Complex vs. simple. Big vs. small. Centralized vs. distributed. Control vs. independence.
There's no doubt on which side corporate IT and its vendors belong. While paying lip service to simplicity, corporate IT keeps building huge, complex systems that require central administration and control.
Yet everyone knows that computing's epoch-making innovations -- Macs and PCs, the Internet, the Palm -- began as simple technologies that needed no central coordination. Chances are, that's how the next innovation in computing will emerge.
Which is why anyone interested in IT innovation should pay close attention to Patricia E. Moody, a manufacturing expert, and Richard E. Morley, a world-renowned entrepreneur and inventor whose work is on display at the Smithsonian Institution.
In the vision of computing's future laid out in their recent book, The Technology Machine: How Manufacturing Will Work in the Year 2020 (The Free Press, 1999), today's big systems lose out to simpler, faster ones that won't need years of expensive implementation followed by intensive supervision.
As the authors see it, tomorrow's factories won't be run by big, complex, centrally run systems like ERP and MRP. Instead, factories will run themselves with little human supervision, using flocks of small, independent, intelligent systems.
Think of it. No human brain could create a system to accurately plan the individual transportation needs of hundreds of thousands of New Yorkers. Yet thousands of taxi drivers operate without central control by following a few simple rules: Stop when someone raises his arm, take him where he wants to go, collect payment and pick up your next job. Overall, that "system" works well.
Similarly, Moody and Morley envision that a host of independent devices embedded in factory equipment, programmed to follow a few rules and to make snap decisions, will run factories. This approach to manufacturing is already a fact of life in a few special plants, such as GM's Fort Wayne, Ind., paint shop. Running a paint shop is an unusually tough job; scheduling is complicated, and a jammed paint gun can throw off the best-laid plans. But the GM plant minimizes scheduling. Instead, a simple local software program, not a mainframe or human, allows the paint booth to select and sequence the work. As each truck comes down the line, the program tells the booths which color the truck should be. The booths, like rival contractors, then place "bids," based on whether they have the right color and amount of paint for the job. A booth loaded with the color the truck requires bids high, a booth low on paint bids lower, and one loaded with a different color bids lowest. The booth with the highest bid gets to paint the truck.
Moody and Morley say the same localized intelligence approach will someday create workflow systems, automate nonmanufacturing business processes and simulate and improve complex systems. Already, Japan's famous bullet trains use this approach to simulate its national railway system (in this program, the "trains" are software agents following a few rules) to improve train schedules and help decide where to build new lines. Moody and Morley write that simpler, smaller systems will enable localized manufacturing: The television sets or PCs you buy at Wal-Mart will be custom-built there by small, portable factories parked behind the shipping dock, staffed by a handful of workers.
Give individual machines just enough intelligence to follow a few rules, then let them do their thing: What a different direction from today's bloated, burdensome brand of computing! Think about it.