Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!think.com!zaphod.mps.ohio-state.edu!wuarchive!uunet!mcsun!ukc!edcastle!helium!cam
From: cam@aifh.ed.ac.uk (Chris Malcolm)
Newsgroups: comp.ai.philosophy
Subject: Re: a naive idea
Keywords: intelligence, human, artificial
Message-ID: <1991May27.014857.3632@aifh.ed.ac.uk>
Date: 27 May 91 01:48:57 GMT
References: <1991May16.134123.6335@cdc835.cdc.polimi.it>
Reply-To: cam@aifh.ed.ac.uk (Chris Malcolm)
Organization: Dept AI, Edinburgh University, Scotland
Lines: 55

In article <1991May16.134123.6335@cdc835.cdc.polimi.it> gin001@cdc835.cdc.polimi.it (Mauro Cicognini) writes:
>It may be a naive idea, but my assertion is that artificial intelligence 
>is not worth for any economical results it may produce. In fact, it will
>always be less expensive to train a human being to do a certain intel-
>lectual job than to devise an artificial system to do the same thing.
>This because the machine has first to be imbued with "artificial" intel-
>ligence; and this task has yet to prove itself easy. I doubt that it ever
>will, though this is only a personal opinion.

>My point is that, even if we reach the point where we can "make" intellig-
>ence out of silicon (or gallium arsenide), it will always be more costly
>than to have intelligence produced to old way, that is, to make babies.

You're a decade or so out of date with this argument. 

	1. A chess machine which can beat me. I can buy one of these for
approx $100. Training a human being to do this is a LOT more expensive.
In fact, since most people couldn't be trained to this level of
performance, actually selecting the individual, before starting the
training, would probably cost more than buying the machine.

	2. A general purpose mathematics assistant, which can do linear
algebra, find roots, integrate, draw graphs, etc., can be bought (in the
form of a computer program) for a good deal less money than training a
human being to the appropriate level; and once again, you would first
have to select a suitable trainee.

	3. A language translation hacker. While machines which can
produce high quality translations between natural human languages do not
yet exist, there already exist machines which can translate badly, but
well enough that they are in regular use by professional document
translators. The machine produces a rough draft, and the human corrects
it with an editor, enabling the human to do a great deal more
translation in a given time. Once again, training a human to the same
level of expertise would be a lot more expensive than buying the
program.

In all these cases not only are the machines VERY much cheaper to buy than
training an appropriate human, but they are also VERY much cheaper to
run than the salary the trained human would require.

Of course, there is one sense in which you are correct: the man-years of
research that were necessary before it was possible to build one of
these machines did far exceed the cost of training ONE suitable human
candidate for the job, just as the man-years which went into inventing
and building the telephone network far exceed the cost of you having to
write letters instead of phoning for all of your life. Fortunately for
industry that is not the cost-equation by which research and development
is costed, and fortunately for you there are so many people who want
telephones that each subscriber need bear only a minute and easily
affordable fraction of the cost of the world's telephone network.
-- 
Chris Malcolm    cam@uk.ac.ed.aipna   +44 (0)31 667 1011 x2550
Department of Artificial Intelligence,    Edinburgh University
5 Forrest Hill, Edinburgh, EH1 2QL, UK                DoD #205