Re: The most powerful approach to AI ever discovered?

From: Anders Sandberg (asa@nada.kth.se)
Date: Sat Dec 15 2001 - 14:43:54 MST


On Sat, Dec 15, 2001 at 02:01:02PM -0600, Harvey Newstrom wrote:
> > From: "John Clark" <jonkc@worldnet.att.net>
> > > http://www.eet.com/story/OEG20011213S0065
>
> Too bad they don't tell us what this great discovery is. They mention
> neural networks, predict great results, and offer to sell training about
> this unknown discover for $50K per person sight-unseen. This sounds like
> another Ginger. If they really had this discovery, they would make more
> money using the technology. Why sell it to competitors for $50K a pop?
> There is so much hype in the technology sector today, that I just assume
> these kinds of things are false until proven otherwise.

Yes. If I had come up with some super-clever general neural network
solution, I would sell it as a black box or hire out consultants to
help embedd it into other systems. If I had come up with something very
cool-sounding but less useful, I would (assuming I wanted to make money
on it) probably try to earn money by selling the hype - for example by
selling expensive courses in my hype technology. If you have paid $50K
for a course in a technology that turns out to be less useful than you
expected, it is human nature to downplay your disappointment and say it
is great - few will loudly say it is junk.

So far neural networks have not been a commercial success as standalone
products; the networks that really are useful and do earn money are all
embedded in other products. Neural network chips have been developed,
but so far none have been commercially viable. With FPGA
performance/price increases it seems likely there will be more hardware
implemented ANNs soon (there is a great deal of interest in that in my
research group, at least).

-- 
-----------------------------------------------------------------------
Anders Sandberg                                      Towards Ascension!
asa@nada.kth.se                            http://www.nada.kth.se/~asa/
GCS/M/S/O d++ -p+ c++++ !l u+ e++ m++ s+/+ n--- h+/* f+ g+ w++ t+ r+ !y



This archive was generated by hypermail 2b30 : Sat May 11 2002 - 17:44:27 MDT