Title: 27. The Turing Machine 28. The Brain as a Computing Device 29. Universal Computation 30. Computatio
1 27. The Turing Machine 28. The Brain as a
Computing Device 29. Universal Computation
30. Computation and Cognitivism 31. The
Machine Brain 32. Functionalist Separation of
Mind from Brain 33. The Physical Symbol
Systems Hypothesis 34. A Theory of Intelligent
Action 35. Could a Machine Really Think? 36.
The Turing Test 37. The Loebner Prize 38.
Problems with the Turing Test 39. Inside the
Machine Searles Chinese Room
Introducing Artificial Intelligence by Henry
Brighton and Howard SelinaSECTION C.
. Chuck Selden
- Name Chuck Selden
- Profession Health Scientist Administrator
- Affiliation NIH-OD-OER-OEP-ESTO
- Relation (to the subject matter) My
Intelligence, unfortunately, is not artificial - Question (that you would like to explore in the
dialogue) - What is minimal number, speed, and level of
complexity for a system of cells/nodes/reactive
memory units needed for a machine to reach AI?
2The Turing Machine
- Imaginary Machine
- Contains a Finite Control that defines all
states in which machine can exist - Machine states reached by what is read off an
infinite tape inscribed with symbols - Turing Machine predated all modern computers
- Exemplar of the theory of computation
3The Brain as a Computing Device
- Warren McCulloch and Walter Pitts in 1943 A
Logical Calculus of the ideas Immanent in Nervous
Activity - The neuron as a computing device
- Configurations of neurons as logic gates can
perform any calculation computable by a Turing
Machine
4Universal Computation
- All computers are restricted to perform
calculations as would a Turing Machine - All other machines and the human brain can also
be reduced to Turing Machines - NOTE The human brain, and other animal brains,
have the ability to initiate thoughts and actions
without relying on a list of instructive symbols
(inputs) ie. spontaneous activity - Thus, the brain is much more than a Turing
Machine
5Computation and Cognitivism
- Turing Machines and computers work in various
ways - TM reads an infinite tape of symbols
- Desktop computer uses random access memory to
carry out computations - The Brain uses a vast network of neurons
- But how does the mind compute?
6The Machine Brain
- Brain function has been explained historically as
using the most complex machinery of the period - Renaissance - clockwork
- Industrial age - steam engine
- Present day - computer
- Brain as hardware, mind as software
- Mind has no mass, no material
- Operates on the hardware
7Functionalist Separation of Mind from Brain
- Operations matter, not the physical process
- Functionalism means many different ways to do
something - Example a spreadsheet will run on many different
kinds of operating systems - Cognition, then, is platform-independent
- Machines could be cognitive just like a brain is.
8The Physical Symbol Systems Hypothesis (PSSH)
- The mind relies on syntactic manipulation of
symbols - Cognition relies on manipulation of symbolic
representations of things in the world - The platform carrying out the manipulations is
irrelevant (neurons, silicon, tin cans and
string, clockworks, steam engines) - A physical symbol system has the necessary and
sufficient means for intelligent action.
9A Theory of Intelligent Action
- Newall and Simon wanted to tell what operations
were needed for intelligent action - The science of AI is to test PSSH
- Acording to PSSH, the right program is all that
is required for a theory of intelligent action - NOTE
- Do computers really use a physical symbol system?
THIS IS THE KEY TO AI - NO, Computers use numerical representations of
symbols, thus, computers cannot become (act)
intelligent.
10Could a Machine Really Think
- Is there a decisive argument that proves machines
cannot think? - Alan Turing in 1950 asked Can Machines think?
but thought the question was ill-defined and
meaningless - Can a computer win the imitation game?
- The test can a person on a computer, linked to
another computer in another room, tell if the
other computer is responding to questions or a
conversation unaided by a human, or not?
11The Turing Test
- Interrogator asks any question
- On basis of the response(s), which do not have to
be truthful, the interrogator decides if is
corresponding with a human or a machine. - Machine passes test if interrogator thinks it is
a human. - Example
- Add 34957 to 70764
- Answer 105621
12Correct Answer 105721
- Did the wrong answer make you think a human was
in the other room, answering via computer?
13But what is thought?
- Is there thinking involved?
- Is it just processing?
- Example, from Noam Chomsky, using the word Swim
- Does a submarine swim?
14The Loebner Prize
- In 1990, start of an annual competition for
100,000 and a gold medal - To be won by first person who designs a computer
program that passes the Turing Test - Some cash and bronze medals have been awardednot
the gold. - The authors suggest that no one will win for some
time.
15Problems with the Turing Test
- Test ignores linguistics and how the machine
operates - Goal for developing AI is not to win a prize or
fool a person, but to make an intelligent
machine, a machine with a mind. - Turing test could be passed by a computer that
memorizes all questions and responses to all
questions ever asked, and, given that any new
question will be like an old one, will give
reasonable answers every time. - It would still not be intelligent.
16Inside the Machine Searles Chinese Room
- In 1980s, John Searle tried to knock out AI
claimants with this argument - A man in a room with a rule book for using
Chinese symbols would get questions in Chinese
and provide answers in Chinese. - But couldnt read any Chinese.
- Did he understand the exchanges?
- NO Even though he might provide the right
answers and so pass the Turing Test. - Similarly, the computer might compute, but will
not, cannot, understand what it is doing...it has
no cognition.