce consequences


Started looking at IF-THEN rules (like later expert systems)
LISP was heavy on computer power – more useful in 1970s

McCarthy also pioneered idea of time-sharing computers
Minsky’s progress at MIT…




Sputnik left US behind technologically
US created DARPA
1963 MIT got over $2M for Machine Aided Cognition
MAC project brought MIT about $3M a year in grants thereafter
 Minsky’s student James Slagle worked on SAINT program


Solved symbolic integration problems
Later evolved into MACSYMA