Intelligent Systems Report €January €1997 €Volume 14 €No. 1
Intelligent Systems Report €January €1997 €Volume 14
€No. 1
Editor
It wasn't supposed to have taken so long. When the whole idea of artificial intelligence first began to be seriously discussed, back in the late 1950s and early 1960s, people actually expected to see results within a decade, such as real-time speech recognition systems, Grand Master-level chess programs and self-navigating vehicles. Today, thanks to our 40-year advantage in hindsight, we can look back and chuckle at how little those early AI pioneers really understood the enormity of the tasks they had set for themselves.
However, it's important to remember that when the nascent AI field was getting started, space exploration was also just beginning. Who back in 1960 would've thought putting a man on the moon would prove to be much easier than developing an intelligent robot?
At the Dartmouth Conference of 1956 (the celebrated birthplace of the AI field), the early AI pioneers -- McCarthy, Minsky, Newell, Simon, et.al. -- set forth this basic proposition: Every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it. While the day-to-day shifts of the AI field often seem to be as capricious as the wind, the overall goals remain rooted in that simple statement of simulating intelligence in a machine.
The AI field has made enormous strides in its 40 years, and 1996
bore witness to many important events in the intelligent systems
arena. Following is our annual look-back feature (also known as "the
editors take their Christmas break").
Item: VR developers embrace VRML 2.0
Perhaps the biggest impact on the virtual reality field within the
past couple years has been the Internet, particularly the World Wide
Web. VRML -- the Virtual Reality Modeling Language -- was developed
as a language for describing multi-participant interactive
simulations -- virtual worlds networked via the global Internet and
hyperlinked with the Web. Last year VRML 2.0, also known as Moving
Worlds, emerged as an open, platform-independent specification for
dynamic 3-D environments on the Web.
Similar in concept to the core Web text standard HyperText Markup Language (HTML), VRML encodes computer-generated graphics into a compact format for transcription over the network. As with HTML, a user can view the contents of a file -- in this case, an interactive 3-D graphics file -- as well as navigate through other VRML "worlds" or HTML pages. In addition, VRML is infinitely scaleable so that as users navigate through virtual worlds and approach objects, greater levels of detail emerge.
With Silicon Graphics at the forefront, dozens of companies stepped forward in 1996 to embrace VRML 2.0, which leverages Sun Microsystems'
Java and JavaScript to create behaviors, motion and interactions.
VRML 2.0 also enables third-party plug-ins to be used unmodified,
adding useful 3-D capabilities to Web-based database, design and
other real-world applications.
Item: Wall Street still bullish on AI
In 1995 no fewer than 10 intelligent software companies went public
on various stock exchanges (usually NASDAQ). The initial public
offering (IPO) furor slowed down a bit in 1996, but not before one of
the best-known and most successful intelligent system vendors went
public: Gensym, developer of the G2 real-time expert system
development tool.
For those who appreciate ironies, Gensym's success was built upon one of the greatest AI failures of all time: Lisp Machine Inc. In the mid-1980s, LMI attempted to go head-to-head against the likes of Symbolics, Texas Instruments and Xerox in the Lisp workstation arena, but found the going a bit rougher than anticipated, and soon went bankrupt. Meanwhile, though, a small group of LMI software developers, who had developed an early expert system tool for process control called Picon, left LMI to form a new company -- Gensym. Over the course of the past 10 years, while most of the wretched excesses of the AI hype have long since been forgotten, Gensym has emerged as one of the dominant players in the intelligent systems arena. The number of companies relying on Gensym's G2 for intelligent process control is already impressive, and continues to grow.
Tecnomatix Technologies, a vendor of virtual reality systems for manufacturing applications, also went public in 1996. The company specializes in what it calls computer-aided product engineering (CAPE) tools, which are designed to allow production engineers to create an on-screen virtual manufacturing environment that graphically displays and simulates actual manufacturing operations. The simple fact that a company can thrive with such products indicates the tremendous strides VR has made in the industrial marketplace, as well as point up the rebound manufacturing technology has made in recent years (and right here it would well behoove us to sneak in a plug for our sister newsletter, Intelligent Manufacturing, which examines just such technological breakthroughs).
And as did General Magic and Verity before it, intelligent agent
vendor Edify decided to wade into public waters. Edify's IPO was
buoyed by an agreement to support Microsoft's Open Financial
Connectivity (OFC) specification. This OFC support enables banks to
use Edify's Electronic Workforce agent software to deploy electronic
banking information and services to users of Microsoft Money. Shortly
after its IPO, Edify introduced its own intelligent financial
solution, the Electronic Banking System. We'll have more to say about
electronic commerce in the next item.
Item: Intelligent agents help banks go online
1996 saw the emergence of the World Wide Web as a new channel for
banks to deliver various services to their customers. Indeed, by
year's end, some "virtual banks" had been established that existed
entirely on the Web -- no front lobbies, no drive-through windows, no
free toasters -- just banking as close as your fingertips and your
PC.
As noted above, Edify helped give electronic commerce a healthy
nudge by launching its Electronic Banking System (EBS), which
combines Web banking services with intelligent agent-based electronic
banking capabilities that support self-service via telephone, fax and
personal financial management software. EBS makes possible such
applications as home banking, bill payment, dynamic target marketing,
and customer service teleconferencing.
Item: Kasparov defeats Deep Blue
Intelligent chess-playing machines have fired the public's
imagination for quite some time. In fact, more than 200 years ago, in
1769, a Hungarian noble named Baron Kempelen invented the Automaton
Chess-Player, which totally captivated Edgar Allan Poe several
decades later when the device was shown throughout the U.S. More
recently, in the 1950s, noted mathematician Claude Shannon wrote the
first paper on computer chess, and soon afterward such AI pioneers as
Allen Newell and Herbert Simon were writing chess-playing programs.
As the level of play improved over the decades, a wealthy inventor named Edward Fredkin established a prize in the late 1970s: $100,000 to the developers of a computer program that could defeat the reigning world chess champion. Needless to say, to date the Fredkin award has gone unclaimed.
Last February, though, the almost unthinkable finally happened:
IBM's computer chess program Deep Blue actually defeated Grand Master
Garry Kasparov in regulation match play. True, Kasparov quickly
rebounded in the six-match competiton, winning three times and
playing to a draw twice. But that single victory of Deep Blue's
brought more attention to AI research and parallel processing than
any other display of AI since the Reagan "Star Wars" days.
Item: Neural networks fight cancer and TB
While Deep Blue grabbed most of the "techie" headlines in 1996,
another intelligent program was performing what some considered a
medical miracle: Neuromedical Systems' PAPNET neural network-based
system for detecting cancer. As originally developed, PAPNET uses
neural networks to detect pre-cancerous cells on routine cervical Pap
smears, has been cleared by the FDA, has been accepted by the
American Medical Women's Association, and is in use nationwide.
Now comes word that PAPNET is being adapted to help detect lung cancer as well. In a recent study at a New York hospital, PAPNET testing correctly identified 97.1% of smears from patients who had lung cancer.
Not to be outdone by Neuromedical Systems, another company --
Physical Optics -- has developed the SSMART neural network system to
detect and identify Tuberculosis and other mycobacterial infections.
Researchers at various university centers are also working on neural
systems to diagnose coronary artery disease and pulmonary embolisms.
Even emergency rooms have fallen under the purview of neural
networks, as a scheduling program has been developed for
perioperative management.
Item: Speech recognition technology hits the big
time
Lernout & Hauspie Speech Products hit the jackpot last year when
software giant Microsoft licensed its speech recognition technology.
Microsoft intends to use L&H's text-to-speech software in future
products, and will use L&H linguistic components for the
development of speech modules. This technology enables electronic
equipment to convert written text to human-like speech.
Meanwhile, two other leading speech recognition vendors signed a merger agreement. Voice Control Systems acquired Voice Processing Corp. in a pooling-of-interests transaction valued at over $40 million, marking the second time in recent years that VCS has gobbled up one of its competitors; in 1993 VCS acquired Scott Instruments.
VCS and L&H, in partnership with a number of other speech
recognition vendors, also announced the development of the Speech
Recognition Application Programming Interface (SRAPI), a standard for
Windows and Java platforms. SRAPI is a cross-platform specification
that seeks to create a standard method for applications to interact
with speech technology providers.
Item: VR cleans up nuclear wastes
And finally, in our salute to the most unusual application of
intelligent systems we uncovered during the previous year, we're
giving the nod to TRW and its CAPS (Characterization Analysis
Planning System), which uses virtual reality technology to gather
sufficient information to decontaminate old buildings once used for
producing radioactive materials for atomic bombs during the height of
the Cold War. Sounds pretty straightforward, until TRW's engineers
explained that the decontamination process includes cleaning up bird
droppings and clearing out poisonous snakes. Anytime you can work the
words "bird droppings" into a story, it's gotta be unique!
And with that, we wish you all a very happy and successful new year!
More News, More Often in '97Beginning this month, our parent company, Lionheart Publishing, will launch a Newsroom page on the World Wide Web that will offer news and information, covering a variety of topics, on a daily basis. The INTELLIGENT SYSTEMS News Page, for instance, will include news, product announcements, conference information, and similar material. Regular ISR subscribers -- both hard copy and online subscribers -- will continue to receive the same amount of news and features they've come to expect from ISR over the years; think of the Newsroom as a bonus, since it will feature some news items that might otherwise never see print. We also invite you to check out some of the other Newsroom pages as well, especially those featuring news of interest to readers of our sister publications, INTELLIGENT MANUFACTURING and ELECTRONIC COMMERCE UPDATE. And by all means, please ask your colleagues to check out the Newsroom as well. |