X-Message-Number: 25531
From: "John de Rivaz" <>
References: <>
Subject: Re: AI and the Singularity
Date: Thu, 13 Jan 2005 12:17:48 -0000

If  highly-specific 'AI' is possible, why can't it be networked to produce a
much better " general AI"?

One possible analogy against this proposition may be the failure of the
hospital system. It is supposed to gather together many specialist experts
to deliver the results of laboratories around the world to the patients. The
system's shortcoming isn't just a matter of safety testing and regulation.
There has recently been an exchange of email on
which indicates how specialist knowledge of the successful treatment of
strokes varies around the world, even between developed countries. The
problem doesn't seem to be whether the knowledge is available or not, but
whether it is accepted.

It appears that if you get a stroke and are admitted quickly to certain US
hospitals it can be a curable condition. Their staff know that they should
examine and treat such patients with urgency. Anywhere else (even Canada)
and permanent disability or death is the only result. It is also not just a
matter of resources -- in the UK the employees of the National Health
Service do know they have to treat certain heart conditions with urgency.
The agency has even installed suitable equipment in their ambulances and
trained paramedics in its use. Although this must have cost very large sums
of money, the money saved from long term care of permanently disabled heart
patients must have greatly outweighed this cost.

But all this could be a false analogy as the so called "specialists" in
hospitals are really specialised general AI units - not quite the same as a
single speciality AI unit.

Such a unit could be made to be capable of evaluating information as
information on its own merits, and would *not* evaluate it on the basis of
its source (eg *not* disregard it because it is from another country or
formulated by someone of a different religion or whatever.) The main reason
people disregard information from unknown sources or sources considered to
be dubious is lack of time and resources. In his book "Strange Brains and
Dr Pickover showed how many of the geniuses who have shaped the
world we live in had lots of very silly ideas as well as their brilliant
ones. Accepting any of these, on the basis of the identity of their
originator, without careful consideration is equally unproductive.
Disregarding information purely for the sole reason is that originator is of
the wrong country or whatever, it is plain stupidity. That is a fault
arising from the generalisation of the intelligence. The best specialist AI
units should have enough resources to analyse anything and reject or accept
it entirely on its merits.

Sincerely, John de Rivaz:  http://John.deRivaz.com for websites including
Cryonics Europe, Longevity Report, The Venturists, Porthtowan, Alec Harley
Reeves - inventor, Arthur Bowker - potter, de Rivaz genealogy,  Nomad .. and

> Subject: AI and the Singularity
> From: <>
> Many people fantasize thusly:
> 1. General AI is a laudable goal whose attainment will herald the
> arrival of the singularity (or something of similarly grand nature);
> The significant engineering breakthroughs to occur in the future
> will be done not by GAI, but by highly-specific 'AI'.

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=25531