The Intelligent Machine Post

Issue 00000001

Dateline: January 25, 1998

EVERY week my scurrilous little searchbots roam the Net spying for scraps of intelligence about intelligence. Actually, they are not really my searchbots. The main ones I currently use belong to Excite and Inquisit. Excite offers not only a free customizable news page which comes up every time I connect but also a free news clipping service that scans lots of online news sources for my topic of interest--AI. Inquisit does essentially the same but goes deeper, and there is a charge for the service. Inquisit's search results are sent to me by email, and while I believe I could get Excite's that way, too, I choose to view Excite clippings directly on the Web.

And every week I post links to four or five of the more interesting clippings on the home page for this site. Sometimes, but not always, I save a link as an html file for future reference, and occasionally I try to search back for something I remember posting a link for, but can't find it. So, I figure, why not put everything into a newsletter that will be archived and indexed on The Mining Company server and will therefore be full-text-searchable.

So here's my first effort, modeled somewhat after the highly successful and very useful Edupage service. I have culled the interesting bits from the contents of my clipping files, paraphrased them, given appropriate attribution, and added a comment of my own here and there. If you find this useful, or if you have something to report, please let me know.


Talking Genes
Geneticist Anthony Monaco and colleagues at Oxford University have found a gene controlling human speech and have named it "SPEECH1," reports Maggie Fox for Reuters. The finding means that we are "hard-wired" for speech; a finding corroborated by the ability of deaf children not taught sign language to construct their own languages, complete with grammar and sentence structure. This is pretty exciting, not only in offering hope of eventual gene therapy for sufferers of various speech problems, but also for AI researchers in the sense of showing that language capability can in principle be coded in machines.
 
Dr. Mysto
This children's education freeware for Windows 95/NT is described as an "intriguing application of artificial intelligence to the old Twenty Questions game." Downloadable from its author Frank DiNunzio's Web page. I haven't tried it--if you do, let me know how it goes.

AI for Power Stations (1)
An AI-based program that runs under Windows was named Best Customer Service and Support Solution and winner of the prestigious Microsoft Industry Solutions Awards for 1997. A San Diego electrical power system design software company uses PHD's Professional Help Desk(TM) to deliver technical support to nuclear power plants that use its software. Professional Help Desk incorporates "highly advanced artificial intelligence to resolve complex, client-specific problems simply and quickly," says the blurb. Check out the Microsoft Industry Solutions Awards site for more info. 

AI for Power Stations (2)
Intelligent software from HSB Group Inc. takes data from accelerometers attached to a generator's turbine and determines if the turbine is singing off-key, which can happen if a turbine blade develops a crack or the shaft develops a wobble. Undetected, a crack or a wobble can turn into a catastrophe. The accelerometers read the vibration frequencies, and the AI software analyzes the data from those readings and figures out what's wrong and how serious it is. The automated system works 24 hours a day and never whines, unlike the army of testy engineers who must otherwise be sent to inspect turbine blades regularly. Now, the engineers can put their feet up and whine about the national debt while waiting for the HSB software to tell them when something needs fixing, what needs fixing, and how urgently it needs fixing. It also points out the costs of replacing or repairing the defective parts, a terrific aid to management decision making.
 
This is Not Cricket
The Daily Telegraph's Simon Davies uncovered (December 16, 1997) a European Commission report alleging that the United States can eavesdrop on every telephone, email, and telex communication around the world. Assessing the Technologies of Political Control, commissioned by the Civil Liberties Committee of the European Parliament, apparently describes a U.S. network of spy stations on British soil and around the world. It is called Echelon. Don't ask why. Echelon  "routinely and indiscriminately" monitors phone, fax, and email messages, and was designed primarily for non-military targets, according to Davies. After randomly intercepting volumes of communications, Echelon uses AI software "like MEMEX" to look for key words. Maybe the National Security Agency's Web site (an oxymoron if ever there was one) has more info, but I doubt it, and anyway, don't tell me--it may be too dangerous to know.

On the other hand, you have to sympathize with the poor old NSA, toiling through the midnight hours and burning billions of taxpayer dollars on Echelon only to have some bearded erk post a 128-bit encryption key for fun on the barrier-free global Internet, nullifying, with a sly stroke to silly-mid-off, the whole daft game.

Blade Runner
Ridley Scott's science-fiction movie has been morphed into a huge (four CDs' worth) interactive 3D game. You, the player, set the parameters for the storyline, and the 70 characters in the game are endowed with their own artificial intelligence (AI), so they are able to make decisions on their own and communicate with one another without your knowledge. As a result, the game is unpredictable and very different every time you play. I haven't played it, but the reviews of those who have (in the Irish Times, Capital Times, Milwaukee Sentinel & Journal, and AP Online) are very positive. The AP reviewer noted a particularly significant twist: at some point in the game, the player faces an explicit moral dilemma, and is forced to decide whether to be a good guy or a bad guy.

CA Buys AI WARE
Computer Associates International, Inc. (CA) has purchased privately-held AI WARE, Inc., a developer of intelligent decision support software. CA will incorporate of AI WARE neural net-based technology into CA's product line to enable clients to make "accurate business forecasts based on historical data," and also to incorporate predictive management and enhanced troubleshooting support facilities into its Unicenter TNG enterprise management systems.  AI technology will also be integrated with CA's application development offerings, including the Jasmine database, for handling complex systems in many industries and vertical markets.
 
AI May Help or Hinder Your Mortgage
Stewart Online, a provider of loan closing document technology, recently announced the appointment of Peter A. Renfro to their management team as senior VP and chief technology officer. What's interesting about this is that Renfro has extensive experience in AI underwriting technology in the mortgage industry. It is another example of Machina sapiens worming its way into all aspects of life, and writing on the wall for human mortgage underwriting officers.

Medical Roundup
Zeneca Pharmaceuticals has developed a virtual reality (VR) simulator designed to give physicians some idea not only of the physiognomy of a migraine attack but also what it feels like. The simulation takes place inside a tilting, vibrating flight simulator-like pod, complete with sound effects, flashing lights, and 3-D images.

"Telesurgery," a method by which it is hoped surgeons will soon be able to operate on patients halfway around the world using datagloves, high-speed communication links, 3-D imaging, and cameras and robots at the patient's location, has advanced to the point where animals have been operated upon from a distance of five kilometers and a human patient had a gall bladder removed from across the room. The main holdup is the speed of the communications link. A DARPA spokesman says that the maximum distance for a wireless link is 50 miles, and 200 for a cable connection. Internet2, a U.S. National Science Foundation-funded research project, is designed partly to find ways to increase the speed of the Internet so it could be used, inter-alia, for telemedical purposes.

DARPA is also working on a five-dimensional total body scanner known as a "Medical Avatar." The scanner will produce enough data to enable a phsyician to examine a living, breathing, holographic VR replica of a remote patient, even reaching inside to feel the heart beating or feel a broken bone.

Though DARPA reckons we are still 25 years away from full-blown telesurgery and telemedicine, it is already here in limited form. Doctors use videoconferencing networks to view x-ray images or examine a skin lesion of a remote patient. VR already helps patients overcome claustrophobia, agoraphobia, fear of public speaking, fear of heights, impotence, premature ejaculation, and to control pain for burn patients and women in childbirth.

Circulation, journal of the American Heart Association, has reported that an AI program combining neural net technology with expert system technology did better than a group of cardiologists in diagnosing heart attacks. The neural networks were found to be 10 percent better at identifying abnormal EKGs than the most experienced cardiologist.

It may appear that telemedical technologies have a future as tools for human doctors and I would agree--but only for the short to medium term. In the long term, computers will make better diagnoses and robots will make finer incisions. In short, they'll make better doctors. Now, if only we could get DARPA interested in lawyers and politicians . . .

  Until next week,

       

NEXT WEEK: IMP #00000011.

Previous Features