density chips and off-chip wiring—all of which make the mainframe processor — The personal computer does not re- guire very high speeds; therefore, CMOS chips that don't need expensive cooling can be used. As a result, PCs have high chip density, relatively few chips and little off-chip wiring. Another new technology that has ma- jor implications for the future is neuro- computer technology. A neurocomputer contains large numbers of electronic models of neurons with a high degree of connectivity between them. Neurocom- x ka ž i | ei : | i Z š Li - ; ia ] k rapid advances, MARTIN. giving organiza- puters are now available on boards that tions significant can be plugged into conventional per- opportunities to gain competitive advan- Z | tage. Many strategically oriented corpo- rations have used improvements in hardware and software technology to restructure their businesses, introduce new products, enter new businesses and improve development productivity. Important trends in computer hard- ware technology include the rapid : movement toward a distributed work- station environment, major increases in the processing speed of microprocessors, ultraparallel processing, specialized in- ference processors, neurocomputers, op- | tical-fiber communication, optical disks, o Talidigital high speed-telecommunica- — — aa tions and advanced microchip technolo- gy. It's important to gauge where this technology is going and what effect it will have on the future competitive - business environment. For example, changes are occurring very rapidly in microprocessor technolo- gy. The cost of chips in PCs has become much lower than the cost of chips in mainframe computers. This difference can be translated into the cost of doing a job. | The cost of mips (million instructions per second) on a large mainframe today is about two orders of magnitude great: er than the cost of mips on a desktop 80386-based PC. There are several rea- sons why this is true. One is that the lead time to build a PC is about a year, while the lead time to build a main- frame is about six years. We can get new technology into the PC much faster than it čan be implemented in a main- frame. The second reason is that there isn't much competition in the mainframe area. A lot of information-systems exec- - utives worry about the fact that there is inadeguate competition in top-of-the-line mainframes, while in the area of PCs there is absolutely intense competition, which significantly drives down the cost of mips. The third reason is perhaps more fun- damental: The mainframe is built out of bipolar technology, and the PC is built using FET (field-effect transistor) or CMOS (complementary metal-oxide semiconductor) technology. With bipolar technology, heat is generated all the time, With CMOS technology, however, heat is generated only when the compo- nent is switched or when information is read from memory. In the mainframe, we need high pro- cessing speed, which reguires bipolar technology and results in the dissipation of a great deal of heat. This leads to a — reguirement for a large number of low- sonal computers. These boards enable a PC to accomplish such tasks as the rec- ognition of the characteristics of a spe- cific face, which we cannot do at the present time with the world s fastest supercomputers. Although there are many potentially valuable applications of neurocomputers in business, most, business and informa- tion-systems professionals don't seem to know much about them. Some of the ap- plications are likely to be of major stra- tegic importance in business. Today's neurocomputers, although they can do things that the Cray 2 can't do in isolated applications, are neverthe- less very crude compared with what we can build in the future. An important PRED TELIGENC bri Affect Business Strategles Changes in Microprocessorf Chip Technology Could Have Major Implications for the Future KE z Rt — | a Mt. de dre. ži na dev Fi Me ča roda! [ zk hi | č ; 4 h Fuj zi zah — a E pta, a me Pika zi s k. ; yi. de daj , sta ? m | A a OE, ae ii ISTA ZAE EM ai EM. ča LAVA da oreh ra ha ki sy Pa s prva Sota! Nal ji v — Ne ari š , [ie 5 Ši še s Kale j j i? ud z x hej java pe pi s si ka Le A Ee AJA 1 zredi i i No ge z RE ee ii poši ij a— s ša d k hu £ k PA. č "a | ; 4 L z pa | k i. Bed ua struj SE | piš ad ič k m 1 a i 4: -, [ dl s > s i Na de ni . s ri j Li: ši a A ni kož ra A pe : zA pa ni ME o a Aa ČA ai got Meg be po o v k PN A AA ei H kk - - j ' di MLA rati! zi mi a j tai je ms Ke Min ie te dr ih so Hrig Hi it) pog k, a kala Poz t rane Na!) Maki UVA Pe are] z Pia s š ne eV a ea He ta h aha ino ER bide] bk A pini J za Me ma ARE rabe ae a a aaa Mi v EEA ZD Jz Ura A A pa iu Mih sr ha ua a ai h dia km : : mit MI - 1 ii; djčeg te 4. E s seči k, H l nj x h k 1 die . z po pa ie JE JA guestion is, How many neurons can we put on a chip or on a large-scale wafer? One of the problems with wafer-scale integration is that there is typically a lot of damage to individual components, rendering portions of the wafer inopera- ble. Perfect wafers cannot, be made with current manufacturing technology. How- ever, with neurocomputers, if some of the electronic neurons are dead, there is little effect on the overall operation of the wafer. Thus, there's a natural match between the operation of neurocomput: ers and the manufacturing technigues of wafer-scale integration. Similar arguments apply to making optical neurocomputers. Bell Labs is | JO d S J o chips with physical components e chips (hat - meastire o chips tHat OE AL LJ 4.1 ei Ka: ud bi! p dimi: r S sei pač Viri TR a ji rt kai aj . RR OSOKTa vrt ma x Y dia 1 ba Uh IV PE ni i di ui duse | mt fe spe s zma at kai poje ni [iča x ojla, Trio dai PP ad. ee ; ka z k ui - Li sei Ii ——- m. PRI dag J A Huje TI .unA ka h ; sj 7 x 4 E emo" 4 po de nisi ko iš ki dr O Gea a ei, a a IME merna, le paj i k ai U m a, ta ši a di si B dem ; ih im j : k ka H Mi: " Ke pli boki Jra Mt il. j ti HM pi: h. s Pe im nI IME M ji PU ču. Ha i He , de tai LI uši TI pare Moke ze h MA ia les NET naja s eluj stili HO NI z s xi? ke . leči keš ; ' a ki 1 s ; z i PRIPETI noj pa di JE Un aa PRI poe a di odi h j Ee -, ' LI " --. doni — ij mije PRIME ea ZA z RA ga ge e ca Ga: Pal ; It's important to gauge where advances in hardware and software technology are going, and what effect they ul have on the business strategies of companies in the future. working intensively on the development of optical neurocomputers. We can pre- dict how many neurons can be put on a wafer in the years 1990, 2000 and 2010. We can be fairly sure that, by about 2010, we can build a neurocomputer with as many neurons as the human brain. However, the electronic neuro- computer will operate 10 million times faster than the human brain! What applications can such an elec- tronic device perform? It won't be able to do the same things a human brain can do because we have such a poor un- derstanding of what goes on in the hu- man brain. It's likely the neurocomputer will appear simplistic compared with the human brain, but it will perform program to build a neurosupercomputer. EBRUARY 27, 1989 simplistic functions extremely fast. There are very important applications of neurocomputing in telecommunica- tions and in defense. Neurocomputers will enable us to build intelligent tele- communication networks and smart weapon systems that can recognize their target from a long distance away and home in on it. The Department of De- fense recently announced a $30 million There is another fascinating type of chip coming into existence. Its a chip that has physical components, not just logical components, and implements many different types of microsensors. For example, there's one chip which contains tiny whiskers, similar to tuning forks, that can detect low-level vibra- tions. Other chips measure the flow of gases or air. Another type of chip devel- oped by IBM can detect exceedingly sub- tle motions. IBM built a ballpoint pen with this chip in the tip of the pen. The pen is able to analyze the subtle motions associated with writing and can be used to detect forgeries. Over the last 15 years, we've seen the marriage of the computer and telecom- munications. In 1965, the telecommuni- cations industry was completely sepa- rate from the computer industry, and nobody thought that they would ever have anything-to do-with each other. By 1995 there's going to be a large amount of overlap between the companies that make things for telecommunicating and the companies that make products for computing. In a similar way, the television indus- try is completely separate from the newspaper and magazine industry, and both are completely separate from the computer business. However, if we look at those industries in about 2009, we will find a high level of overlap between the television and broadcasting business, and between the newspaper/magazine business and the computer business. Those industries are converging as we enter the age of hypertext and hyper media, intelligent television and intelli- gent. books. Every television will become a computer and every computer a tele- vision set. Again we need to ask, what are the implications of the merger of technologies? Who is taking advantage of it? Who is completely unaware that that is happening? Most importantly, if we put these dif- ferent, sets of facts about. technology to- gether, what machines can be built with those electronic components, what appli- cations will become possible, and what is the strategic importance of these ap- plications? Next, week, I will look at the effects of improvements in technology on so- ciety. M The James Martin Productivity Series, an %nformation-service updated guar- terly, is avaslable through High Pro- ductivity Software Inc., of Marble- head, Mass. (800) 242-1240. For infor- mation on seminars, contact (in the United, States and Canada) Technology Transfer Institute, 741 10th St., Santa Monica, Calif. 90402 (213) 394-8305. In Europe, contact Savant, 2 New St., Carnforth, Lancs., LA5 9BX Umited Kingdom (0524) 734 505.