|Volume 1: No. 18|
Unisys lost $1.3B in the second quarter, and plans to eliminate 22,000 jobs this year. (Nearly all of the financial loss is attributed to the cost of the downsizing.) [SJ Mercury, 7/24.]
IBM lost $1.7B, but is blaming it mainly on accounting quirks. The company plans to cut 17,000 employees this year, from a year- end workforce of 373,000. [Alan Gersten, UPI, 7/19.] (IBM employees are not laid off, of course -- they are given incentives to leave. I've heard that about a third of IBM's workforce are contractors rather than employees, and these can be laid off at any time.)
Silicon Graphics has introduced a new low-end system, the $8,000 Iris Indigo. (Its other UNIX-based 3-D graphics workstations cost $15K to $175K.) Sun has responded by improving its $5,000 SLC system and dropping the color Sparcstation to $7,000. Apple and HP have been dropping their system prices lately, and Compaq Computer Corp. has just cut notebook prices by 22% -- its third product reduction this year. [Lee Gomes, SJ Mercury, 7/23.]
Apple Computer shipped 85% more computers last quarter than they did a year ago, and is gaining market share with its competitively priced models. [Macworld, 7/91.] Due to declining margins, though, Apple has decided to back off on its excellent pay and benefits. It will aim for a position in the top 30% of computer companies instead of the top 10%. [Ron Wolf, SJ Mercury, 7/3.]
Congress is investigating whether Japanese companies in the U.S. are refusing to hire or promote women, minorities, and workers over age 40. Recruit U.S.A., for instance, has been fined $100,000 by the EEOC for seeking 25 young workers of Asian descent for IBM Japan. [Kristin Huckshorn, SJ Mercury, 7/24.]
DEC has agreed in principle to buy out N.V. Philips' European computer business, including imaging systems for document storage. Philips' PC activities are not included. [SF Chronicle, 7/24.]
Alan B. Lefkof, 38, has moved from president of Grid Systems Corp. to president of Farallon Computing. Reese Jones, Farallon's 32-year-old founder, remains as chairman and CEO. [SJ Mercury, 7/23.]
IBM has introduced a color CRT display that is pressure sensitive as well as touch sensitive. Resolution is 4,096 x 4,096 by 256 pressure levels. (Pressure is detected by sensors at the corners of the faceplate.) [Richard Doherty, EE Times, 6/17.]
NTT has developed a 15" 3-D color display that does not require special glasses. A lenticular screen directs separate images to your two eyes, with a head-tracking system used to adjust the picture alignments. Viewing is from 80 cm and can be done in a brightly lit room. Commercialization is not yet projected. [Junko Yoshida, EE Times, 6/24.]
Close to 1,000 Ph.D.s recently blew it, with GMU's Robert Sachs a publicly identified [and apologetic] casualty. Marilyn vos Savant recently gave a puzzle answer in her Parade magazine column, and 10,000 readers disputed her solution. It's the Monty Hall problem, first published in 1976 in American Statistician. There are three doors, one with a car and two with goats. After you choose a door, Monty will open one of the others to show you a goat and will then let you reconsider. [This constraint was not made clear in the Parade version.] You choose Door 1, and Monty reveals that there is a goat behind Door 3. Now, if given a choice, should you switch to Door 2? Or doesn't it matter? (Answer given below.) [John Tierney, NY Times. SJ Mercury, 7/22.]
Have you forgotten the password you used to encrypt a WordPerfect text file? John Gilmore of Cygnus Support (Palo Alto, CA) says he can work it out in about ten minutes. The key to breaking the Vigenere cipher is the presence of printer configuration information in predictable places. [Lee Gomez, SJ Mercury, 7/3.]
Neural Networks and Artificial Intelligence; a 1992 issue of Information Sciences. Solicited applications include neural logic programming, feature detection, knowledge representation, search techniques, and learning. Contact Subhash Kak (email@example.com) by 9/30/91. [George Georgiou (georgiou @rex.cs.tulane.edu), comp.ai, 7/23.]
Fault-tolerant computing; IEEE Transactions on Computers, May 1992. Submit papers to W. Kent Fuchs (firstname.lastname@example.org) by 7/1/91. (OK, I'm a bit late.) [IEEE Computer, 6/91.]
Case-based expert systems; Int. J. of Expert Systems: Research and Applications, March 1992. Evangelos Simoudis (email@example.com) by 7/1/91. [IEEE Computer, 6/91.]
J. of Computer and Software Engineering. This is a new quarterly starting in Fall 1991. E.K. Park (firstname.lastname@example.org) by 7/15/91. [IEEE Computer, 6/91.]
ACM Letters on Programming Language and Systems needs concise papers (5,000 words) similar in scope to TOPLAS articles. Charles N. Fischer (email@example.com). [CACM, 7/91.]
Group Decision and Negotiation is a new Kluwer Academic journal edited by Melvin F. Shakun, NYU. Material is sought in decision and negotiation support, AI and management science, applied game theory, and cognitive or behavioral sciences. [Katia Sycara (firstname.lastname@example.org), DAI-List, 7/24.]
Knowledge-based decision support systems will be the topic of a late-1992 issue of Decision Sciences. Submit papers by 2/1/92 to James Hershauer, Dept. of Decision and Information Systems, College of Business, Arizona State University, Tempe, AZ 85287-4206. [IEEE Expert, 6/91.]
Research & Education Networking (Meckler Corporation) is a monthly newsletter covering development, use, and impact of NREN, the Internet, electronic networking, publishing, scholarly communication, and other concerns of the academic, library, government, computer, and communications communities. Contributions are solicited. Erik Jul (email@example.com), (614) 764-4364. [PACS-L, 7/15.]
The Glass Ceiling is a new $150 monthly newsletter about women's legal struggle to enter the executive ranks. If you've got a story, call DataLine in San Francisco. [Lloyd Watson, SF Chronicle, 7/24.]
The Computer Science Lab at Xerox PARC (Palo Alto, CA) is looking for an experienced builder of computer hardware to work with hardware and software scientists constructing prototypes of future computer systems. Contact Mark Weiser (firstname.lastname@example.org). [This is for a Computist. Please post it to friends who might be interested. -- KIL]
UCalifornia is helping to establish the American University of Armenia in Yerevan, to be moved later to Abovian. Mihran S. Agbabian, USC's civil engineering chairman, has been proposed as the first president; UCB's Armen Der-Kiureghian as Dean of Engineering; and Stepan Karamardian, former UC Riverside dean, as Dean of the College of Business Management. Visiting lecturers in CS are needed for stays of 3 to 12 months. Contact Prof. Der- Kiureghian (email@example.com) [Diane Curtis, SF Chronicle, 7/22. Roupen Nahabedian (firstname.lastname@example.org), comp.edu, 7/24.]
A Silicon Valley startup working on multilingual OCR is looking for Mac or MS Windows engineers with AI, pattern recognition, and OCR experience. Contact Greg Buechler (email@example.com), Oryx Executive Search, Inc. (Sunnyvale, CA), (408) 749-1567. [misc.jobs.offered, 7/5.]
The PANGLOSS interactive machine-translation project needs an experienced R&D manager for a team at the CMU Center for Machine Translation, the NMSU Computing Research Laboratory, and USC ISI. Knowledge of NLP, LISP, and PROLOG is required. Contact Sergei Nirenburg (firstname.lastname@example.org). [misc.jobs.offered, 7/16.]
A Palo Alto R&D startup needs several MS/PhD computer scientists and software engineers. Contact Barbara Lynch (barbara @aurora.com). [Dov Rosenfeld (email@example.com), misc.jobs.offered, 7/16.]
Boeing Computer Services' Advanced Computing Technologies Center (Philadelphia, PA) has openings for NLP developers. Contact Krishna Nanda (firstname.lastname@example.org). [NL-KR, 7/17.]
The UEdinburgh AI Applications Institute (a tech-transfer organization with the Dept. of AI) needs professional staff members in AI modeling (decision support, qualitative reasoning, numeric modeling, simulation) and in knowledge-based scheduling. [Robert Rae (email@example.com), misc.jobs.offered, 7/18.]
The Software Verification Research Centre (SVRC) in Brisbane is looking for a Ph.D.-level Research Fellow for a team working in AI-based real-time verification of embedded systems. Contact John Staples (firstname.lastname@example.org) by 8/19/91. [Susan Dowrie (email@example.com), misc.jobs.contract, 7/24.]
GTE Labs (Waltham, MA) needs an experienced MS/PhD knowledge engineer to develop expert systems for automating company operations. [William Papp (firstname.lastname@example.org), misc.jobs.offered, 7/8.] GTE Labs also needs an experienced MS/PhD DAI researcher to work on interagent coordination for cooperative fault management and operations support. [Robert Weihmayer (email@example.com), misc.jobs.offered, 7/22.]
Engineering productivity tools unquestionably help bring in projects on or under budget, but they are best for problems that require cloning of old solutions. A 1987-8 study by David Murotake (GE Aerospace) and Thomas Allen (MIT Int. Center for Research on the Management of Technology) showed that use of complex tools during conceptual design tended to stifle creativity, whereas use during development freed time for creative work. Results might be different [or not] if the latest generation of creativity-enhancement tools were considered. [Michael F. Wolff, IEEE Spectrum, 1/91.]
A British study by Edward McDonough III (Northeastern University, Boston) has shown that experienced, "associative," systematic leaders excel in producing minor extensions of technology, but that creative, M.S. or Ph.D.-level leaders are best for ground-breaking product-development projects. Routine applications projects would seem ideal for exploiting project experience, but McDonough found that older (bored?) leaders were outperformed by younger ones. He recommends using these projects as training grounds. [Michael F. Wolff, IEEE Spectrum, 1/91.] (Does this introduce age discrimination?)
Businesses have nearly all the PCs they need. The personal computer market is still healthy, but channels are shifting to superstores selling name-brand and generic hardware. The software market will change if prices drop to match hardware costs. Most industry pundits are excited about multimedia -- and are again predicting the death of paper -- but I think that entertainment, user-friendly home applications, and access to electronic bboards will drive future software sales. Anything complicated is going to be a tough sell.
James J. Mitchell [SJ Mercury, 7/21] notes that software is becoming more important as the hardware market saturates. Low- cost computing is coming, and companies are working to establish their positions. Companies like Apple, Sun, HP, Octel, and LSI Logic are heavily into software development, often with more programmers than engineers. Venture capitalists are drawn by the annual growth rates of major software companies: Symantec's 1985- 90 sales rose 124% per year (!), Oracle 111%, Adobe 106%, Cadence 81%, Borland 58%, Software Publisher 30%.
"Borland people used to work 12 hours a day, six days a week. ... Now it's going to be 15 hours a day and seven days a week." -- Philippe Kahn, president of Borland International (Scotts Valley, CA).
"Almost anybody can start a software company. It's not capital intensive. There's no other business that Philippe could have been successful in." -- Gordon Eubanks, CEO of Symantec. [James J. Mitchell, SJ Mercury, 7/21.]
Last week, Matthew Witten asked about career opportunities for generalists and interdisciplinary scientists.
Mark Weiser (firstname.lastname@example.org) says that being a generalist has made for a richer life and has led to territory unexplored by others. Getting tenure was a problem, though, since he was not the best in any of his fields. Universities do not really value generalists. He recommends working on the problems of one discipline with the point of view of another; since there aren't many generalists, you're likely to be the first. Also, take your many areas all to heart, and try to have each thing you work on be significant to all of them. Otherwise some fields are just hobbies, and you are not truly a generalist. Mark considers himself lucky to be at PARC, where generalists are valued.
I'm reminded of a delightful book, The Way of the Ronin: A Guide to Career Strategy, by Beverly A. Potter, 1984. I'd like to share some notes that I've interwoven with my own views.
Workers form companies in order to accomplish more than they could individually. Companies also permit division of labor, with each member doing those tasks that most suit his or her own skills and personality. This should benefit individuals as much as it benefits the group. Cooperatives and communes are often based on such ideology; unfortunately, they lack the efficiency and stability of alternative systems.
In career feudalism, an owner forms the company and retains or delegates all authority. The entire company serves the owner, and reflects the goals and personality of its leader. Workers trade their autonomy for protection and economic support -- salary, job security, health insurance, retirement benefits, credit unions, journal subscriptions, etc. The good of the company (i.e., its owner) is paramount, but the company will take care of you as required by law and to the extent necessary to keep your job slot filled.
Feudal systems are bureaucracies, and bureaucracies are exceedingly efficient at repetitive tasks. Low-paid workers handle routine tasks (often involving material processing or customer contact); supervisors provide guidance and quality control; middle managers implement new policies and intervene when problems arise; and executives monitor outside forces and set strategic direction. Line functions (those producing revenue) form the core of the pyramid; outer vertical "faces" take care of purchasing, accounting, personnel, and other staff functions. (In an accounting firm, of course, accounting is a line function. At NSF, Physics and Chemistry are the core; Sociology and Education are further out.)
Bureaucracies achieve their efficiency and stability by dividing tasks into narrowly defined functions, each of which can be done by anyone with minimal skills. Machine-like workers become interchangeable parts in an organizational machine. The company need not invest in employee development, relying on personnel screening instead. Headhunters help by cannibalizing scarce parts from other corporate machines.
Unfortunately, we have tried to fit R&D into a bureaucratic mold. R&D is necessarily creative and integrative. Universities manage by maintaining a stable educational bureaucracy and letting professors do research on the side. R&D labs and software houses are more problematic because research results must be coupled efficiently to the needs of the company. That means that the company chooses the research problems, provides the resources (including salaried time), and must obtain results in a usable form. This coupling is at odds with creative research, and little pure research actually occurs in most companies.
We sometimes denigrate industry for jumping on the bandwagon of each new technology, hoping for a quick fix or the big win. (Neural networks is the current hot topic, with fuzzy systems running a distant second. Expert systems and genetic engineering are fading, or have at least slowed. Optical computing is making a comeback, and micromachines are waiting in the wings.) Well, industry is doing the best it can. Fad technologies are the only scientific domain where bureaucratic science can compete and succeed. Neural-network researchers appear plentiful and interchangeable at the moment, and even incremental results are likely to confer competitive advantage. Any breakthrough that can't be absorbed can easily be licensed to others.
Progress in traditional areas, on the other hand, is slow, dependent on scarce individuals, and unlikely to match immediate corporate needs. Even a critically needed research effort is dangerous if the loss of a key scientist would kill the project (or benefit a competitor!). Companies find it safer to wait until technologies are developed, then license the results. A U.S. company can probably license foreign technology more easily than it can tap the research of its U.S. competitors, so foreign research labs can be a significant aid to the U.S. economy. Note also that government facilitation is more important in mature research areas than in hot new ones, so you can expect NSF's emphasis to be on the trailing edge.
To have respect and job security in a bureaucratic system, a scientist must either specialize in an essential technology or be a generalist willing to adapt to current corporate needs. The specialist is tied to a specific worksite and technological milieu, and possibly to a specific customer. He (or she) is only valuable while still actively productive, and then only if that value is perceived and bankable by upper management. Fortunately, active publication increases visibility and establishes market value; unfortunately, it makes you less of a competitive asset and less trusted to remain with the company. (Publication helps to open up jobs in other companies, but don't count on this for job security. If you're looking for work, it may be that the entire field is closing down.)
An extreme case of specialization is the "wizard." These are people who know a resource -- a computer system, say -- inside and out. They become invaluable, and hence unpromotable as well. Some are superstars, likely to leave if not pampered; some jealously guard their domains by refusing to train successors. Such people make managers very nervous, but are usually retained as long as possible -- and hired as part-time consultants if they leave. Wizards are not on either a vertical or horizontal career track, but may make out pretty well financially.
Generalists have a better shot at job security, but don't get to do their own research. They hire people for specific projects, then fire them when the customer is sufficiently satisfied (i.e., the problem is solved) or dissatisfied. (Company policy may transfer scientists to other projects instead of firing them, but that conflicts with the nature of research and greatly lowers productivity. Startups gain tremendous advantage by hiring a product-specific team, then lose the advantage when the same team must be put to work on a second product. Long-established companies are notoriously poor at innovative research.)
Scientists in industry cannot choose their own projects. [Xerox PARC excepted.] Some welcome outside direction and the chance to solve important problems, others flee to universities where they can delve deeply into just those problems matching current interests and prior training. Unfortunately, the latter are very poor at selling their research results to those who might benefit. It's like the difference between journalists and the literati who write only for their own kind. We need something in between -- for personal gain, for professional satisfaction, and for the good of society.
Beverly Potter suggests that the best course, for those with the ability and disposition, is to follow the philosophy of the ronin. Ronin, or people who do ronin, were samurai with no masters -- by circumstance, by choice, or as an imposed sabbatical to learn new skills. They were without the benefits of feudal protection (salary, health care, pension), and thus had to live by their wits. All were outlaws, in a sense, and many survived as hirelings, mercenaries, or bandits. (Have you considered a sabbatical as a bandit?) Others found or created new positions in the feudal order. Many ronin became teachers at progressive universities, and helped to mediate contact with the West.
In modern terms, a ronin is one who makes lateral moves instead of following a traditional career path. Ronin may never make it to the top positions, or they may get there quickly and then leave for something else. They may start their own companies. If they can afford to, they may also drop out entirely. The pragmatic philosophy that leads to success also reduces the desire for success, giving the ronin a win no matter what the outcome.
The ronin (or samurai) philosophy is more than I can present here. Briefly, ronin are warriors, trained to disregard any consideration that does not achieve their goals. Many people experience life as a mix of blessings and curses; ronin see only problems to be solved and tasks to be accomplished. In war, the goals and problems are imposed externally. In peace, we must choose our own. The Japanese had several centuries of peace in which to contemplate the proper role of a warrior. I'll pass along a bit more in a later issue.
Your first pick had a 1/3 chance of being right. States CGG and GCG were equally likely a priori, but only in half of the CGG cases would Monty have [randomly] chosen to show Door 3. A posteriori, then, CGG is half as likely as GCG. There remains a 1/3 chance that the car is behind Door 1, with a 2/3 chance that it's behind Door 2. You should switch. (Not convinced? Get some playing cards and run a simulation -- as Monty himself did.)
If Monty were NOT forced to open a second door, or to give you a second chance, his decision to do so might depend on many factors; a purely statistical analysis is impossible. The answer would also be different if Monty didn't know (or care) where the car was, but vos Savant's problem said that he did know.