The press were delighted over the weekend to announce
that finally a computer has passed the Turing Test. The test, devised by Alan Turing in 1950 is a measure
of a machine's ability to exhibit intelligent behaviour indistinguishable
from that of a human. If the human participant in a conversation between
man and machine cannot reliably tell that he is not talking to another human
being, the machine is said to have passed the test. And now a programme called
Eugene has managed to convince a third of its human subjects that (while not
necessarily representing humanity as we know it) they were conversing with a thirteen year old boy.
I don’t see what all the fuss is about. For a start, half the perverts on teenage internet chat rooms
have been posing as thirteen-year-olds for years. But also it has come to my attention that such research has already gone way
beyond a mere five-minute keyboard-based ‘chat’. I have recently been made aware of
the findings of a Cambridge robotics laboratory’s work in producing a life-sized,
human-emulating android. The robot has been involved in field trials under
controlled conditions for a number of years now.
The following are quotations from the papers I have been privileged
to have access to and summarise some of the group’s findings:
“25/09/2010: Android first trialled on unsuspecting
public. General reaction was muted acceptance, although the jerky and sometimes spasmodic
physical articulation soon gave away non-human core and more work is needed to
model a more realistic and believable head. Speech somewhat stilted and the lack
of fluency and focus needs to be tweaked. Conclusion of observers: Not human.”
“30/06/2011: New, smoother voice software installed and physical
movement toned down. It was decided to reduce emotion facsimile to a minimum
and have the machine sitting down and stable while answering a series of
questions. First responses good, but on entering a verbal feedback loop it
became impossible to exit; the robot began to repeat answers and very quickly
its interviewer expressed doubts about dealing with an autonomously cognitive
entity. Some difficulty in switching off and terminating the experiment. Still
need more work on facial prosthetics? Conclusion: Not human.”
“17/04/2013: Tried exhibiting machine on small dais in
Cambridge market, using a scripted speech with crowd-interactive opportunities.
Robot successfully engaged small gathering for several minutes with detached
monologue but was not so convincing when attempting to respond to questions in
a human manner. Some odd answers resulted, often unconnected to actual queries,
so experiment was terminated as crowd drifted away. Some educationally
challenged participants and juveniles were taken in for a while but a number of
parents complained about the strange grimaces and nasal whine frightening small
children. Conclusion: Not human.”
One of these humanoids isn't really human...
“21/05/2014: After much work on physical presentation,
movements are still somewhat spasmodic and unconvincing. However, determined to
push on, we embarked on a last ditch nationwide tour, in order to try and fool
as many people as we could by only allowing brief exposure to the latest
prototype and limiting its speech to meaningless sound bites. Robot lasted less than twenty-four hours before the frankly
embarrassing bacon sandwich incident, after which the team is seriously
considering ‘recycling’ it or melting it down for scrap. Conclusion: Project Ed Miliband is unlikely to ever fool anybody.”
No comments:
Post a Comment