----- Original Message -----
From: "Richard Baker" <[EMAIL PROTECTED]>
To: "Brin-L" <[EMAIL PROTECTED]>
Sent: Monday, March 04, 2002 10:23 AM
Subject: Re: SCOUTED: Science Meets Spirituality, and Wireless Nanotech VR


> The Fool said:
>
> > I'll place my bets on 2035.
>
> I'm going to be a conservative and say that Moore's Law will continue as
> it is for the next century. Given some advances in software and
> cognitive science, this means that by 2030s the AI population will have
> be thinking as many thoughts as a human city. The early 2050s will see
> "equality" - the AIs and humans will be producing the same amount of
> computational activity. By the late 2060s, the output of humans will be
> about a thousandth of that of our civilisation. The end of the century
> will see the AIs out-thinking us by a factor of a billion. (I expect
> the first human-equivalent mainframes around 2025.)
>
> Rich
> GSV Pessimist

Well, I think that you, as well as the others here,  are overwhelming
optimistic when it comes to the potential for AI.  I've been around claims
for neural networks and AI for about 20 years now.  I've seen companies put
tens of millions into problems that I thought could easily be solved by AI.
Twenty years later, they still are not solved.  My former company has thrown
millions away to an AI company that was formulated by professors from a name
US institute (i.e. most people would list them in their top 5 technology
schools).  All they did was sell smoke and mirrors.  They had a cute robot
film they showed and talked a lot about "the big picture."  Yet, their AI
work added nothing to the project, and all the company got for their
millions was a glowing report on what could be done.

The field I was in required downhole sensing and decision making, the type
of stuff that AI robots would be perfect for.  Nothing has come out of it.
Indeed, the very best groups like this have done has been writing software
that allows decision making algothithms that have already been thought out
by the people in the field to be implemented in a more efficient fashion.

Dan M.

Reply via email to