Until
next week, 
Intelligent Machine Post
Dateline: March 8, 1998
MY intended review of Neuroshell Easy will have to be postponed for two weeks. In the first place, I'm still trying to get hold of some decent datasets, having discovered that my own thesis survey data is lost, and in the second place I'll be up in Michigan's Crystal Mountain resort most of next week, talking about the Internet to a group of community health leaders. I can whet your appetite, though, with the news that one of my readers wrote to say he's been using Neuroshell successfully to pick stocks.
Links to some of the more interesting AI stories each week are given on the home page for this site, but since the home page links change every week I shall summarize the more interesting stories in IMP so they don't get lost and forgotten.
One more change to IMP: The focus of this site is not the technological detail of AI, but its implications. One of the key implications is the loss of jobs to machines. Yes, I know; joblessness (in the United States) is at an all-time low. My hypothesis is that (1) much of the new employment (in the United States, at least) comes from the technology sector, but that jobs in that sector will eventually shrink once machines start to design and manufacture themselves, and (2) employment is increasingly self-employment, not traditional jobs.
I predicted in last week's feature about the future that intelligent machines would take over many if not most jobs within 30 years, including even the job of a surgeon, and I stick by that prediction. The summaries below are part of the growing evidence. Where a specific news item seems to me to put writing on the wall for any given job, I will henceforth place the job in Ellis's Endangered Employment Katalog (EEEK!), and note the entry under the relevant summary.
Enough chatter; here's the beef:
HAL is Loose in the Networks!
Not really, but something like
it. MARS (Multi-Agent Reasoning System) is a coordinated suite of intelligent
agents that can monitor, troubleshoot, and reconfigure a distributed network
all on its own. It learns over time and needs little to no setup. Network
managers oughtta love it, at least until, like HAL, it takes their jobs.
The underlying AI engine is Bayesian. The program is written in Java, so
should run on any platform. According to Data
Communications of March 1998, MARS is being marketed by France
Connexion Ingenerie. (When I tried it, that link only brought up a
server notice, not the FCI site. Seems like it's under construction.)
The name MARS and the functionality of the software is remarkably similar to dMARS, a coordinated suite of intelligent agents that can monitor, troubleshoot, and reconfigure the network aboard a NASA space shuttle and the planned space station.
(EEEK! Network administrators.)
. . . And in the Factories!
Take a coordinated suite of intelligent
agents that can monitor, troubleshoot, and reconfigure the routers, servers,
modems, etc., that make up a telecommunication/data network, and put them
to work in a factory, office, or "smart home" monitoring, troubleshooting,
and configuring machinery and appliances and you end up with intelligent
buildings and with factories that can run themselves. That is what Echelon,
a company that sells industrial control networks, is all about. The factory
manager, the line engineer, the process control specialists . . . all will
one day join the communication network manager in the dole queue.
(EEEK! Factory managers, factory workers, industrial systems engineers.)
And the answer is . . .
Stephen Hawking, the Cambridge
University physicist and author of the best-selling A Brief History
of Time, predicts that we will have the complete answer to the origin
of the Universe within 20 years. Hawking will be discussing his views at
a series of lectures about the Milennium this week at White House. The
lectures were broadcast over the Internet -- the RealAudio (and
video) of Hawking's lecture is at Sun
Microsystems Web site. Hawking says that complexity is where
the action is going to be. "It is in this complexity that I expect the
major developments of the next millennium. I foresee biological complexity
increasing with genetic engineering and the development of self-designing
computers with artificial intelligence. The future certainly won't be static."
The importance of complexity, the fact that we are making progress in understanding and predicting it, and the notion that the next thirty years are going to be absolutely mind-blowing, are familiar refrains on this Mining Company site, but it's nice to know we're in such august company.
(EEEK! Physicists.)
Credit Card Fraud
Version 4.0 of HNC
Software's Falcon credit card fraud neural-net software has
been released. The new version meets year-2000 compliance criteria and
includes several performance enhancements. Falcon is used by 16
of the 25 largest credit card issuers in the world, and a majority of
the top 50 in the United States, according to the American
Banker. An HNC executive claims that "under some circumstances,
the average U.S. client might see as much as an 80% increase in dollars
saved."
(EEEK! Fraud investigators, private detective agencies, credit underwriters.)
Customer Service
We've come across many instances
of neural net or CBR (case-based reasoning) software being increasingly
used in a customer support role. Business Wire of February 26, 1998,
reports another one. Dell Computer Corp. has adopted Inference
Corp.'s CBR software to assist staff in its U.K. and Ireland call center
to answer more queries on the first call by prompting them with questions
and guiding them through a knowledge base as they speak to customers. "Our
technical support staff currently resolves more than 90 per cent of customer
queries on the telephone. Using CBR for diagnosis has further improved
our first time fix rate, where we solve the problem in a single telephone
call", said Dell's customer service director. "This is a win-win situation
for everyone -- faster diagnosis means an improved service for our customers,
greater job satisfaction for our staff and a more cost-effective way of
doing business for Dell." Win-win? The Inference system currently runs
on Dell's intranet, so only staff can access it directly, but the plan
is to let customers access the software themselves, on the Internet.
Picture this (I'd draw it for you, but I'm no cartoonist): A bunch of unemployed customer service reps waiting in line for their turn to get help from the automated unemployment service representative. The fellow at the front of the line turns to the chap behind him and says, with pride, "You know, I taught that machine everything it knows." Now there's job satisfaction, for you.
(EEEK! Customer service representatives, technical support staff.)
SaxEx
No, don't get all excited. This
is about music. Researchers at the Artificial Intelligence Research Institute
in Barcelona have developed a program called SaxEx,
which allows your computer to play music like Charlie Parker. The only
input it needs is the sheet music, yet it can add all the nuances
humans employ when blowing the sax, by drawing on a repertoire of performances
by real saxophonists stored in memory.
EEEK! Saxophonists.
And on that wailing note, let's
wrap it up for this week and return next with more of the same. Thanks
for reading.
Until
next week,
NEXT WEEK: Another edition of Intelligent Machine Post.
Help Wanted: Got questions or comments on this article or on any other AI-related subject under the sun? Post it in the AIBB!