> > Emlyn wrote: > > The short story... A guy got Eliza, hooked it up to an AOL
> >>messenger account... The chatters don't know that they are talking to a
> > >
> > > http://fury.com/aoliza/
> > >
So we now have two *limited* versions of the Turing test. This
one is limited because the posters on Emlyn's AOLisa site were
likely young people who had never heard of Elisa. Had they been
briefed beforehand, they would not have been fooled. I assume
most of those reading this have played with Elisa or at least know
what it is. We would likely have recognized the joke early. Yet
this experiment definitely must be considered a Turing test, of sorts.
The other interesting *limited* case of a Turing test would be
in chess software. I have watched this field develop over the
past 25 years. Up until about 15 years ago, it was not difficult
to distinguish between a human player and a machine. As the
machines got better, they became more human-like in their play.
They became more likely to give up wood for position, something
the early software would seldom do. They became much more
strategic and less tactical.
A most interesting case study for post-singularity historians would
be the match between then-world champion Gary Kasparov and
Deep Blue. In the last game of the match, after Deep Blue
made its last move, Kasparov resigned and asserted that some
human must be helping the machine! Its play demonstrated such
a deep "understanding" of positional chess that the grandmaster
was convinced no machine could ever do it. Only after it was
repeatedly demonstrated did Kasparov finally accept that his
beloved game really can be reduced to a set of equations.
Modern chess software running on modern computing hardware
plays in a way that is indistinguishable from a really sharp human
player. I sure cant tell the difference.
My friends, thats all it is, and thats all we are, wet computers
that somehow do what we do. Both cases give me hope that
we will eventually figure out a way to encase ourselves in a
more robust and permanent "body". spike
This archive was generated by hypermail 2b30 : Mon May 28 2001 - 09:59:42 MDT