Minds and Computers : An Introduction to the Philosophy of Artificial Intelligence

Exercise 10.2
What are your initial reactions to the Turing test as an
indicator of machine intelligence? What problems do you
110
  

---

envisage with using this method for determining the presence
of a mind?
Keep in mind that the claim is not that passing the Turing test is
su
fficient for having a mind. The thought is that passing the Turing test
gives us good grounds to suppose that the test subject has a mental life.
As it stands the Turing test remains underspecified. We have not yet
set a length of time for conducting the test and a lot hangs on this. If,
for instance, the test is to be conducted for only five minutes, then we
might rightly have qualms about its reliability as an indicator of men-
tality. If, on the other hand, the interrogator is allowed to spend as
much time as she requires until she feels capable of making an ade-
quate assessment, then it seems we can have a lot more confidence in
that assessment – particularly if, as we should, we take the assessment
to be revisable, contingent on the results of further such tests.
Since 1991, an annual competition for the Loebner Prize (named
after the philanthropist sponsor) has been conducted, in which par-
ticipants submit to a formalised version of the Turing test. The com-
petition is held over one day with a $100,000 prize and a gold medal
on o
ffer for the a machine which is held by the judges to be indistin-
guishable from a human being. As yet no machine has come close to
this. Each year, however, a $2,000 prize and a bronze medal is awarded
to the machine which performs the most impressively of the entrants.
Turing envisaged, in 1950, that by the year 2000 machines would
be able to convincingly pass a Turing test of moderate duration. No
doubt he would be disappointed with the current state of play – the
transcripts of conversations with Loebner Prize entrants show that
some rather unsubtle questioning su
ffices to distinguish the machine