Manton Matthews Department of Computer Sc.

1
Manton MatthewsDepartment of Computer Sc.
Engr.
How computers think but do they understand?
Scholars Day April 5, 2008
2
What is Thinking?

• Merriam-Webster Definitions
• Thinking- The action of using ones mind to
  produce thoughts
• Can a machine think?

http//www.merriam-webster.com/dictionary/
3
What is Intelligence?

• the ability to learn or understand or to deal
  with new or trying situations
• the ability to apply knowledge to manipulate
  one's environment or to think abstractly as
  measured by objective criteria (as tests)

http//www.merriam-webster.com/dictionary/
4
Computers Acting Intelligently

• Interacting with humans
• Analyzing situations and making appropriate
  decisions

5
Examples of Intelligence? Interact with
customers in a business

• Worker must be able to
• Understand customers wishes
• Address the request
• Maybe handle the details of the sale

6
More Specifically a Bank Teller

• Teller must be able to
• Understand customers wishes
• Count checks/money
• Give cash and receipts to the customer

7
Computer Bank Teller

• Teller must be able to
• Understand customers wishes
• Count checks/money
• Give cash and receipts to the customer

8
Chess Player

• chess is gymnastics of the mind

9
Evolution of Chess Playing Computers
10
Deep Blue defeats World Chess Champion

• "You have to be on full guard every move of the
  game, which means it is more exhausting. I think
  Deep Blue is stronger than all but a handful of
  top human players." - Garry Kasparov

11
Aibo

• .

12

• .

13
Examples of Aibos thinking

• Waking up on back (114)
• Picking up the Aibone ( Aibos bone) (303/200)
• Aibo juggling ball (351/212)
• Aibo ball balancing trick (505/142)
• Time for rest and recharge. (126)
• Solving a maze (259)
• Following Directions
• Remote photographer

14
Aibos maze solving

• Aibos sensors
• Solve a maze
• Walk forward till we sense an obstacle
• Pan head from left to right looking for longest
  distance that is unobstructed
• Turn that direction and repeat 1-3
• Else if all directions obstructed turn 180
  degrees and repeat 1-3

15
Email me a photo!

• x

16
Star Treks Doctor

• Why consider him here?

http//en.wikipedia.org/wiki/Doctor_(Star_Trek)
17
Practical Computers Acting Intelligently

• Expect computers to do some things well (even
  better than a human)
• Number crunching
• Combinatorial search (chess)
• But how about tasks requiring subjective
  judgment?
• Example Predict success of a student in a
  training program
• 1955 Paul Meehl showed simple statistical
  learning algorithms out performed experts in 19
  or 20 studies
• Since 1999 ETS has used a program to grade essay
  questions on the GMAT, 97 agreement with human
  expert graders (the same as with other human
  experts)

18
Alan Turing

• 1912 (23 June) Birth, Paddington,
  London1931-34 Undergraduate at King's College,
  Cambridge University1932-35 Quantum mechanics,
  probability, logic1935 Elected fellow of King's
  College, Cambridge1936 The Turing machine,
  computability, universal machine1936-38
  Princeton University. Ph.D. Logic, algebra,
  number theory1938-39 Return to Cambridge.
  Introduced to German Enigma cipher
  machine1939-40 The Bombe, machine for Enigma
  decryption1939-42 Breaking of U-boat Enigma,
  saving battle of the Atlantic1943-45 Chief
  Anglo-American crypto consultant. Electronic
  work.1945 National Physical Laboratory,
  London1946 Computer and software design leading
  the world.1947-48 Programming, neural nets, and
  artificial intelligence1950 The Turing Test for
  machine intelligence1954 Death (suicide) by
  cyanide poisoning

19
Turings Imitation Game

• Version 1
• Three participants A,B and the interrogator
• One of A and B is a woman, the other a man
• In separate rooms, communicate only through
  terminal
• Assuming the man tries to fool the interrogator,
  can he still identify the woman?
• Version 2
• Replace the man with a computer.
• Turings claim If the computer can make the
  interrogator miss 50 of the time, then the
  computer has acted intelligently.

20
Loebner Prize

• The Loebner Prize is an annual competition that
  awards prizes to the Chatterbot considered by the
  judges to be the most humanlike

21
John Searle's Chinese room

• Thought experiment By John Searle 1980
• to show that a symbol processing machine like a
  computer can never be properly described as
  understanding
• Suppose a computer that behaves as if it
  understands Chinese so well it passes the Turing
  Test
• Chinese characters ? program ? Chinese response
• ? means pages with the characters are slid under
  the door
• Now suppose Searle is in the room reading a book
  in English that has the program instructions in
  it And he follows them to produce the output
  character

22
But can a computer understand?

• Circle of terms revisited
• Intelligence
• Think
• Understand
• Mind
• Conscious
• This is the realm of philosophers!

23
Weak AI vs Strong AI

• Weak AI hypothesis
• It is possible for a machine to act
  intelligently or even less objectionable to act
  as if it were intelligent.
• Strong AI hypothesis
• It is possible for a machine to actually think
  to understand to be conscious

24
Examples of Quirkiness in Meanings

• Consider the definitions of two forms of
  locomotion
• flying through the air and swimming through
  water.
• Fly - 1 a to move in or pass through the air
  with wings
• So airplanes fly
• Swim - 1 a to propel oneself in water by natural
  means (as movements of the limbs, fins, or tail)
• But boats dont swim

Norvig and Rusell AI a Modern Approach
25
So now do we agree?

• Computer can act intelligently but they cant
  understand.

26
The Human Mind

• How do humans
• think
• remember
• understand

27
Your Brain

• 100 billion nerve cells, neurons
• Interconnections
• Neurons have
• Cell body
• Axon
• Dendrites
• Synapses
• Neuro-transmitters
• They stimulate each other by firing

28
http//health.howstuffworks.com/brain1.htm
29
Neurons Firing

• Neurons that fire together get better at it they
  form associations

The Brain Fitness Program
30
Hebbs Law

• Donald Hebb a Neuro-psychologist (1949)
• When an axon of cell A is near enough to excite
  cell B and repeatedly or persistently takes part
  in firing it, some growth process or metabolic
  change takes place in one or both cells such that
  A's efficiency, as one of the cells firing B, is
  increased.
• "Neurons that fire together wire together."

31
Organic vs Inorganic Chemistry

• 1828 Wohler wrote
• "I must tell you that I can make urea without the
  use of kidneys, either man or dog. Ammonium
  cyanate is urea."
• This organic synthesis dealt a severe blow to a
  widespread belief called vitalism which
  maintained that organic chemicals could be
  modified by chemistry but could only be produced
  through the agency of a vital force present in
  living plants and animals.

http//www.3rd1000.com/urea/urea.htm
32
Organic vs Inorganic Minds

• Can we build an artificial Neuron?
• Not yet certainly
• But artificial hearts,
• Suppose we could and further suppose that we
  could replace one in the brain
• And replace another
• Did that mind lose its ability to understand when
  we replaced the first artificial neuron? the 2nd
  ? the last

33
Moores Law 1965 for next 10 years

• The number of transistors in an Integrated
  Circuit is doubling approximately every two years

http//en.wikipedia.org/wiki/Moore's_law
34
(No Transcript)
35
(No Transcript)