F E A T U R E S    Issue 1.07 - November 1995

The Media Lab at 10

By Fred Hapgood

Beyond the technology it has imagined, the thinkers it has encouraged, the students it has attracted and seeded, the greatest achievement of the MIT Media Laboratory may be the invention of - itself.

For a few years after it officially opened in 1985, the MIT Media Laboratory may have been the most celebrated research institute in the world, at least as measured by inches of newsprint or minutes of television time. Perhaps it still is. Major periodicals and TV science series have covered the organisation - taking the reader or viewer through a carnival ride of speech-driven task managers, holograms, lip readers, eye trackers, gesture recognisers, virtual reality modellers, and more. A book called The Media Lab, in which writer Stewart Brand did the same (though at a slower pace), became one of the major pop-tech sellers of recent times. The argument used to rationalise all this attention was that the Media Lab was "inventing the future," a phrase used by Brand in the subtitle of his book and by The New York Times in its August 23, 1987, profile of the facility. Inventing the future is by definition the agenda of any technical research institute, but in this case the expression referred to a more ambitious claim: that the Media Lab was defining the next keystone technology, the theme around which the other technologies of the time would orbit, like steel production in the 1800s or electric-power generation in the first half of the 1900s. The candidate being primed for this role for the late 20th century was human-machine interactivity.

"Imagine," wrote MIT professor Nicholas Negroponte, the lab's founder and director, in his 1968 book The Architecture Machine (The MIT Press), "a machine that can ... discern and assimilate your conversational idiosyncrasies ... (that can) build a predictive model of your conversational performance." OK. And what would be the point of that? The point, Negroponte said, is that dialogue with such a machine "would be so intimate - even exclusive - that ... (it would) bring about ideas ... unrealisable by either conversant alone."

When Negroponte wrote those words, he had in mind a house - he began his career as an architect - a building that would act as a consultant on the question of its own redesign, engaging its owners in an ongoing dialogue on such issues as ventilation, illumination, and drainage. When both parties pronounced themselves content with said redesign, the owners would go off for a few days, perhaps on vacation, and the house would reconstruct itself autonomously, metamorphosing into the desired form.

As a concept, this "Architecture Machine" may have been a great example of interactivity, but it obviously faced a few short-term fabrication constraints. So, through the 1970s, Negroponte pushed his ideas into more flexible and plastic contexts, specifically the media: broadcast, publishing, movies, and telecommunications. He liked to talk about asking the TV to summarise or expand on a point, or have the set watch a show and then tell you if you'd like it and why. Your newspaper could track what you skipped and reread and where you paused, then use those cues to evolve into a composite "Daily Me" that would carry only the news you cared about most. Advertisements would watch people watching them and therefore continuously adapt to their responses.

At another time, these ideas might have made no more than an interesting essay, a counterpoint to the conventional opinion - as developed by psychologist Bruno Bettelheim - that normal people did not have "intimate and exclusive" relationships with machines. However, Negroponte wrote against the background of the extraordinary success of the personal computer. The intimations of change that spread in the wake of that success sent thousands of corporate executives streaming into the vision marketplace, where they went milling about, crying out for guidance.

Given occasion, Negroponte could soar as high as any futurist ("Monologues will become conversations; the impersonal will become personal; the tradi-tional 'mass media' will essentially disappear," he proclaimed in The Media Lab), but his argument about autopersonalising machinery was more focused, tighter, easier to grasp than most of the nostrums being peddled in this sector. Business leaders flocked to hear his analyses. When he proposed organising a facility that would pursue this vision firsthand, supporting itself by selling ringside seats to the adventure, corporations responded enthusiastically.

In 1985, after being housed for two years in the MIT computer science department, the Media Lab opened in a cool, closed-in block of a building known locally - after its architect, I. M. Pei, and its distinctive exterior white tilings - as "the Pei toilet." On the most gen-eral level, the lab's agenda was the man-machine interface, a phrase that encompassed ways to talk to machines (such as speech recognition), ways for these machines to talk back (such as holography), and processes which supported this interaction (such as video compression).

Many of these technologies, let alone the general theme of machine interactivity, were being researched elsewhere, but the Media Lab came at questions a little differently. Unlike in-house researchers or contract labs, it was looking not for new applications but for new application domains. The idea was not to cut a specific path but to illuminate the landscape, usually by simultaneously prototyping several different but related applications in a domain.

This was a new niche in technical research, somewhere between industrial R&D, with its near-term orientation, and the academic engineering sciences, in which real-world constraints are assumed away in order to define problems of "professional interest." These issues are old enough - and popular enough - to have attracted a pool of expertise sufficient to staff the committees that control funding and editorial decisions within the professions.

As is often the case in nature, occupying a new niche requires radical mutations of the underlying phenotype. In this instance, what mutated was the look, feel, and structure of the institutions that had been responsible for the technical research of the past 50 years. For example, members of these facilities have traditionally assumed that visitors understand the core concepts of the technology. But the Media Lab had a second constituency: executives in marketing, strategic planning, and corporate development. These were big-picture/high-concept types on tight schedules, corporate movers who needed to be able to glance at something, get it, and be off. And so the Media Lab became famous for its "demos," colourful, witty, and stylish presentations - usually in code or on tape (or videodisc) - that demonstrated how a particular project would work in a real-world context.

A second difference was that the breadth of the lab's mission - to illuminate new technologies by prototyping them in several contexts - required it to draw resources and expertise from a wide range of cultural sectors. The small faculty included an opera composer and a filmmaker, an AI scientist, graphics specialists, and a team of educators. A person walking the halls might see demos of new musical instruments in a concert hall, holographic modelling in a manufacturing company, and speech-driven task schedulers in an office, all side by side. There was not a hint of the professional partitions that divide - some prefer "focus" - the attention of traditional research institutes.

Most facilities have a small number of sponsors, often one; research is naturally dominated and defined by the interests of that sponsor. Work in industrial labs is controlled by corporate manage-ment. Research in academia is controlled by professional panels that make sure that every project funded falls somewhere on the short list of venerable theoretical issues deemed to be "of professional interest." (Not every interesting question qualifies. The Rockefeller University professor emeritus Donald Griffin, who discovered echolocation in bats in the late 1930s while still a student at Harvard, recalls that his advisor suggested he drop the topic. Who would peer review his work?) In either case, a researcher who wants to pursue an interest not shared by his sponsor is out of luck.

By contrast, the Media Lab's product is not a "product" but a seat on an expedition across the technological frontiers. This positioning provides the lab's work many hundreds of potential buyers, each with their particular perspectives and needs.

Unlike institutes that live off government money, the Media Lab's financing opportunities are not restricted by national borders - half its funding comes from overseas.

Consequently, the place has unusual freedom of action. "If one sponsor isn't interested in a project, another one will be," says the lab's associate director Walter Bender.

"I remember getting one of those do-this-do-that letters from a sponsor," recalls a Media Lab student. So what did he do? "I sent him this vicious flame," the student says. "And I copied it to everyone. They know they're not supposed to do that." This spirit of independence has helped the lab to form and follow its own identity. Four years ago, for example, the Media Lab came out swinging for digital high-definition television, despite the fact that several of its sponsors at the time had made substantial investments in the rival analogue technology.

Finally, the size of its market allows the Media Lab to choose to go down branches of research that might lead nowhere. "Coming up empty" is the ultimate nightmare in other funding models, but for the Media Lab one annoyed sponsor is not the end of the world.

From the point of view of the sponsors, what the Media Lab sells is the opportunity to follow and examine in detail aggressive investigations of new technologies. "It's an idea factory," says Gary Bottger, manager of external technology at longtime sponsor Eastman Kodak Company. And these are the terms on which the Media Lab has to produce.

Sponsors are naturally feeling happiest if they can see a steady flow of ideas, perhaps not on each visit but regularly. This imposes intense pressure on the students to keep cranking.

"To have no demo is not a good state to be in here," a graduate student says dryly. On the other hand, the lab tries to pay 100 per cent of the tuition for its students and frees nontenured faculty from fund-raising responsibilities - nobody has to write any grant proposals - so both can concentrate on keeping the river flowing.

And so the river flows, at the Media Lab's distinctive pace. Touring the lab once, I passed a game demo running on a wall-size display. The game could read a player's silhouette through a mounted camera and react accordingly. Turn one way, and the view on the display would slide in the other direction; if you fired a gun, the display would figure out what you were doing and where the gun must have been pointed, then detonate the proper portion of the landscape, synthesising the appropriate sound effects. "He wrote that in a couple days," my guide said, pointing to a student. The student heard him. "Actually it took me more like three weeks," he said modestly.

Perhaps what made the Media Lab so celebrated when it opened a decade ago was less the romance of human-machine interactivity than the spirit of the place itself. The cultural variety gave it the outline of a complete society; the relentless flood of projects made the place feel as if, wherever it was going, it was getting there without losing a moment; and its collective self-possession gave it authority and stature. One felt the different qualities and characters of the culture mixing, reacting with new energies, bursting out of the dead hands that shape the atmosphere of other institutes. It may have been that the technology was the least part of the future being invented.

Ten years later, it seems fair to say that the age of intimate and exclusive relations with machines is still a bit over the horizon. The impersonal has not become personal. (Whatever that might mean, it hasn't happened.)

The mass media have not "essentially disappeared," though they might be turning a touch brown around the edges. Both interactive TV and interactive movies are moribund; the interactivity that has developed most rapidly over the last decade has not been human-machine but human-human - e-mail, newsgroups, talk radio, home shopping. And one of the biggest pieces of news in machine interfaces has been Microsoft's struggle to make its operating system look more like that of the 1985 Macintosh, an example of inventing the past, if anything.

Perhaps the spread of CD-ROMs and the popularity of the Web might allow machines to claim a gain of a couple of IQ points, though even at this stage, not many applications autocustomise in a very interesting way.

Yet today's visitor wandering the passages of the Pei toilet will find projects extending far beyond the original media horizons: narrative recognition, the gender of machines, directable cameras, neighbourhood nets, remembrance agents, augmented realities, automated performers, steganography (hiding one signal in another), haptic holography, storyteller systems, fast physics simulations, wearable networks, embroidered circuitry, evolution-driven animation, and on and on - about 100 projects in all. Perhaps paradoxically, the Media Lab has prospered even more than its vision.

Over the last decade, the lab's budget has grown at an average annual rate of 30 per cent a year to a present level of about $25 million.

By contrast, in the last five years MIT, which gets much of its money from the government, has lost approximately 30 per cent of its gifts, grants, and bequests. The Media Lab now has more than 100 students, up from 58 in 1986, and like its neighbour MIT, the lab relies heavily on its students' work. Applicants to the Media Lab don't apply through MIT; the lab has its own procedure: students are admitted by specific faculty members with whom they then work very closely - faculty members sometimes refer to their having "hired" a given student. Students work cheap, and low costs are probably necessary to console executives with giving up control over research. (The average amount paid by a sponsor is about $200,000 a year, but entry-level memberships go down to $75,000 a year for a three-year commitment, about the price of four quarter-page ads in The New York Times.) The regular and rapid turnover of students maintains the flow of new ideas that attracts sponsors, and, according to Kodak's Gary Bottger, Media Lab graduates are in intense demand.

Employability, of course, is not the only test of a student's success. When Joshua Smith, now a masters student at the Media Lab, started thinking about graduate school, he faced a problem common to intellectually active students, almost symptomatic of the breed: his interests did not respect professional boundaries. He had one BA in computer science and philosophy (a double major), and a sec-ond BA in physics.

"I wanted to combine the mode of inquiry of physics with the domain of computer science," Smith says. "I was interested in whether we could make models of computation that worked as well as our models of the physical world. Questions like that don't fit in at many graduate schools."

Smith had heard about Neil Gershenfeld, a physics professor at the Media Lab who shared similar interests, and came by to talk. Gershenfeld hired Smith, who joined his project on interface transducers. The idea was to fill a volume of air with electric fields such that when a user waved his hands through the air, like a symphony conductor, the device could infer changes in the position of the hands from changes in the fields. The tech-nology has been prototyped in gaming and instructional contexts, as musical instruments, and as a pointing device, like a 3-D mouse.

From Smith's perspective, electric-field sensing combined all his interests - including philosophy, given the issues of logical inference. Yet Smith knows he is doing an end-run around peer review: "As a potential academic," he says, "that makes me feel a little funny."

But there is no professional academic community of electric-field sensing - the whole field is too new. His "peer" group is the sponsors. Sponsor engineers critique his designs, but the opportunity for technical review seems to matter less than the support and enthusiasm of this small community that has sprung out of the sponsor mix. Smith says that overall the lab shaped his interests in physics and information into a commitment to the problems of sensing, which he says he expects to be thinking about for some years yet.

When Nicholas Negroponte started the Media Lab, his aim was to build an institution that reached the very highest levels of influence and innovation, such as MIT's Research Laboratory of Electronics in the '40s, Bell Labs in the '60s, and Xerox PARC in the '70s.

Probably - history retains its freedom of judgment - the Media Lab flunked this standard in the '80s. But the standard itself is absurdly high: If the Media Lab was not the Xerox PARC of the '80s, neither was anyplace else. Over the last 20 years, MIT as a whole has received billions of dollars in research funding. How many people can name one big idea or significant technical innovation - besides X Windows - brought about by the expenditure of all that money?

When the sponsors are asked about the Media Lab's contributions, they talk less about being inspired to beam themselves into a new commercial universe and more about getting help with the hard technical questions that come with adapting a given industrial mission to a changing world.

Ed Horowitz, CEO of Viacom Interactive Media, says the lab has played a big role in his company's work on image archiving. Gary Bottger credits the lab with helping to expand Kodak's ideas about digital camera applications and develop desktop publishing standards. Sam Fuller, vice president of corporate research at Digital Equipment Corporation, says his company was strongly influenced by the lab's thinking on digital HDTV. Bill Molteni, senior scientist at Polaroid Corporation, sees the lab as a high-end user's group, a place that can be depended on to know more about the best new technology than its own manufacturers.

These are all exactly the kinds of services that industry has looked to MIT to provide for most of its history - the services that were a major source of support for MIT before the government started funding academic engineering research at such a high level.

The lab, however, missed what turned out to be the decade's best ideas in machine interactivity: hypertext markup language and browsers. And the Media Lab might be constitutionally incapable of generating these types of ideas. Both programs were thrown together by people who wanted something cheap and crude that they could hand out to their friends for free. Because neither program pushed out the wall of software development, few working in the Media Lab might have thought such programs interesting enough to develop.

One has to wonder at this point how far the Media Lab model - private funding, lots of sponsors, a sweeping mission, institutional autonomy, the deep mixing of a wide range of cultural modalities - can be pushed.

Can there be a Media Lab for biotech? For materials research? Can we stop subsidising the research professions and be the better for it? A decade after its inception, the lab serves as a model for the organisation of technical research and the relation of research to industry. It is also a model for education, in which art and engineering are combined in a project-centred curriculum where each student's work is reviewed by real-world experts instead of academic professionals. No doubt this model has its limitations, but we live at a time when the old ideas on these issues are simultaneously rusting out. Perhaps the single biggest achievement of the last 10 years will turn out to be the Media Lab itself.

Fred Hapgood is a freelance writer specialising in science and technology. His article on Internet telephony appeared in Wired US 3.10. He can be reached at hapgood@pobox.com.