Xerox Won't Duplicate Past Errors

Now, its innovations will get used in-house--or spun off

It was one of the great fumbles of all time. In the 1970s, Xerox Corp.'s Palo Alto Research Center (PARC) developed the technologies that would drive the personal computer revolution. "By 1979, we had it all--graphical user interfaces, mice, windows and pull-down menus, laser printing, distributed computing, and Ethernet," recalls M. Frank Squires, a PARC founder in 1970 and now chief administrative officer of Sematech Inc., the chip-industry consortium in Austin, Tex.

Xerox had the PC and networking businesses firmly hooked--but didn't try to reel them in. It didn't even patent PARC's innovations. Management was too preoccupied with aggressive competition from Japan in its core copier business, says CEO Paul A. Allaire. "If we'd been really good, we could have done both. We probably should have," he admits.

Instead, PARC's technologies became the foundations for such icons as Apple Computer Inc. and 3Com Corp. Apple co-founder Steven P. Jobs visited PARC in 1979 and was astonished at what he saw. His PARC tour inspired the Macintosh. That same year, Ethernet inventor Robert M. Metcalfe left PARC to start 3Com. These are just two famous examples of the great Xerox giveaway.

It won't happen again, asserts Allaire. The butterfingers days are over. PARC's mission has always been to blaze the way for tomorrow's Xerox--but PARC's research is more critical than ever, because Xerox' future will be increasingly digital. So Allaire visits regularly to get his arms around PARC's leading-edge technology. "It also gives [the researchers] encouragement," he adds. "Besides, it's a lot of fun."

That's because Xerox PARC supports a stimulating melange of research. PARC is home not just to the usual complement of computer wizards, programmers, and engineers. There's also an assortment of anthropologists, psychologists, linguists, and cognitive scientists. They explore the social contexts that surround technology, looking for ways to make digital systems conform to human expectations.

This eclectic crowd often hatches potent ideas for which Xerox has no ready use. But when something looks like a winner, Xerox doesn't simply let it slip away as it did before. Now the company bankrolls an entrepreneurial spin-off. The parent company figures to profit from these ventures, even if the start-

ups remain a small part of the $17.4 billion giant.

The entrepreneurial move kicked off in 1989 with Xerox Technology Ventures, which provides seed capital. Early last year, Xerox New Enterprises (XNE) was formed to oversee PARC spin-offs and help with logistics and finding outside funds for growth. XNE now has 10 companies under its umbrella. Already, they're generating $700 million in annual revenues (although half comes from Xerox Engineering Systems, an existing unit that was folded into XNE this year).

XNE is modeled on Thermo Electron Corp., a Waltham (Mass.) company that for 14 years has been making a business of creating businesses--21 public companies so far. "A lot of people thought we were crazy" when Thermo Electron started bankrolling spin-offs, says President John N. Hatsopoulos, but the entrepreneurial culture it engenders has paid off big. "The company is twice the size it would have been without the spinout strategy," he boasts.

Allaire says some Xerox executives also had initial misgivings about the spin-off strategy. Several startups funded by Technology Ventures went bust, "and for a while, that looked like it might be just wasting money," he recalls. But Documentum Inc. more than compensated. Founded in 1990, it went public last year "and paid a really good return on investment" (table).

One of last year's spin-offs, InXight Software Inc., shows why PARC Director John Seely Brown believes social scientists are vital to developing new technology. Windows and pull-down menus did a laudable job with simple programs. But with today's complex programs, the screen is often so crowded people grow frustrated trying to figure out what icon to click, he notes. Worse, the conventional windows interface lacks perspective. "It's like walking around with two toilet-paper tubes on your eyes," says Brown. "There's no sense of things moving smoothly from the periphery into your center of vision." As a result, everything is a surprise--and people don't like a constant diet of surprise.

InXight's Hyperbolic Tree interface is a first step toward a more fluid interaction. Instead of jerking the user from spot to spot, it flows as a mouse pointer moves, magnifying a particular region in the center of the screen while shrinking the bordering areas--but keeping them in sight through a sort of fish-eye lens. "It was designed with a deep understanding of human perception and cognition," says Brown.

CALMPUTING. Comshare Inc. in Ann Arbor, Mich., will shortly release a new version of its budgeting program that features InXight's Hyperbolic Tree. "This new stuff allows you to see quantitative relationships in ways never before possible," says Comshare Senior Vice-President Steven J. Tonissen.

Clever screens are just the openers of "calm computing." There's also dpiX Inc., which was founded last year to build high-resolution displays that soothe the eye with paperlike, no-flicker images. Back in the lab, PARC researchers are working to embed artificial intelligence in documents to fashion a sort of digital society, with software agents roaming among the documents and collecting information. "We want to do knowledge mining like we do data mining now--to find the knowledge that lies between documents," explains Brown. When it all comes together, he predicts, "knowbots" and intelligent documents "could unleash all the latent knowledge of the entire enterprise."

This sounds like it would be part of the Xerox mainstream. But for now, InXight is trotting out the initial offering. It's called LinguistX, and it lets computers understand human language. PARC has been hammering at so-called computation linguistics for 25 years, says Romano Rao, chief technologist at InXight, "and we've now gotten it to the point where it actually works." People will no longer have to memorize abstruse commands to search a database.

Computational linguistics goes beyond just translating words into computer code. It also involves mapping the brain's folds and convolutions. "Something in that shape makes humans genetically capable of learning language," says Rao. Using these maps, researchers have created computer models that represent the spatial relationships among language elements. With LinguistX, computers can automatically analyze digital documents to extract specific information, generate summaries, and alert a person to documents that require action.

Still more sophisticated tools are coming, says InXight President Mohan Trikha. Soon, he says, "you'll send an E-mail inquiry, and the computer will read it, interpret what's wanted, and automatically initiate the processes needed to fulfill the request." That would lift electronic commerce to unprecedented levels of productivity, he notes. Computers that smart will be so adroit at dealing with routine matters, Trikha predicts, that people will be able to pursue four or five careers in parallel.

Xerox may have missed opportunities in the PC age, but it hopes to pioneer an age of societal computers. And if PARC's grand vision doesn't quite happen, Allaire is confident his Palo Alto crew will continue to earn their keep. PARC is a bargain, he points out--its budget is less than 1% of the company's $1.6 billion R&D spending. Just one PARC creation, the high-end Docutech printer, raked in $1.8 billion last year "and probably justifies the entire cumulative investment in PARC," says Allaire. "PARC is one of the best investments we've ever made." And it's fun to visit as well.

Before it's here, it's on the Bloomberg Terminal.