June 22, 2010 Handout for OLLI summer
workshop on
COMPUTERS VERSUS HUMANS*
Leader: Theo Pavlidis (t.pavlidis@ieee.org)
NEWS OF THE DAY
The cover page article of the June 20 issue of the NY
Times .Magazine is titled: "Who is Watson? I.B.M. has taken artificial
Intelligence to the Next Level. But does that mean a computer can actually
think?" The article describes the work behind a machine that
plays the TV game Jeopardy. On the web: http://www.nytimes.com/2010/06/20/magazine/20Computer-t.html?hp
IMPORTANT FACTS ABOUT COMPUTERS
Computers can deal only with tasks that can be expressed
in mathematical terms. This includes not only "number crunching"
but also mathematical logic. The latter is the basis of search engines
(such as Google), and game playing programs, such as Deep Blue (chess)
and Watson (Jeopardy). If a problem can be expressed in mathematics, computers
will usually outperform humans because of their speed. (Of course, before
they can do that, they must be programmed by human experts.)
Computer ability to interpret an object is much better
for written text than for pictures, speech, music, etc. This is because
stored directly in machine memory is characters (letters, etc) that have
meaning for humans. The information stored in memory for other types of
objects has no explicit human meaning. This is called the semantic
gap by optimists and the semantic abyss by pessimists.
Computer "Learning" is quite different from
human learning. The term actually refers to a class of mathematical techniques
that have only superficial resemblance to human learning. For examples,
computers need thousands of samples of each letter before they can learn
to "read" an alphabet. Humans need fewer than 10 samples of
each letter.
Caveat: There are problems that can
be expressed through mathematics but still cannot be solved by either
computers or humans. Weather prediction is the best known example. Some
cryptographic techniques rely on the inability of computers to "break
the code" in less than several years of computer time.
CHAOTIC SYSTEMS
The term "Chaotic" is used to describe a system
that its outcome is predictable in theory but not in practice. Often,
the outcome may be predictable for a short time but the prediction becomes
less and less accurate as time goes on. For example, weather predictions
are fairly good for, say, the next 12 hours but their accuracy drops rapidly
after that.
Some people say that chaotic systems have free will.
(For more on that issue see
http://www.theopavlidis.com/essays/FreeWillChaos.htm.)
For more on chaos see: James Gleick, Chaos - Making
a New Science, New York, Viking, 1987. (A paperback edition came
out in August of 2008. It sells from Amazon for $13.60.)
* Copyright ©2010 by Theo Pavlidis
|