🚀 go-pugleaf

In article <5lq9qj$m35$1@pheidippides.axion.bt.co.uk>,
    Paul McIlroy <paul.mcilroy@bt-sys.bt.co.uk> wrote:
>If you were asked to develop a computer like HAL from 2001, Holly from
>Red Dwarf or Deep Thought from the Hitch-hikers Guide to the Universe,
>and to complete it within 10 years could you do it ? Would you be first
>across the finishing line, or would someone else beat you? Where would
>you start?  And where would you end up?

    I'm as skeptical as anyone else when it comes to the idea of AI
being right around the corner - having celebrated HAL 9000's birthday
on January 12 of this year, like many others, almost three decades
after the story was written - but I'm a bit more optimistic about the
ability to create something that *seems* intelligent.
    Speech recognition programs have become real sooner than I ever
expected, but not because the techniques have gotten any better, but
because the raw computing power now exists to throw at the problem.
Discrete speech is usable, and continuous speech recognition may be
less than five years away.
    They don't have much in the way of an AI component - mainly just a
set of heuristics and fuzzy logic - but they still manage to
mindlessly transcribe speech, often as well as a fully sentient
secretary.
    I really wouldn't be surprised now if, within another 30 years,
raw power makes it possible for computers (and even robots, finally!)
which respond in a very human-like way, despite having nothing
resembling a thought process in their entire system.
    In other words, their apparent 'intelligence' would be entirely a
user interface. Not Turing test winners, but functional, useful, and
practical. Currently around 70% of computing power is going to GUIs,
and I wouldn't be surprised to see that figure keep shooting up as the
techniques and processor speed are further developed. Many programs
already provide popular 'wizard' functions meaning AI-UI components
aill probably have a strong marketing potential, and that means
development effort.
    As for 'how', I have no idea. My advice is to look at how video
games are being developed these days - they're already around 95% user
interface.

--
John Bayko (Tau).
bayko@cs.uregina.ca
http://www.cs.uregina.ca/~bayko