o unit of the Massachusetts Institute of Technology suffers a more dubious reputation within the scientific community than its all-too-famous Media Lab, the victim of its persistent tendency to overstate its claims. There was its birth in a burst of big-name corporate funding. A breathless book by Stewart Brand. Its association with the supernova that is Wired magazine. The style of its founder, the larger-than-life magazine columnist and fund-raiser Nicholas Negroponte. Its determination to refer to its remarkably ill-designed home as ''the Wiesner Building'' (after the sainted MIT president who created the lab) instead of calling the building by its designated number, E-15, like everyone else at Tech.
All the while, an embarrassingly small stream of products has been issued from an organization that had promised to ''invent the future.'' The firesale last year of Patty Maes' Firefly Network Inc. to Microsoft was only the latest in a string of disappointments. Media Lab's most familiar product has been a blue streak of high-tech blarney.
For the past five years, however, things have begun to change, according to people outside the lab. Negroponte, though still very involved, has ceded on-site leadership to others. The young people who have been hired, promoted, and tenured within the lab have been selected along different lines: Stronger backgrounds in traditional disciplines are the rule. Instead of being governed exclusively by the principle ''Demo[onstrate] or Die,'' the lab has found a new role for an older imperative designed to weed out those who simply talk a good game: ''Publish or Perish.'' Student quality has improved.
The single most persuasive piece of evidence of this subtle transformation is the appearance of a new book by one of its faculty, Neil Gershenfeld. Not his trade book, ''When Things Start To Think,'' which is slowly climbing onto the best-seller lists. That in itself will make him heir-apparent to Negroponte as celebrity-in-chief. Instead, I mean a 344-page tract on the nature of mathematical modeling.
Gershenfeld, 39, is Exhibit A among the lab's next generation. A Cornell physicist, veteran of Bell Labs, a former junior fellow of the Society of Fellows at Harvard, Gershenfeld moved to MIT five years ago and loved it. He now rhapsodizes about the lab's extra-disciplinary approach as happily as Negroponte ever has. Indeed, the back of the jacket of ''When Things Start to Think'' illustrates to near perfection what it is about Media Lab-speak that chafes.
Negroponte writes, ''Gershenfeld gets it. He was the first person to introduce me to the relative merits of bits and atoms. This book elaborates his vision in the narrative tone of a native of, not a visitor to, the future. It is an important story about why and how computers will disappear, when and where your things will think.''
Penn Jillette, half the high-brow magic act of Penn&Teller, writes, ''Hey, I know Neil personally. I've gotten to hang at the Media Lab. I was part of a stupid showbiz gag that turned into a way to make air bags safer for kids (don't blame me, I was just going for laughs). He's a smart guy who thinks, and that's so much rarer than I used to imagine.
''If you can find the time to put a silly little show together, have it play in Boston, meet some of the MIT cats, invite them to your show, get in good with them, meet Neil, do a project with him, bring him to Vegas, and sit up all night at a coffee shop talking about the future - do it! If you have better things to do, read this book.''
Now the funny thing is, if you do read ''When Things Start To Think,'' you'll be startled by how much deeper in spirit it is than its blurbs. The story of the magic trick into car seats turns out to be a lucid meditation on the nonlinear process by which basic science is applied to practical problems.
''I never would have taken funding for automobile seat safety development - it's too remote from what I do, and too narrow,'' writes Gershenfeld. ''NEC Corp. would never have supported magic tricks internally - it's too remote from what they do, and too crazy.'' Yet the result of the collaboration in the free-play zone of the lab will arrive on the market this spring: a sensor that automatically switches off the air bag for rear-facing infant seats. In the future: air bags automatically adjusted for passengers' weight and positions in all seats.
Likewise, a chapter called ''The Nature of Computation'' provides an introduction to the mechanics of quantum computing - an enormous increase in power that apparently is suddenly a realistic possibility. It is much easier to read than Gershenfeld's paper on the topic in Science magazine, or even the Scientific American popularization of it, for that matter. The chapter on ''The Personal Fabricator'' describes an act of the imagination by which the tool currently at the center of Gershenfeld's course called ''How To Make Any Thing'' - a three-dimensional printer consisting mainly of a computer-controlled glue-gun - eventually might be turned into a machine that would make all others.
''Information and Education'' explains that the Media Lab is really a front for a more interesting project: Jerry Wiesner's long-cherished dream of an institution designed to diminish the power of university departments to dictate the shape of education. The pressure for specialization within disciplines often led to a lack of awareness of common problems and alternative approaches to them, Wiesner believed.
Hence, for example, the two undergraduate courses that Gershenfeld devised to make certain that students in the otherwise unstructured Media Lab program had the tools to understand both the physical world outside of computers and the logical world within. ''The Nature of Mathematical Modeling'' is one result; a second soon-to-be-published book, ''The Physics of Information Technology,'' is another.
''My students loved them,'' he writes of the lectures. ''Some of my peers hated them.'' The reason: each week he would cover in relatively few strokes material that a specialist ordinarily might take a semester to teach. Reducing whole disciplines to 10 pages or so of essential ideas makes for a remarkable guidebook. Virtually every present-day technique for modeling systems is displayed, like so many tools hung on a pegboard.
In ''The Nature of Mathematical Modeling,'' Gershenfeld writes, ''How would you describe the flickering of a flame? The texture of an oil painting? Highway traffic during rush hour? Twinkling stars? Breaking glass? A bowling ball hitting pins? Melting ice? The flight of a paper airplane? The sound of a violin?'' Different processes require different tools, he says, and there are many possible levels of description.
So old college warhorses such as differential equations and the calculus of variations are presented side-by-side with more recent techniques of describing dynamic systems such as neural networks and simulated annealing. Cellular automata and genetic algorithms are explained, along with time series and state estimation - all pitched to the level of an as-yet-inexperienced practitioner. There are appendices devoted to math packages and graphics tools, to network programming and benchmarking techniques.
None of these tools is required for newspaper discourse - nor do I possess any skill in their use whatsoever. But anyone who wants a sense of how the language of mathematics has changed in the last 50 years will marvel at Gershenfeld's concise map; most persons will be reassured by its existence; and some may even work their way through various applications. (Yes, I mean you, Owen Walker!) It remains to be seen where the various paths lead from the Media Lab, but pedagogic innovations such as this one are likely to be judged a considerable success.
Toward the end of ''When Things Start To Think,'' Gershenfeld acknowledges that operating the Media Lab in the environment of MIT is a little like driving a bumper car in an amusement park. There is steady jostling, he says, but it would be impossible to function without the institutional setting.
A spirited competition is under way within MIT between the more traditional Laboratory for Computer Science and various affiliates and the Media Lab. Both are planning new buildings; as architects they have engaged rivals who are contending to be known as the best of their day, Frank Gehry and Fumihiko Maki. The buildings will stand only a couple of blocks apart. And so it goes. Only good can come of it.
Dow:
