(FORTUNE Magazine) – Bell Labs gave birth to the transistor. And the laser. And motion pictures. And long-range TV broadcasts. And real-time language translation. And on and on, so that over time this venerable institution rightly became known as America's invention factory. But the next transistor may well come from a small company you've never heard of.

Meet Mark Reed (pictured at left), 45, co-founder of Molecular Electronics Corp. and a digital hero in the making. Four decades after Intel co-founder Bob Noyce embedded Bell Labs' transistors in silicon to create the first manufacturable integrated circuit, Reed is leading an effort to put computer chips in the shade by making so-called molecular transistors, which are thousands of times smaller and 100 times faster than their silicon predecessors. Incorporated into tiny supercomputers, these new transistors could bring unprecedented intelligence just about anywhere.

The molecular transistor is only one of a number of epochal discoveries that are emerging from small entrepreneurial companies. No longer do the likes of Bell Labs and GE point the way to technology's future. Big corporations are cutting back on basic research to concentrate on shorter-term goals. More than anytime in the past, new outfits are springing up all over the place to drive technology forward. They have scored with other big breakthroughs such as these:

--Materials Innovation Inc. of West Lebanon, N.H., has not only drastically reduced the size of a press that makes industrial parts from metal powder but has also developed new powders for production of components that increase the efficiency of electric motors by 30%.

--Bios Group of Santa Fe has used complexity science--the study of such seemingly far-fetched subjects as how foraging ants find food--to improve supply chains and factory outputs, putting Procter & Gamble on track to cut inventory by 25%.

--Catalytica Combustion Systems Inc. of Mountain View, Calif., is equipping electricity-producing gas turbines with a catalyst that eliminates pollutants before they form.

Corporate R&D isn't dead, to be sure. In a few exceptional cases, such as Xerox's Palo Alto Research Center (PARC) and Du Pont, basic research continues. But ever since the cost-cutting fad triggered a bloodbath at the big labs in the mid-1980s, which saw thousands of researchers unceremoniously forced to leave for university and other jobs, corporate R&D has been shifting more and more from the "R" part of the equation toward the "D." (The Defense Department didn't help by scaling back its R&D spending, either.)

The change in priorities has continued even though overall R&D spending has rebounded. According to the Industrial Research Institute, an organization of FORTUNE 1,000 research directors, corporate R&D budgets increased from $97 billion in 1994 to $166 billion last year. Yet the innovative content of corporate research has been greatly diminished in favor of work that supports short-term goals, such as improvement of existing products. A few companies simply closed their R&D labs; others atomized central labs by assigning the researchers to divisions. Says Arthur Chester, president of the Hughes Research Laboratory in Malibu, Calif.: "Basic research has been on the decline in large organizations since the mid-1980s, as more companies have routed more of their research funding through their business units, which have more specific goals." His lab, once part of Hughes Aircraft, now serves as a corporate research facility for Raytheon and Boeing (which own it jointly), and also does research for GM. Adds Chester: "The decline has been more pronounced in some industries, such as aerospace and electronics, than in others, such as chemicals and pharmaceuticals."

Companies with R&D labs are trying to expand and enrich their activities in other ways. Paul Sheldon, former R&D director at machine-tool maker Giddings & Lewis, says, "Corporate R&D has both opened up and shut down." By "shut down," Sheldon means the diminution of internal research activities. By "opened up," he means greater collaboration with outside research organizations, not only small companies but also national labs, private research institutions, and universities. (See accompanying story, "Where R&D Is Thriving.")

As a result, MIT, the nation's leading research university, has seen a flip-flop in its sources of financial support. MIT used to get two-thirds of its outside support from Uncle Sam and one-third from industry. Now the proportions have been reversed. While five years ago MIT received $56 million in research funds from corporations for sponsored research on campus, last year the figure reached a record $73 million.

But purchasing research conducted by others is no panacea. "It's been my experience that buying from the outside is risky unless you have enough in-house expertise to be a smart buyer," says Hughes Lab's Chester. "And technology developed elsewhere--for example, in a university--usually needs adaptation and additional development before you can use it in your business." In-house expertise is getting harder to come by too. According to Karl Koster, director of corporate relations at MIT, "Big companies used to have their pick of MIT students. Now they are having a hell of a time recruiting them."

With the reorientation of industrial labs toward immediate results, the fear is that the U.S. may never duplicate the kind of critical mass achieved in years past by Bell Labs, where top scientists in different disciplines collaborated closely with one another. Corporations are shunning basic research because results are unpredictable and cannot be prescribed. This year's three winners of the Nobel Prize in chemistry discovered largely by accident how plastics could conduct electricity. That's the way high-temperature superconductivity was discovered as well, along with many other advances.

To be sure, basic research continues in academia, and the U.S. boasts the world's best research universities. But even Arno Penzias, a convert to "practical" science in his later years as research director at Bell Labs, scoffs at the idea that part-time research professors assisted by part-time graduate students can match the brilliant staff of the old Bell Labs. Universities are still split into departments in which, for example, physicists seldom work with chemists, while progress in science most often occurs at the intersection of disciplines.

Luckily, America has a great history of responding to challenges with vitality and resilience. New players are stepping forward to create innovative technologies outside corporate labs.

Technology-savvy venture capitalists, both in and outside Silicon Valley, are eager to play a role in forming new R&D machines. They've just about had their fill of helping start those sometimes shallow dot-coms. Explains Yogen Dalal, a partner in the Mayfield Fund of Menlo Park, Calif.: "There has been little fundamental technology in the B2C, B2B, and other applications startups." Instead, these companies trade on their knowledge of particular industries, using their insights to make plain old marketing land grabs. "As we return to fundamentals," adds Dalal, "VCs are looking for the next wave of real technology."

Renowned veteran R&D executive George Heilmeier, who has served as director of the fabled Defense Advanced Research Projects Agency (DARPA) as well as director of R&D at Texas Instruments, and more recently as CEO of Telcordia Technologies (formerly Bellcore), notes that the shift to venture-backed small companies in many ways addresses problems that have always plagued corporate R&D. Among them: Many CEOs don't understand R&D and delegate responsibility for it to the research director, who often doesn't participate in major corporate decisions. Big companies frequently view R&D as an expense rather than an investment. Division managers at large corporations are often in conflict with R&D directors because central R&D is outside their control.

VCs know what it takes to commercialize technology in a competitive marketplace. Today they have a great spectrum of small companies to pick from. Everyone and his cousin who has anything to do with R&D is trying to start a technology company, collectively spawning offspring as if they were egg-laying octopuses.

But it's not just trend-spotting venture capitalists at work here. High-tech startups are being formed by national labs too. Sandia National Laboratories in Albuquerque, for instance, boasts that it has launched 41 companies (with VC help) since 1994. Private research institutions are also playing a role in company formation, in the process creating a financial lifeline at a time of scarce funding for basic research. SRI International and its subsidiary Sarnoff Laboratories have founded 31 small companies. Some of these initiatives are beginning to pay off big, not only technologically but financially as well. Battelle of Columbus, Ohio, the largest nonprofit research institution, recently collected $148 million in stock for its minority interest in a small fiber-optics company that was sold to a bigger one.

Consulting companies are financing R&D ventures as well. Bios Group, for example, owes its origin to, of all entities, the accounting firm Ernst & Young, now known as Gemini Cap Ernst & Young. Christopher Meyer, the outstanding director of Ernst & Young's center for business innovation in Cambridge, Mass., became enamored of complexity theory in the 1990s and persuaded his company to help start Bios Group in 1997; Ernst & Young invested $5 million in return for 40% of the stock. Since then Bios has attracted other investors, such as P&G and Ford Motor. "Our mission," says Meyer, talking about his center, "is to anticipate and shape the future, like the old Bell Labs did, only on a very small scale." Some new companies are self-financed. To start Materials Innovation, Alan Beane, a former CPA who made millions from the sale of his computer accessories company, dipped into his own pocket.

The success enjoyed by the research startups has not gone unnoticed by large corporations. Many have recognized that in their panic to align R&D to immediate needs, they have cut the future out of the equation, and they are now beginning to form venture firms of their own to conduct research. Heilmeier, the former DARPA director, is encouraged by this trend. "The story here is that corporate R&D is breaking out of its mold," he says. "The venture companies remove the problems endemic in big R&D organizations." The small fry are able to focus much more sharply on the task at hand. And some big-company scientists and engineers assigned to run corporate spinoffs reap undreamed-of rewards when the companies go public or are sold. (Company executives don't like to talk about this, apparently for fear of agitating scientists and engineers who didn't participate in the payout.)

Lucent Technologies, the parent of Bell Labs and one of the record setters in the number of companies spawned (31 since 1997), finds the technique so attractive that it recently reeled back three small companies into Lucent after they had sufficiently advanced technologies in Internet telephony, digital video broadcasting, and electroplating parts for electronics. But Lucent was willing to be patient. Back in the late 1970s and early 1980s, when "intrapreneurship" was all the rage at big companies, the idea generally failed because corporations were unwilling to leave their offspring alone. "Today," says Thomas M. Uhlman, president of the new-ventures fund at Lucent, "companies are much more relaxed. And as a result, they benefit more."

The research labs are not ready to cede the future to their upstart brethren, but even in achieving a radical breakthrough, their motives and methods are in stark contrast to the ways of the past. Perhaps the place one would least expect to find a tightly knit entrepreneurial team is inside Bell Labs. Yet David Bishop, director of micromechanics research there, led a group of 14 scientists and engineers who worked day and night last year to come up with the first optical switch, the Lambda Router. Optical switches direct photons to their destination without converting them into electronic pulses, as conventional switches and routers do. They are ten times faster than their electronic counterparts and, claims Lucent, 25% cheaper to operate. With the explosive growth of traffic on the Internet threatening to create gridlock, optical switches are widely expected to replace electronic ones. The Lambda Router will start carrying traffic across the Atlantic on Global Crossing's fiber-optic cable by the end of this year. By being first, Lucent hopes to capture a big share of a market that industry analysts say will reach $4 billion a year two years from now, and challenge router-colossus Cisco Systems in the process.

Bishop's team has operated very much like a small, independent company. He proudly shows off a tiny factory inside Bell Labs where Ph.D.s work as high-paid assemblers, putting those switches together. One factor that energized his scientists, Bishop adds, was Cisco CEO John Chambers' description of a Bell Labs-designed Lucent telephone as having been built by "dinosaurs." Says Bishop: "He probably doesn't know that dinosaurs were warm-blooded and fast-moving. Now our mission is to kill Cisco." What could better illustrate the dramatic change in the nature of the big R&D organizations?

Good work will continue to emerge from corporate labs, but the shift in brainpower means that more and more of the really striking advances will come from small, entrepreneurial ventures. Below are some of them.

Molecular Electronics Corp. The Transistor Of Tomorrow

When Bell Labs invented the transistor in 1947, it planted the seeds of the First Digital Revolution, which transformed business and industry as well as our lives. The Second Digital Revolution will spring from the realm of the invisible, and it may come from a company that didn't exist until last December. In contrast to the big corporate muscle behind Bell Labs, Molecular Electronics Corp. is the creation of scientists at three universities and a grab bag of businesspeople from disparate walks of life. Co-founder Mark Reed, who is also chairman of the electrical-engineering department at Yale, started the company with colleagues at Rice University and Penn State, and with the backing of a group of Chicago investors.

The CEO is Chicago entrepreneur Harvey Plotnick, 59, who studied physics at the University of Chicago and founded the largest U.S. book publishing company outside New York City, Contemporary Books, which he sold to the Tribune Co. in 1993. Molecular Electronics is building off-campus labs to house development teams in each of the three university towns. In two years, Reed promises, the teams will demonstrate the first working prototype of a molecular transistor.

Each transistor will consist of a single molecule of plastic-like material that, unlike today's semiconductors, will be made in flasks. The new transistors will be thousands of times smaller than their silicon predecessors and will work 100 times faster. Eventually they will take digital electronics into uncharted terrain where walnut-sized supercomputers will spread digital intelligence of a degree unattainable today throughout factories, offices, cars, and homes.

The startup is the outgrowth of a decade-long effort that culminated last year when Reed and his co-workers became the first to show that specially designed molecules can be switched on and off with a trickle of electric current. A small charge induces slight changes in the configuration of the molecules, blocking current flow. When voltage is cut off, the molecules spring back to their original shape, letting the flow of current resume. These on-off capabilities qualify a molecule as a transmitter of the ones and zeros of computer language.

Connecting these molecular switches, which are visible only through an electron microscope, is one of the big challenges facing Reed. Short of solving that problem, he envisages an intermediate molecular electronic circuit in which standard, ultrathin wires used in silicon chips will connect colonies of molecules set up in tiny depressions in a circuitboard.

Even the first such circuits should be a lot cheaper to make than silicon chips. That's because production of molecular electronic devices involves a kind of chemistry well established in the pharmaceuticals industry. Reed and his colleagues start with familiar compounds, such as carbon molecules used to make plastics. They then add reagents that bond with the target molecules at specific sites, gradually forming molecules with the desired electronic structure.

Next they attach a molecular fragment with a high affinity for gold to the bottom of the molecules. Finally Reed and his associates dip boards with a gold surface into a beaker rich with the molecules--and presto, the molecules attach themselves to the gold surface by the millions, lining up like trees in a microminiature orchard. This self-assembly is what should make molecular chips a lot cheaper to produce than their silicon predecessors.

Although big obstacles to mass-producing such devices and linking them into working systems still loom, Reed and his associates are confident of achieving that feat. One reason is that Reed knows a lot more than most university professors about submicroscopic manufacturing. Before Yale nabbed him ten years ago, he was a star researcher at Texas Instruments, where he pioneered the construction of devices even smaller than molecular transistors--quantum electronic devices, in which an electrostatic force confines a single electron into a structure called a quantum dot. In its quantum state, where it exists as both a particle and a wave, the electron sloshes about like water in a glass. Molecules, in contrast, are larger and easier to handle.

So far, Reed and his team have been able to demonstrate electronic responses only in single molecules--not enough of them, if strung together, to constitute a chip. But even in the earliest experimental stage, they have shown that a molecular switch, when used as a random-access memory device, can retain its data for nearly ten minutes. Silicon random-access memory devices, on the other hand, must be recharged every few milliseconds to hold their data. The new transistors also have a huge advantage over silicon devices in the amount of heat they produce. Molecular devices consume electric power at a trickle and hence generate 100 times less heat than their silicon counterparts. The latest microprocessors emit about 100 watts of power--as much as an average light bulb. Most need fans to cool them.

Who needs the transistor of tomorrow? Just about everybody. Although silicon has proved amazingly resilient as a material for building chips with ever smaller transistors, physical and financial limitations are expected to catch up with silicon in the next decade or so. The physical reality is that Moore's law--which holds that the number of transistors on a sliver of silicon will double every two years--is getting increasingly difficult to obey. The law was postulated by Intel co-founder Gordon Moore 35 years ago, at the beginning of the semiconductor revolution. To keep the law in effect, individual transistors on a chip would have to be shrunk to the point of invisibility, with each transistor holding a charge of only eight electrons, compared with thousands today. Such a minuscule number would be hard to confine, and leakage of electrons from one transistor to the next would make silicon chips unworkable. So experts expect Moore's law to expire by 2020.

Even before then, the lesser-known Moore's second law could cause insurmountable problems. This one postulates that the cost of building a chip fabrication facility, known colloquially as a fab, will double every three years. Since a fab already costs as much as $2.5 billion, in a decade or so the cost would soar to an astounding $30 billion to $50 billion--or 10% of the estimated revenues of the semiconductor industry in 2010 just for a single plant. The cost of building a molecular-device fab is hard to estimate now because of many unknowns, but Reed says that if it matched the price tag of today's chip fabs, "we'd be in the wrong business." Suffice it to say he thinks the cost will be considerably less.

Meanwhile, where are the large semiconductor makers as far as molecular devices are concerned? Pretty much fast asleep. The only big company with a significant initiative in molecular electronics is Hewlett-Packard. Not only does HP's participation in this field support the notion that imaginative research can make a difference to the survival prospects of a corporation, it also throws down the gauntlet to competitors that remain on the sidelines. HP's eloquent chief scientist, Joel Birnbaum, established the program in molecular electronics four years ago because, he says, "I wanted to show the other big companies that it's possible to continue to do basic research."

Birnbaum sees eerie parallels today with the 1950s, when CEOs of many of the large electronics companies that made vacuum tubes believed the tubes held an unassailable competitive position against the newly invented transistor. So far, Birnbaum has been shouting in a vacuum where other large companies are concerned. This year he and a colleague wrote in the science journal Physics Today that "...many researchers and corporate executives seem to have a blind optimism that somehow [the problems facing silicon will be resolved]. We think we cannot risk it."

Materials Innovation Inc. The Shrinking Metal-Powder Press

Amid the white pine, poplar, and birch forests of central New Hampshire, three unlikely entrepreneurs --a CPA, a sculptor, and a government scientist--are rewriting manufacturing's future. They have taken the enormous press that turns powdered metal into industrial parts and have shrunk it from the size of an up-ended tractor-trailer to that of a small refrigerator. They have also created particles of coated steel and other metals for the manufacture of new parts that are giving products like motors a heretofore unattainable compactness and efficiency.

The metal-powder industry is an important part of manufacturing. According to the Metal Powder Industries Federation, in 1999 the industry produced $5 billion of components that go into the making of products like valves, pumps, and drive gears. Pressed-powder parts are made by filling a die with powdered metal and then rapidly compressing the particles with a punch press that delivers a million pounds of pressure. This is a cheaper and faster way of making components than carving them from blocks of metal using grinding and milling machines. But the gigantic presses designed more than 50 years ago suffer from geriatric diseases. Sometimes the parts they make crack, often the components are not uniform enough in structure, and certain parts can't even be made on them.

Into the breach leapt Materials Innovation Inc. (MII) of West Lebanon, N.H. The brainchild of entrepreneur and former accountant Alan Beane, this little company had been working since its founding in 1985 to develop new powdered metals, coating the microscopic particles in novel ways to assure production of better parts. A few years ago Beane decided that big opportunities lay in making composite pressed parts for electronics and other uses. He brought in his brother Glenn, a Syracuse University-trained sculptor and an expert in metallurgy. The Beanes hired prominent materials scientist David Lashmore from the National Institute of Standards and Technology.

The trio started making what they knew were superior metal powders intended to produce so-called soft magnetic parts for electric motors and electronic devices. Unlike the permanently magnetized hard magnet you used in school to attract metal particles, soft magnetic materials become magnetic only after a current is passed through them.

MII acquired one of those gigantic conventional metal-powder presses. But when the three pioneers tried to make parts from their new materials on the big press, they found that it wasn't precise enough to give them the tolerances they needed. Says MII President Glenn Beane: "We were stopped dead in our tracks."

Two years ago the MII team put aside work on metal powders, deciding that they needed a new type of press. Software-savvy Glenn Beane led the design effort, which he finds as satisfying as creating sculptures. To build the new machine, Beane and the other MII wizards applied computer intelligence where brute force had reigned, embracing principles that were new to their industry.

Unlike the huge old presses, the small MII press is entirely computer controlled, so anyone who knows how to run a PC can operate the machine. This will let manufacturers tap a large pool of workers, not just highly specialized--and scarce--experienced press operators. By emphasizing software controls, MII has also introduced another novelty to the industry: Its presses can be diagnosed and run remotely around the clock, bringing the digital factory closer to reality. The new press is more precise too. It weighs the powder for each part, while the old presses measured powder by volume. The MII press distributes the metal particles more evenly inside the die and compresses the powder with twice the force of conventional presses, resulting in denser and more precisely engineered parts that require no additional machining.

MII's powder press lets manufacturers avoid costly downtime as well. Unlike existing presses, which tower 18 feet above the ground and extend six feet beneath the surface, the small MII press can be moved around on a forklift and plugged into production lines at special docking stations, which will house the hydraulic system that powers the press. When manufacturers finish a production run of a component, they can unplug the MII press and hook up another one programmed to make a different part. The process takes minutes. To switch production runs using a conventional press, the partmaker must take its multimillion-dollar machine out of action for days to retool it. MII will also get manufacturers up and running more quickly. It will deliver a press in weeks, vs. 12 to 18 months for a conventional press. And unlike the larger machines, MII's presses will be available to lease. By year's end, CEO Alan Beane expects ten of its presses to be in use at MII and customer sites.

MII promises to reshape the metal-powder industry not only by letting factories operate around the clock with a minimum of workers but also by making the plants smaller and cleaner. Old-economy metal-powder plants are usually full of dust and noise. MII's first machine operates in a spotless, high-ceilinged, airy facility at company headquarters.

On this pioneering press, MII is making parts for a novel electric motor used to drive water-purification and other industrial pumps, produced by SHURflo Manufacturing Co. of Santa Ana, Calif., a unit of Wisconsin Energy Corp. The motor part, called a stator, could not have been made to the user's specifications by any other manufacturing technique. A stator serves as the supporting structure for a coil of wire that is connected to an electrical source. When current flows, the coil creates a magnetic field in the stator, which transfers the energy to a part called a rotor that does the desired work, such as turning a shaft.

MII's stator is made from a specially coated iron powder with a uniformly dense structure, which makes the stator more productive. Motors using MII's stator turn more electricity into work instead of dissipating it as heat, and can pack more horsepower into a smaller volume. SHURflo says its new motor is 30% more efficient than competitors' older motors. MII's soft magnetic powder, which consists of ceramic-coated iron particles, can also be used to make high-performance electromechanical valves, parts of automotive fuel-injection systems, and many other components.

For another customer, Dudley Lock, a major lock manufacturer in Hemmingford, Quebec, MII makes the plunger, a critical part of a padlock. A weak plunger compromises the security of a lock. Dudley's plungers are made on the new press from MII's sinter-hardened steel alloy powder. The MII part can withstand more than twice the force that plungers made of traditional powdered steel can.

Larger versions of the MII press will be able to make much bigger parts, including car doors. Lanny Pease, who runs Powder Tech Associates, a consulting firm in North Andover, Mass., calls MII's new press "without a doubt a major advance of the past 50 years." Adds Woody Haddix, president of Helsel, a division of Hawk Powder Metal Group in Campbellsburg, Ind., which makes parts for the hydraulics industry: "[MII has] revolutionized the design and application of powdered-metal compacting technology." When Alan Beane calls his press a new-economy manufacturing system, he has supporters.

Bios Group Simplifying Complexity

What do the apparently chaotic pathways of foraging ants have to do with improving the allocation of resources in a manufacturing plant? Or the application of genetic algorithms that follow the laws of natural selection--the survival of the fittest--with boosting the efficiency of a company's supply chain? Or the use of specialized bits of software code called agents with determining the number of models an automaker should sell next year?

Plenty, it turns out. In the shadow of the spectacular Sangre de Cristo Mountains of New Mexico, Stuart Kauffman and his colleagues at Bios Group delve into the fascinating subject of complexity theory. Bios Group is a pioneering Santa Fe company that applies biological solutions to business problems. Most notably its software has successfully used ant algorithms--rules that imitate the movement of ants--to improve supply chains at a host of prominent companies. Among its clients: Procter & Gamble, Ford, Unilever, Boeing, and Texas Instruments.

Ant algorithms are math formulas that are incorporated into software. They stem from observations by naturalists that when ants start looking for food, they send out a number of scouts. As a scout treads along, it lays a trail of chemicals called pheromones that its fellow ants can recognize. The first ant that returns after a successful hunt is presumed to have found the food source closest to the ant heap. It has marked this route with pheromones twice--coming and going. Other ants will now follow this preferred pathway. If something should block the route or if the food source is exhausted, the ants will follow alternate routes to more distant food sources, marked by other scouts that took longer to return to the nest.

Despite the apparent chaos of life in the ant heap, individual ants demonstrate flexibility in response to changing events. Older, more experienced ants are usually the foragers. But if food sources dry up, the experienced scouts are joined by younger nursemaid ants that can adapt to a new role when the colony needs to bulk up the numbers of the search party. In an ant algorithm, software agents (bits of software representing individual ants) perform rerouting automatically, similar to the way ants readjust their paths to food sources. The algorithms mirror the fact that cooperation in an ant colony is largely self-organized. The formulas help identify good solutions to static problems and can quickly adapt when circumstances change.

Ant algorithms were introduced to Bios in 1998 by French scientist Eric Bonabeau, now president of EuroBios. Bonabeau had earlier devised the algorithms to help optimize the routing of British Telecom's network. Bios has several patents pending on its algorithms. The company has also applied genetic algorithms, agent-based modeling, data mining (detection of patterns in large databases), optimization (seeking the best possible solution), and other components of complexity science to projects at more than 30 companies.

At Unilever, for instance, Bios used its algorithms to identify the most efficient allocation of manufacturing resources. Under Bios' direction, the company changed the location of mixing and storage tanks and packing machines to best cope with breakdowns and to meet fluctuating customer demands in the shortest amount of time. Kauffman, chief scientist at Bios, says, "We were able to tell Unilever to put more redundancy into its manufacturing system--to install machines that are multifunctional, machines that can do three different jobs." Instead of making products sequentially in a fixed order, with one machine passing the products to another machine, Bios suggested intermingling production schedules. Unilever is now installing Bios scheduling software on top of its new manufacturing software; the company expects a big productivity payoff next year.

Ant algorithms are especially useful when corporations are having trouble identifying the most efficient sequence for a series of actions. To help Southwest Airlines reduce cargo handling bottlenecks, for example, Bios simulated the destinations, shipments, freight-house operations, and ramp personnel of the airline's cargo operations. The software agents used in the simulation first followed the established rules for controlling the flow of freight throughout the network. One of those rules called for loading the freight on the first plane going in the right direction.

The Bios software showed that following this rule slavishly leads to unnecessary handling and bottlenecks at certain locations--the precise cause of most delays. Bios came up with a different rule: "Try to put the freight on an airplane that is landing in the destination city within the delivery deadline." This rule could result in more circuitous routing, but it sharply reduced freight handling--and that was a greater expense than air miles. With its new software model, Bios predicted that Southwest could cut overnight transfer weight by 71% and overall freight-handling labor by 20%. Field trials bore out the forecast. Southwest implemented the new rules across its system last year for an annual saving in labor costs of $10 million.

Procter & Gamble came to Bios in 1998 for help in cutting its inventory. The $38-billion-a-year giant had already achieved a 50% reduction and was looking for another 25 percentage points. Bios found that the key to further inventory reduction was "constraint relaxation"--that is, the removal of restrictions such as never sending out trucks that aren't fully loaded. Introducing flexibility into both delivery schedules and the size of truckloads worked wonders.

Bios scientists found that P&G's seemingly logical policy of sending out only full trucks actually created disruptions along the supply chain as trucks waited until they were filled. With such delays, P&G's greatest fear came true: supermarket shelves that were empty of its key products, like Tide and Comet. Likening the flow of a supply chain to that of a river, Kauffman says the "lumps" introduced by the full trucks were like rocks that disturb water flow. Letting some trucks travel with partial loads and making delivery times more flexible removed the lumps.

P&G completed the first field trials of the Bios system in September and found that results exceeded expectations. So impressed was P&G that it recently became a part owner of Bios by investing $5 million. More trials are scheduled for December.

In an ongoing project for Ford, Bios is trying a novel approach to help the automaker decide how many models it should be selling and what their attributes should be. Today Ford uses many sources of information to gain insight into customer preferences, including focus groups, sales figures, J.D. Power surveys, and more recently click-stream data. The problem is that data from different sources often conflict, and no one source is inherently superior. A further problem with focus groups is that consumers' answers about their preferences don't always reflect what they would actually buy. "If I'm in a focus group and I'm asked what I would purchase," says Bios marketing director Christine McLorrain, "I'm not constrained by my pocketbook, my husband, my need for a place to put the dog. I'm more likely to answer with my fantasy car. Since dealers are interested in actual purchases to keep inventory low, the information gleaned from focus groups is not very helpful."

Taking such shortcomings into consideration, Bios scientists are building software agents programmed with buying characteristics--that is, values and preferences for particular vehicle features. If you could talk to one of the software agents, it could tell you exactly how much it would be willing to pay for air conditioning or four-wheel drive. The software agents will give Ford insights that it won't have to psychoanalyze. The automaker will be able to focus on building the cars that are most likely to attract buyers, rather than trying to appeal to everybody with a multiplicity of models.

In other projects, Bios Group has shown Disneyland how to get people to take more rides in less time. It helped Temple-Inland figure out what timber it should bring into its mills based on lumber prices. And it has improved maintenance scheduling for United Airlines.

These diverse Bios successes show how a small firm with a clever idea can point to a new way of thinking about corporations and other complex systems--and how insights into the ways that simple events influence one another can help us understand a complicated whole. Its thesis is that in an increasingly connected economy, businesses should not follow the rules of the Industrial Age, which depended on control and predictability. Complexity theory instead calls for self-organization, distributed control, and the study of so-called emergent behavior--actions that emerge, or become predictable, only when looking at an entire system, not at its parts, like the flocking of birds or the foraging of ants or the appearance of cargo bottlenecks at a particular airport. The new science seeks to give companies robust new approaches for managing rapid growth and change.

Bios' thesis, furthermore, holds that as connections proliferate in the new economy, it begins to resemble a living ecology. Where guesswork has traditionally prevailed in scheduling production or deliveries, science-based adaptive scheduling, borrowed from the supposedly lowly ants, will help companies become efficient as they make products in smaller lots to respond better to consumer demands.

Applications of complexity science will eventually lead to what Chris Meyer, a co-founder of Bios, calls inanimate evolution. For example, when a motorist spins his wheels in the snow, the car automatically signals a customer-service representative for help by cell phone. The rep responds with a software download that upgrades the traction control system, and the car drives away. In the complexity world, predicts Meyer, software agents will deal with other software agents, arranging such things as complex B2B transactions faster and maybe better than people can.

Catalytica Combustion Systems Inc. Clean Turbines To Make Electricity

Most big companies that make gas turbines for the production of electricity go about the cleanup of noxious exhaust gases in a backward way. They pass gaseous emissions through a shedlike "smokehouse" where other poisonous chemicals scrub out the pollutants. It took a small Silicon Valley company to figure out how to clean up the gases at the logical spot--where combustion starts. The elegant technique not only eliminates those after-the-fact smokehouses but also produces power in a manner that is almost pollution-free, and more efficient than the conventional technology.

For more than a year now, a turbine equipped with a small catalytic converter from Catalytica Combustion Systems of Mountain View, Calif., a subsidiary of Catalytica Inc., has successfully operated in test mode at Silicon Valley Power, a municipally owned utility in nearby Santa Clara. The first commercial use of Catalytica's system will be at Enron's Pastoria energy plant near Bakersfield, Calif., where it will be fitted on six GE gas turbines. Three other turbines, made by Kawasaki, will soon be installed with Catalytica's technology at an electric plant in Massachusetts. Says Cliff Baxter, CEO of Enron North America: "Catalytica's technology offers an attractive alternative to traditional emission-control systems."

The inventor of the clever new technique, which Catalytica calls Xonon Cool Combustion, is vice president and chief scientist Ralph A. Dalla Betta, 54. The centerpiece is a palladium-based catalyst that is incorporated into a package the size of an artillery shell and fitted into the side of a turbine. The catalyst works by reducing combustion temperature, thus almost totally eliminating the formation of nitrous oxide (NOx), the major ingredient of smog and other pollutants--without compromising the performance of the turbine.

It hasn't been easy for Catalytica to get where it is. Dalla Betta first demonstrated the Xonon principle ten years ago. But a small company, notes Catalytica CEO Ricardo Levy, "often has to bridge the perception gap. It certainly takes longer than one expects."

Now that large turbine makers such as GE, Kawasaki, Solar Turbines, and Rolls-Royce are beginning to use the Xonon Cool Combustion system in their products, Catalytica is looking to capture a big share of important new markets. One major opportunity is the distributed electrical power market, created as high-tech companies and municipalities set up their own generating facilities. They want to be assured of a reliable supply of electricity, and they don't want to worry about pollution.

FEEDBACK: gbylinsky@fortunemail.com