|
Notebook: November 19, 1997
Princeton helps build next-generation Internet Princeton is one of 115 research universities in the nation helping build the digital backbone of the next-generation Internet, dubbed Internet2. The high-capacity computer network is expected to operate 100 times faster than the current Internet and to support a new class of applications in ways not possible on today's network, said Ira H. Fuchs, vice-president for computing and information technology. Internet2 will make it possible for faculty members to speedily transmit multimedia-rich content and conduct distance-learning courses. And it might make possible virtual labs and shared virtual work spaces where researchers from different institutions could work together. The effort started about a year ago. Project Director Mike Roberts expects the new network to be operational by the end of 1998. In addition to the research universities (all the Ivies are members), Internet2 collaborators include federal agencies and computer and telecommunications companies such as MCI, Sprint, and IBM. Internet2 members plan to develop research and teaching applications that will take advantage of the network's increased capacity. Several applications were demonstrated at a meeting of Internet2 leaders in October in Washington, D.C. One is a scanning electron microscope that can be controlled over the network; another is a "data mining" project that allows information to be culled from distant high-powered computers and analyzed almost instantly. Joining Internet2 requires financial resources and substantial technical staff. An institution has to commit to spending roughly $500,000. For now, Internet2 is reserved for academic use. But within three to five years, said Roberts, it will be available to all users online. It's not clear yet what phone or cable companies will end up providing access to the fiber-optic cables and switches. There's no telling exactly how Internet2 will be used by faculty members and researchers. Said Fuchs, "Internet2 is being built on a Field of Dreams approach-if you build it, they will come." How to make use of it will require imagination on the part of users, he explained. Currently, groups of universities are building gigapops, which stands for gigabit capacity point of presence (part of the Internet2 infrastructure). A gigapop is a location with new switching equipment that combines traffic and puts the material it receives onto a higher-capacity link that will connect to other gigapops, said Roberts. Princeton is building a gigapop with Rutgers University, the University of Pennsylvania, Lehigh, and the University of Delaware, said Fuchs. The gigapop, he said, might be located at one of the Rutgers campuses and will probably be completed in less than six months. Gigapops will be connected to each other via the National Science Foundation's Very High Speed Performance Backbone Network Service (VBNS)-the skeleton of Internet2, said Roberts. VBNS is a high-capacity network provided by MCI and funded primarily by the federal government. According to Roberts, it will take several years for Internet2 to completely replace the current Internet, which member institutions will continue to use for e-mail and searching the Web. The biggest problem facing Internet2, said Fuchs, is one that isn't addressed by its builders: how people will pay to use it. Currently Princeton pays for its pipeline into the network, and students, faculty, and staff never see a bill for their time on the Internet. The new system will have to prioritize the transmission of data and information sent over Internet2 by charging for use. The Web site for Internet2 is located at http://www.internet2.edu/. n
By the numbers: Ivy Minority Enrollment
Class Act: Examining the limits of learning What is knowledge and how do we acquire it? These questions, central to the enterprise of any university, are being examined in an innovative interdisciplinary course called Epistemology and Learning Theory, taught by Professor of Philosophy Gilbert H. Harman and Associate Professor of Electrical Engineering Sanjeev R. Kulkarni. This union of apparently disparate approaches-that of a philosopher and that of an electrical engineer-in the study of learning theory makes perfect sense, say the instructors. "In recent years," observes Harman, "aspects of the philosophical problem of induction have been studied by logicians, computer scientists, psychologists, engineers, philosophers, and statisticians-among others." Harman, whose broad interests within philosophy include theoretical and practical reasoning and cognitive science, joined the university faculty in 1963. Currently he directs the Program in Cognitive Studies. When Kulkarni, who studies statistical pattern recognition and machine learning, came to Princeton in 1991, Harman contacted him to explore areas of common interest. "We agreed that there are fundamental advances that cover so many areas, they should be made accessible to a wide audience," says Harman. The result is Philosophy 218/Electrical Engineering 218, a course taught for the first time last year. "We are attempting to develop as precise as possible a definition of the nature and limits of learning," says Harman. "There are a number of paradigms that can be used to explore the issue." Drawing inferences from data and deciding if we are justified in drawing those inferences are philosophical problems, explains Kulkarni, "but there are mathematical methods that allow us to quantify those problems."
Lectures in Epistemology and Learning Theory display team teaching in its highest form. "In each lecture," said one student, "professors Harman and Kulkarni both speak, often alternating, presenting differing views on the same material." As Emrah Cevik '98 observed, "It is more like a conversation than a lecture." This "conversation" proceeds in an almost Socratic way, with questions posed, contested, and resolved. Harman and Kulkarni don't so much finish each other's sentences as extend each other's thoughts. In a recent class, they examined the mathematical concept of finite and infinite sets, intrinsic to an inquiry into what can be known and with what degree of certainty; they were trying to establish the concept of proof. Kulkarni, who admits that his greatest challenge in preparing the course has been to "find ways of bringing specialized technical material to students who don't have a technical background," is constantly alert to concepts or terms with which students may be unfamiliar. To illustrate finite sets with the same cardinality (the same number of members) he points to the notion of a one-to-one correspondence, noting that it is "just like what a child might do-line up a row of pennies and a corresponding row of beans." The class discussions are both theoretical and technical; what's surprising is that they are often entertaining as well. The professors clearly enjoy their joint enterprise, and the students enjoy the spectacle of, in Cevik's words, having "each professor talk under the watchful eye of the other." In addition, Harman's idiosyncratic humor frequently leavens discussion without ever distracting from it. During the talk on cardinality, he described "two sets that are finite-the set of people on this side of the room," indicating himself and Kulkarni, "and the set of people on that side of the room," indicating the 13 students. "We can certainly map the set of people on this side of the room in various ways so that each of us goes into distinct members over there. But there's no way to map all of you people into us," continued Harman. "You may not have realized it, but that's how we know with certainty that there are more of you over there than there are of us over here." Students attend two lectures a week and one precept. In addition to a term paper, there are also weekly problem sets such as: (1) A process produces a sequence of numbers as output. Supposing that the first three numbers produced are 2, 4, and 8, name two of the most likely rules for the sequence. (2) What would you predict as the next number? (3) What justification could be given for that prediction? (See below for answers.) For those who lie awake at night pondering such problems, there is an online discussion group. In creating what Kulkarni calls "a bridge between humanities and engineering," the seasoned philosopher and the young engineer have attracted students of varied interests-half A.B. and half B.S.E. students, with a range of departments represented. Epistemology and Learning Theory is probably unique, the instructors believe. "As far as we know, there is no course anywhere that talks about these matters at the level we do," says Kulkarni. In fact, comments Harman, "We tried to find a course like this somewhere so we could look at the materials, but there just wasn't anything. We may end up writing our own text." -Caroline Moseley Answers: (1) The number doubles each time; the difference between one number and the next increases by 2 each time. (2) 16. (3) The first rule is simpler, and so more likely than the second rule.
Understanding Knowledge Thinking Things Through: An Introduction to Philosophical Issues and Achievements, by Clark Glymour (MIT Press, 1992)-A discussion of some key issues about inductive learning, including inductive skepticism, notions of infinity, and the nature of computation.
"Language Acquisition," by Stephen Pinker, in Language, Lila R. Gleitman and Mark Liberman, eds., vol. 1 of An Invitation to Cognitive Science, 2nd ed. Series edited by Daniel N. Osherson. (MIT Press, 1995): 135-By asking how aspects of one language might be learned, linguists are able to discover universal principles that apply to all languages.
Machine Learning, by Tom M. Mitchell (McGraw-Hill, 1997)-Looks at how we can design machines that learn, for instance, a computer program that recognizes faces.
"Concepts and Categorization," by Edward E. Smith, in Thinking, Edward E. Smith and Daniel N. Osherson, eds., vol. 3 of An Invitation to Cognitive Science: 1-Examines how humans form concepts and categories.
Pattern Classification and Scene Analysis, by Richard O. Duda and Peter E. Hart (Wiley, 1973)-A classic text on some basic approaches to statistical pattern recognition.
Expanded Woolworth Center Dedicated Designed by Juan Navarro Bal-deweg of Madrid, the new center features an atrium illuminated by huge skylights. The Scheide Music Library--named for William H. Scheide '36, a leading expert on Johann Sebastian Bach--represents the largest portion of the expansion. Before the library's completion, holdings of the music department were housed in Fire-stone Library.
In Brief
Prize: Director of the Creative Writing Program Paul B. Muldoon won the 1997 Irish Literature Prize for Poetry, sponsored by The Irish Times, the Dublin-based newspaper. Muldoon, who is Irish, was honored for his 1996 volume New Selected Poems 1968-1994 (Faber and Faber, London). The newspaper cited the manner in which the collection "highlights Muldoon's witty, personal voice and inventive use of language, which often draws on daily routine and the topical." The prize carries a cash award of about $8,000. Former Irish President Patrick Hillery honored Muldoon and winners in other categories at a ceremony in Dublin on November 20.
ADLAI Papers: A recently completed preservation and cataloging project at Seeley Mudd Manuscript Library will give researchers improved access to the documentary legacy of Adlai E. Stevenson '22, who served as governor of Illinois and UN ambassador. Stevenson and his three sons gave his papers to the university in the 1960s. As part of the recent project, the Stevenson family has added many personal letters to the collection, which was not fully catalogued until this year. Also at Mudd Library is an exhibition that chronicles 75 years of Native American history titled "A Voice in the Wilderness: Selections from the Archives of the Association on American Indian Affairs." The exhibition runs through February 1, 1998.
|