Celebrating Science and Engineering Breakthroughs III
Sallie Chisholm, Lee and Geraldine Martin Professor of Environmental Studies and Professor of Biology, MIT; Regina Barzilay, Associate Professor of Electrical Engineering and Computer Science, MIT:; Anette Hosoi, Associate Professor of Mechanical Engineering, MIT; Nergis Mavalvala, PhD '97, Professor of Physics, MIT
Description: Although these three speakers travel in quite disparate worlds -- natural language processing, mechanics of tiny organisms, and violent cosmic events -- they convey a comparably infectious enthusiasm for their research.
In the early days of artificial intelligence, "people had the na've idea that if you took a computer, and fed it with enough human knowledge, it could eventually understand human language," relates Regina Barzilay. In the 1990s, after this "spectacular failure" in methodology, a new approach evolved: training a machine to infer knowledge from piles of data. Some successes are already in evidence (IBM's chess" and Jeopardy"playing programs). Barzilay wants to move this learning further, toward "grounding interpretation of language in the real world." She describes algorithms that could take over the often exasperating grunt work of installing new Windows software on computers, by reading common instructions and executing the actions. Her system learns to map words with increasing accuracy by using feedback. She is also investigating how to get a machine to understand and act in a much more complex environment, such as the Civilization computer game. Using simulations, her algorithms make predictions about the best possible moves (e.g., invade another nation), and can even absorb instructions from the game's manual, which should make the system "good enough to play against humans." Her ultimate goal: "enabling computers to function competently in a world rich in unstructured information."
For tiny organisms in a fluid environment, a key challenges is viscosity, says
. In particular, Hosoi focuses on eukaryotic cells with spherical heads attacked to flexible tails. Remarkably, the diameters of these tails are the same across all species -- between 250"400 nanometers. Whether a hair in a lung, or a cell in green algae, these tails are all made of microtubule structures that can slide and bend in similar ways. By analyzing these common properties, Hosoi can predict optimal morphology and kinematics of comparable tailed microorganisms. For instance, analyzing sperm cells, Hosoi's team determined that the best ratio of tail to head for swimming efficiency is 12 _ and not just for sperm. "I don't care what the species is, what it's made out of, the tail should be 12 times as long as the headGetting something as clean as that is very exciting!" One outlier in the study: the Bandicoot, whose sperm's fat tail did not share the same radius as all the others. This kind of optimization research, says Hosoi, can only improve, as the computational costs of analyzing vast repositories of biological data continue to drop. When you begin to understand underlying principles in biological structures, she says, "you can move on to inform engineering designs."
Nergis Mavalvala takes her audience "to a slightly uncomfortable side of the universe, warped and violent." She searches the cosmos using "a completely new messenger -- gravitational waves that travel to us from distant sources." Mavalvala credits Einstein (by way of Newton) for first proposing these waves. He developed a picture of space and time "as a fabric" that can be dented by massive objects, exerting a gravitational pull. Massive objects bouncing around or vibrating caused "ripples of space time itself" -- hence gravitational waves. To produce these waves, says Mavalvala, you need lots of mass and rapid acceleration, explosions and collisions _ produced by compact stars slamming together, merging black holes, and the conditions that immediately followed the Big Bang. Using special detectors (laser interferometers), Mavalvala has been trying to detect gravitational waves. The first measurements taken by these detectors found a gamma ray burst explosion, but no gravitational waves. A next"generation detector is on the way that makes it possible to "listen to more distant sounds," but Mavalvala notes that at this level of "exquisite measurement," you pay the price of "quantum uncertainty." However, she is certain the elusive gravitational wave will finally be captured, and "we will be testing general relativity: the first direct observation of ripples of space"time."
About the Speaker(s): In addition to her other appointments, Penny Chisholm currently serves as co"director of Terrascope, an MIT learning community for freshmen. She is also a visiting scientist at the Woods Hole Oceanographic Institution. From 1988"1995, she served as the MIT Director of the MIT/Woods Hole Joint Program in Oceanography.
Chisholm received the 2005 Huntsman Award for Excellence in Marine Science, and is a Gordon and Betty Moore Foundation Investigator in Marine Science. She has published papers in PNAS and Nature. She received her Ph.D. in Biology in 1974 from S.U.N.Y. Albany.
Host(s): Office of the President, MIT150 Inventional Wisdom
It looks like no one has posted a comment yet. You can be the first!
You need to log in, in order to post comments.
More from MIT World — special events and lectures
Added almost 5 years ago | 00:56:49 | 4427 views
Added almost 5 years ago | 01:27:00 | 3986 views
Added almost 5 years ago | 00:59:49 | 3060 views
Added almost 5 years ago | 01:28:00 | 3203 views
Added almost 5 years ago | 01:21:00 | 8689 views
Added almost 5 years ago | 00:55:06 | 4555 views