Teaching machines to listen.
Computer scientist Paris Smaragdis is building some of the world’s most advanced “machine listening” systems–software that uses sound to locate people moving through rooms, monitor machinery for impending breakdowns, or activate traffic cameras to record accidents. It’s work you have to hear to understand. Click here to listen to a podcast about the technology.
Disguising data as noise.
Apostolos Argyris has joined theoretical mathematics with solid-state physics to demonstrate a novel, nearly unbreakable encryption method. As a graduate student at the University of Athens, the Thessaloníki native implemented the first long-distance demonstration of “chaotic synchronization”: using a pair of laser diodes, amplifiers, a mirror, and more than 120 kilometers of underground fiber, he disguised a message as white noise and still retrieved it clearly on the other end.
The feat is a demonstration of the “butterfly effect,” the founding tenet of chaos theory. The theory holds that in sufficiently complex systems, even a slight adjustment of initial variables will produce a dramatically different result; conversely, if you can replicate input conditions precisely, you should be able to replicate the output, even if that output appears random at first glance. Argyris applied this principle by combining a digital message with a chaotic, rapidly fluctuating laser signal; the chaotic signal’s irregular shape masks the message and allows it to evade standard eavesdropping methods. An identically generated chaotic signal on the receiving end cancels out the first, leaving only the original data.
The technique currently delivers data at one gigabit per second (good enough for voice and video encryption). Argyris hopes to raise that to 10 gigabits per second by 2009. He doesn’t see chaotic encryption replacing software encryption, but it could soon offer an extra layer of protection for the most sensitive communications.
His passion is finding ways to connect mobile devices, sensors, and robots directly--without the need for a base station. It's called "ad hoc" networking.
Sending and receiving data can be difficult for people on the move, especially when they travel out of the range of a stationary wireless base. Enter “ad hoc networks,” in which radios communicate directly with each other–no central base station required.
Prithwish Basu, a scientist at BBN Technologies in Cambridge, MA, has developed algorithms that dramatically reduce the chance that a wireless network will drop connections or fail, all while decreasing the energy consumed by battery-powered radios. The algorithms will work for networks of sensors, for people carrying mobile computers, or even for groups of robots with onboard radios.
The U.S. Defense Department is interested in testing these designs as a way to help keep soldiers in touch on the battlefield, but Basu’s true passion is finding ways to incorporate ad hoc networks into everyday life.
TR: What are ad hoc networks useful for?
Prithwith Basu: To keep mobile soldiers in contact with one another and with command, ad hoc networking is the most attractive solution. On the commercial end, people were a little skeptical a few years ago. But sensor networks have really taken off recently. People are buying sensors and deploying them to monitor different phenomena: traffic, pollution, wildlife habitats, train schedules. So I’ve been pushing my company to look more into commercial aspects.
TR: What nonmilitary applications have you considered?
PB: When I was a graduate student at Boston University, I proposed networking parking meters. You put wireless transmitters on each parking meter in a city; then you add a sensor that can tell whether there is a car in that spot. If my car also has a transmitter and I am entering Harvard Square from one direction, I can ask my user interface for a parking space near the law school. My car will query the nearest meter, and if that one’s full, it will propagate the query until it finds a free spot, and then even reserves it, if possible. The network could also allow parking police to determine violators almost instantly. A number of undergraduates are now implementing a network based on these principles.
TR: Could the mobile-robot networks you’ve developed be useful in daily life?
PB: If you have a group of robots running the appropriate algorithms, you can program them to perform any collaborative task. Suppose you have a network of Roombas [robot vacuums] and you want to clean a … hotel floor. If one room is dustier, one Roomba would send a message to three others: “Hey, this space is very dirty, so please come over.” If your protocols are working correctly, you’ll get a distribution of [robots] that follows the dust pattern. Robots connected this way could also help with relief operations in large disaster sites.
In this video clip, Basu further explains “ad hoc” networks.
Making screens crystal clear.
As a graduate student at MIT, Seth Coe-Sullivan created a possible way to make amazingly bright and colorful displays using quantum dots, tiny semiconductor crystals that emit specific colors of light when energized. Researchers had made light-emitting devices out of quantum dots before, but Coe-Sullivan improved their efficiency by developing a technique that produces a uniform, ultrathin layer of the dots between two flat electrodes. He then cofounded QD Vision in Watertown, MA, with the goal of arranging quantum dots into pixels to create full-color, flat-panel displays for cell phones, laptops, and televisions.
As chief technology officer, Coe-Sullivan has already led the fabrication of small monochromatic displays as well as the development of a process that will lead to full-color versions. These screens will produce brighter images and more saturated colors than conventional displays–while reducing both component cost and power consumption by half. They could appear in cell phones by 2009.
Better virtual crash dummies.
About 1,000 fetuses die every year in the United States when their mothers are injured in car crashes. But little was known about how automotive restraints might be improved for pregnant women until
Stefan Duma created the first computer model of a pregnant abdomen. Duma, director of Virginia Tech’s Center for Injury Biomechanics, has also produced important models of the human eye and thorax. And he developed a sensor-rigged football helmet that has been commercialized to help doctors understand impacts to the head.
TR: We’ve all seen those iconic crash-test dummies. What’s wrong with them?
Duma: We’ve come a long way using these dummies, but they are not 100 percent humanlike. They only measure discrete locations where sensors are located. They are fairly rigid compared to a human. In a computer model, we can make them more flexible, and we can have an infinite number of sensors.
How did you create your model of a pregnant abdomen?
When my wife was pregnant, I realized there was little research in this area, so I got into it. We created the model using data from sources like studies of placenta strength and validated it by studying the involved pregnant women. With this computer model, we can now “see” inside the abdomen during the accident.
There’s a notion that lap belts might harm rather than protect pregnant women. What does your model show?
We’ve shown that the three-point belt is far superior to just a shoulder belt. But we’re hoping to help make devices to better protect pregnant women. We’ve been better able to show the injuries they are susceptible to. We’ve transferred this model so that it is commercially available, and we are working with [vehicle] manufacturers on prototype pregnant belt systems.
How do these prototypes differ from existing seat belts and air bags?
There are a lot of things you can do. You could have different attachment points, different sizes, different materials construction for seat belts, for example. And we’ve found that the steering wheel is a key issue. That’s where a lot of the fetal deaths happen.
Simple technologies save energy and lives.
Back in graduate school, Christina Galitsky could boil her life’s work down into something like the title of a journal article: “The reversibility of proteins absorbing onto a surface,” she says. But since she dropped off the PhD track and, later, took a job up the hill at Lawrence Berkeley National Laboratory, the question “What do you do?” has turned into a stumper. “I guess now I say, I try and work on … sort of innovative solutions to … wait, what do I say?” she says with a laugh.
Officially, Galitsky spends about two-thirds of her time developing tools to help companies diagnose energy inefficiencies and find new technologies that conserve power without sabotaging profits. But a glance around her office suggests that a host of other problems occupy her mind. On the floor lies an aluminum contraption, an efficient cookstove designed to fight deforestation in the poor world; she and her colleagues believe it might also keep refugee women in Darfur, Sudan, closer to their camps and out of the path of sexual assault (see “Christina Galitsky on her work with refugees”). Later, she’ll don a rust-stained lab coat and check on her students, who are testing a low-cost scheme to filter arsenic from the drinking water in Bangladesh.
“I’m involved in a crazy range of things,” she admits. “But it would be hard to work on one thing all day long, five days a week.”
When Galitsky left the chemical-engineering program at the University of California, Berkeley, with a master’s in 1999, she found work testing the quality of California’s surface waters. Quickly, she recognized that much of the contamination she encountered came from energy-related sources, such as the power industry. Eager to fight pollution rather than just measure it, she joined Berkeley Lab in 2001. There, she began diagnosing energy waste in nearly a dozen industries, from concrete to beer.
A couple of years later, when the California Energy Commission put up seed money for research into energy efficiency, she got more ambitious. Technologies such as occupant-sensing ventilation systems can help businesses conserve energy, and they often pay for themselves in just two or three years. But traditionally, business owners have had to discover those technologies and determine the costs and benefits by themselves–a huge barrier to adoption. Galitsky and her colleagues decided to test a new approach with California’s wineries, offering them a system that would make it painless to spot their energy waste and find cost-effective ways to do better. (The wine industry requires huge amounts of power: 400 gigawatt-hours–enough to power nearly 60,000 homes each year in California alone, and most of that during the summer and fall, when conservation matters most.)
Galitsky and her colleagues partnered with Fetzer, a large California winery, and started collecting data. It was tricky. In some industries, managers put power meters all over their plants, so they know how much energy each step of the manufacturing process demands. Wineries, however, tend to install just one meter for the whole operation. So Galitsky tallied everything from the number of grapes crushed to the sizes of Fetzer’s refrigeration tanks and pieced the data into rough estimates of the power used at each stage of the winemaking process. Then she and her team surveyed wineries around the globe to identify the most energy-efficient technologies employed at each stage. The result is a tool called BEST-Winery, based on Microsoft Excel. It poses a series of questions, then spits out a score that compares the winery under review with a hypothetical winery of the same size and scale that uses the industry’s best conservation technologies.
Other systems for measuring energy efficiency stop there. But BEST-Winery suggests more than 100 conservation technologies and runs a cost-benefit analysis for each one–a significant innovation for this kind of tool. Winery owners can mix and match different technologies and find comprehensive approaches that fit their budgets.
The state of California presented an award for energy innovation to Fetzer and to Galitsky’s team, which is readying a European version of the software. But wineries are only the beginning. Galitsky thinks a similar tool could work for a wide range of businesses. Soon the Berkeley team will test that theory at the national level, working with the six countries of the Asia-Pacific Partnership on Clean Development and Climate. These countries want to measure the environmental footprints of their cement, aluminum, iron, and steel industries; instead of evaluating a single plant, the software would grade entire countries.
All that energy research is good for the planet, but after a while Galitsky wanted to find more immediate ways to help the worlds poor. So she started hanging out with do-gooder research groups. At a Berkeley meeting of Engineers for a Sustainable World, she met Ashok Gadgil, a senior scientist at Berkeley Lab who had interests similar to hers. Together, they began to look at crises in Darfur and in Bangladesh.
The problem they identified in Darfur is simple, and gruesome. More than 1.6 million citizens of this Sudanese region have been displaced by civil war, with hundreds of thousands crammed into refugee camps. They have to eat, and to eat they have to cook, and to cook they need firewood, but they have already stripped the areas around their camps bare. Local women must wander for hours outside the relative safety of the camps to gather wood. This leaves them vulnerable, and international observers have documented an epidemic of rape by roving gangs.
Nongovernmental organizations have suggested that, along with other measures to protect women, better cooking tools could reduce the need for firewood. While there have been a ton of competing ideas–everything from clay ovens to solar cookers–and a ton of opinions about which ones work best, none of them had been tested in Darfur with any rigor. So Galitsky and Gadgil went to Darfur, partnering with aid group CHF International.
Traditionally, Sudanese women balance their cooking pots on three stones over a wood fire. But lots of heat escapes, and much of the wood simply chars and smokes. As a better option, Galitsky and Gadgil looked to a simple metal stove designed in the 1980s by the Indian nonprofit organization Development Alternatives. Galitsky held a demonstration in Darfur–kind of a cross between a lab experiment and a Ron Popeil infomercial. Before a large crowd, she set up the traditional three stones, the metal stove, and a mud stove popular with many aid groups. A handful of community leaders chopped wood and stacked it into 250-gram piles. Then Galitsky cooked three separate meals, so the women could see how much wood each stove used. “The stone fire used ten piles, the mud stove used nine, and the metal one used only four or five,” she recalls.
Despite the metal stoves performance, the researchers knew it would need modifications to fit life in Darfur. So Galitsky interviewed dozens of women about their lives and their cooking duties. She determined that the stove would need a windshield, to control the gusts that whip through the camps, and stakes for stability when the women stir their assida, a sticky dough that makes up most meals. She and Gadgil also need to make sure the stove can be manufactured quickly and cheaply. But the technology shows promise. “We are very excited,” says Maha Muna of the United Nations Population Fund in Sudan. “The U.N. and [aid groups] have funded so many projects on fuel-efficient stoves as pilots, but CHF and Berkeley Lab are actually carrying out the analysis we need to be able to determine what should be replicated.” The Berkeley researchers plan to begin delivering test stoves to refugee families this fall; they hope to produce 300,000 by next year.
In Bangladesh the problem isn’t food; its drinking water. In the 1970s, Unicef dug wells all across the country so that Bangladeshis could stop drinking contaminated surface water. The aid groups motives were pure, but the wells were not. Most were in areas with high concentrations of arsenic–in some cases, more than 100 times the level the World Health Organization has deemed safe. “It has been called potentially the largest mass poisoning in the history of the world,” Galitsky says.
Recently, the United States lowered its own limit on arsenic in drinking water by 80 percent, and states are interested in new technologies to meet the tougher standard–interested, and putting up money. Gadgil and Galitsky saw an opportunity. With a $250,000 grant from the California Energy Commission and $100,000 from the American Waterworks Association Research Foundation, they are developing a filtration system that could work both here and abroad.
Arsenic is easy to filter at a big water-treatment facility, but engineers can’t scale existing technologies down enough to serve individual families, or make them cheap enough for the poor world. Gadgil had an idea. Iron particles act like arsenic magnets, bonding tightly to the arsenic for easy disposal; but a filter made of pure iron powder would be prohibitively expensive. Layering a thin coat of iron onto waste ash from coal-fired power plants, however, would offer similar arsenic–attracting surface area at a fraction of the cost.
Getting the ash and the iron to stick together turned out to be a challenge. But after a dozen failed attempts, Galitsky and Gadgil came upon the solution: washing the iron-coated ash particles with lye and letting them get good and rusty.
The result, which looks something like dark curry powder, will capture nearly all the arsenic in a beaker of contaminated water. The researchers still need to figure out how water should pass through their hybrid ash-and-iron substrate, and what real-world conditions might interfere with its performance. But they believe filters made with their new medium could be effective enough to meet stringent safety standards yet still affordable enough to sell to Bangladeshi households.
With Galitsky and Gadgil’s method, a family could filter a years worth of water for less than about $2; it would cost at least $58 with today’s cheapest comparable technology. Galitsky talks about all her research with a real sense of urgency, and not just because people and the environment are suffering. For the problems she is addressing, big gains are tantalizingly close, and the rewards will be great–for the poor communities this kind of science can help, and for Galitsky as well.
“I felt so helpless,” she says. “And I still feel helpless. But at least now I’m doing something.”
Ram K. Krishnamurthy
Problem: As chip makers increase computing power by packing more and more transistors into the same space, they could end up with chips that become too hot to operate. In fact, if the pace continues, by 2010 computer chips will theoretically generate as much heat per square centimeter as a nuclear reactor.
Solution: By carefully plotting chip circuit paths, Intel engineer Ram Krishnamurthy has minimized energy leakage and improved performance; his prototype circuits run fi ve times as fast as those in today’s PCs but consume 20 to 25 percent as much power. In less than a decade Krishnamurthy has amassed 53 U.S. patents relating to circuit design. Intel, IBM, and Hewlett-Packard are already incorporating aspects of his work into advanced circuits that will lead to servers requiring no costly cooling systems, as well as to laptops with longer battery life. And if PCs soon run at a cool 10 gigahertz, it may be thanks largely to Krishnamurthy’s work.
Designing games with new realities.
The pay phone rings. This much you know: in the 26th century, a young girl’s personality was uploaded into an artificial intelligence called Melissa aboard a military starship, which was then thrown back through time to your own year, 2004. Now the starship’s crew is dead; Melissa survives, damaged and stranded on the Internet server of a beekeeper in California’s Napa Valley. From that beekeepers website, ILoveBees.com, you gleaned this phone’s GPS coördinates. You cup the receiver to your ear as Melissa speaks…
If this seems familiar, you were probably among the 600,000 active players of I Love Bees in 2004. To date, it remains the most widely played “alternate-reality game.” Jane McGonigal, who recently completed her PhD in performance studies at the University of California, Berkeley, was tapped to become one of the game’s four “puppet masters” after she began working with 42 Entertainment, a company in Emeryville, CA, that creates entertainment-based marketing campaigns and original content.
Microsoft had hired the company to promote Halo 2, the video game it was about to introduce, by creating a massively multi player alternate-reality game. McGonigal designed the real-world “missions” that took advantage of and shaped the way the players organized themselves. Creating and engaging this worldwide community helped make I Love Bees so successful. McGonigal argues that alternate-reality games use network technologies–e-mail, websites, Internet chat rooms, text messages, and phone calls–to construct new types of communities whose “collective intelligence” lets them solve problems no member could solve alone. In 2005, she and the I Love Bees team won the Game Developers Choice Awards’ Innovation Award and the International Academy of Digital Arts and Sciences’ Webby Award.
McGonigal has continued working with 42 Entertainment. In 2005 she developed Tombstone Hold ‘Em, part of a 2005 promotion for Activision’s game Gun; crowds congregated in historic cemeteries to play poker using tombstones instead of cards. Such novel uses of public spaces are another way she engages players. Her own work as a game designer is fed by watching players interpret the missions she designs: “They always think of far more interesting things than anything I could imagine.”
Making fuel ethanol more cheaply.
Ethanol offers a renewable supply of auto fuel, and it can also reduce pollution. But in the United States, fuel ethanol is made almost exclusively from corn kernels, and it provides little more energy than raising, harvesting, and processing the corn consumes.
Determined to help wean the world off petroleum, bioengineer Michael Raab is putting enzymes into corn that will make it easier and cheaper to convert the entire plant–kernels, husk, stalk, and leaves–into ethanol. These proteins allow processors to break the complex carbohydrates that make up most of the corn plant into simple sugars that can be easily fermented into ethanol.
TR: What is your company, Agrivida, doing?
Raab: We’re taking the processing enzymes used to break down the leaves and the stalks and reëngineering them so they have no activity when they’re in the plant. Then after you harvest the leaves and stalks, you can discretely activate the enzymes with a mechanism that we can control; we are using increases in temperature and acidity that coincide with normal ethanol-processing conditions.
Why is it important to design these enzymes to be inactive until needed?
Putting enzymes in plants is not a new idea, but it has not been very successful because the enzymes have a dramatic negative effect on plant development. Our engineered enzymes get around this problem by delaying their activation, which allows the plants to grow normally.
How does this process improve ethanol production?
Our enzymes allow us to more efficiently process the plant, which increases ethanol yields per acre by about 50 percent while decreasing costs per gallon by about 30 percent. In a commodity market that is growing as fast as ethanol is, that’s a huge deal.
Will this process work on other plants?
We are beginning with corn; once it’s proven there, we will transfer the traits into other crops, such as switchgrass, poplar, and sugarcane. That will allow geographic regions other than the Midwest to make ethanol, which decreases transportation costs; it also expands growing seasons and provides farmers with greater alternatives in their crop selection.
Making mobile secure.
Problem: PCs have long been under siege from malware writers and identity thieves, but as more mobile devices connect to the Internet, they too are becoming targets–which could have even worse repercussions. Viruses that infect cell phones or PDAs can spread via Bluetooth, a wireless technology commonly used to connect handheld gadgets to PCs and onboard computers in cars. The consequences could be severe if a virus hopped from a phone to a networked PC behind a corporate firewall–or to a car’s navigation system, jumbling GPS information or worse.
Solution: To make mobile devices more secure, Anand Raghunathan and his team at NEC Laboratories America have given them a supplementary processor, dubbed Moses. The processor performs all of a device’s security functions, such as encryption and user authentication. Using a separate security processor isolates all the system’s encryption keys, which protect passwords and personal information. So if a virus did hit a device, it couldn’t access the passwords needed to log in to a bank account or an office computer; its effects would be limited. In addition, because Moses is specially designed to encrypt and decrypt data efficiently, a phone using the processor requires one-fifth the time and consumes one-third the power of a traditional cell phone performing the same tasks.
The security processors will be installed in millions of cell phones over the next few years, Raghunathan says. And he predicts that other devices will benefit from Moses, too. Radio frequency identification tags, networked sensors, and MP3 players–any small gadget with a limited power supply–could use the technology.
Faster defenses against computer viruses.
Problem: Current antivirus systems rely on humans: when a network attack comes to light, security analysts begin looking for a string of bytes–a “signature”–that uniquely identifies the malicious program. The signature must be downloaded (often automatically) before software can identify and block attacks. But the whole process takes hours–or days–while attacks can infect up to 55 million computers per second.
Solution: Sumeet Singh has completely automated worm and virus detection, putting defenders on the same footing as attackers. As a graduate student at the University of California, San Diego, Singh realized that worms and viruses move through a network differently from normal traffic: malicious code strives to reproduce and propagate itself rather than simply to travel from point A to point B. So he created software tools that scan for snippets of data that exhibit such behavior.
Incorporated into a network router or switch, Singh’s software can identify malicious code when it first enters a network and generate “vaccines” to combat its spread. In June 2004, Singh cofounded NetSift with his PhD advisor, computer science professor George Varghese, to commercialize his technology. Cisco purchased NetSift just a year later; Singh has since led the integration of his techniques into Cisco routers and switches. He hopes that this technology, able to scan more than 20 gigabits of data per second, will eventually stop viruses and worms as soon as they pop up.
Lighting cellular movies.
Alice Ting’s movies won’t fill any theaters, but they are breaking ground in using what’s called “fluorescence imaging” to reveal the minute inner workings of cells in unprecedented cinematic detail.
To make her biology-in-action movies, Ting, an assistant professor of chemistry at MIT, needed a bright and efficient light source. So her lab developed a way to, effectively, glue superbright fluorescent “tags” directly to proteins of interest. The standard way to make proteins glow is to use green fluorescent protein (GFP), originally isolated from jellyfish, and cousins that fluoresce in different colors. But these proteins produce relatively dim light, making it difficult to see single molecules or in vivo processes. They also have to be genetically fused to the proteins being studied; this can alter the proteins’ behavior and prevent them from freely moving around and into cells. In contrast, Ting used quantum-dot tags that are up to 100 times brighter than GFP and interfere less with the observed proteins.
Other labs had labeled proteins with quantum dots–nanocrystals that fluoresce in different colors depending on their size–but attached them with the help of bulky antibodies. Ting’s lab did away with these clunky connectors and, at the same time, created far more secure ones, fusing small protein “linkers” to both the quantum dots and the proteins of interest. Ting then used an enzyme to join her two linkers, and voilà–she could observe a living cell in action. Nor is the linking system limited to quantum dots: it can be used for any tag.
“Alice Ting is a true innovator and is one of the best chemists of her generation,” says Timothy Swager, chair of MIT’s chemistry department. “Scores of research groups around the world are already applying her methods.” One of Ting’s latest projects is to fluorescently image the junction between nerve cells, illuminating a biochemical process that appears to play a key role in learning and memory. So it may be possible one day to see an actual film of how a brain learns. “Mammalian cells are so beautiful and funky,” says Ting–with the appreciation of a true director.