Amidst all the recent talk of improving science education at Yale, WEEKEND set out to find some of the wackiest research projects around. Naturally, we turned to monkeys and robots. Little did we know that in examining our primitive ancestors and the machines that will one day kill us to rule the world, we would actually learn something about ourselves. From monitoring the behavior that makes us human to studying its origins, WEEKEND investigates.
In the hit television show “The Wire,” dealers transfer drugs to buyers through an intricate web of deception and secret signals. To the untrained eye, the interactions seem hardly suspect; subterfuge passes by the law unnoticed. But for Andreas Savvides, associate professor of electrical engineering and computer science, this does not have to be the case. He imagines a world in which sensors could identify these interactions and, based on algorithms of human behavior, pinpoint when and where a criminal conspiracy takes place.
“If you have a group of people conspiring to do something, and people possess information about how groups work, then you can use that prior knowledge to say that a group of people is selling drugs or plotting to plant a bomb,” he said. “Just by analyzing how people interact, we can figure out what is going on and say which people are the primary suspects.”
The sensors, which would be small enough to fit in the palm of your hand, would even be able to figure out each person’s role in the scheme. For example, Savvides continued, if there were five people involved in a drug deal, the sensors would be able to tell you who was selling the drug and who was transporting the package. To aid in his analysis of group functioning, Savvides works with a team of behavioral scientists at Johns Hopkins University’s applied physics laboratory.
For Savvides, his work is rooted in an exploration of how sensors can perceive their environment and operate within a larger network of sensor intelligence. Like a video game.
“If you [are] an avatar in the game, my work looks at how your environment can perceive your existence and use it to provide services to you,” he said, adding that his work is difficult since sensors lack the sophistication of the human brain.
Savvides said the idea for his work with sensors came to him when he first started teaching at Yale in 2003. He noticed that people’s use of networks was limited to data collection. The main focus was how to make better networks with more reliable data. That approach is practical, he said, but not very innovative. So he became interested in making sensory systems “speak” to him.
There are two steps to the sensor system. First, the sensors search their surroundings and identify interactions between people, some of which are benign and others of which may be conspiracies. Next, human experts input a set of data based on how groups act in conspiracy situations, and the sensors build a graphical model against the incoming data to check for a match.
“The human analyst is solving a puzzle of which they know only bits and pieces,” he said. “Sometimes, like in a drug deal, the sensor system will glue together the entire scenario for them, but for a big insurgency in Baghdad, for example, it will show what different groups are doing and the experts can say, ‘Hey, they are getting ready to strike somewhere.’”
Needless to say, his work has attracted the attention of security agencies. “It’s better not to name them,” he added with a laugh.
He stressed that the sensors do not violate privacy since they interpret moving bodies as dots, rather than bodies or faces. This can pose a problem for identification in situations when two people cross paths, because the sensors cannot distinguish which person is which after the interaction. But Savvides said one of his students found that if a person is carrying a smart phone, the data in the phone will interact with the sensors to identify him or her.
“Let’s say you have an iPhone and you start walking around in Commons,” he said, “the system will know that the dot is you.”
Aside from their sleuth work, the sensors also have more everyday applications. For example, a sensor planted in an elderly person’s home could send information about that person’s well-being to family members. Or a sensor placed in an office building could isolate individuals to determine how much energy each person uses and help make a person’s energy consumption more efficient. The common theme, Savvides said, is that the sensors are doing much of the work for you.
“You are an English major, and you probably own more than 10 microprocessors already,” he said. “But you are always the bottleneck, you have to explicitly touch the buttons to make them work. I am into this research because I think that intelligence that can be derived from things embedded in space can complement your cognitive ability.”
He added that sensor networks could help build super soldiers or assist people with physical or cognitive disabilities; if the applications pan out, intuition may become a thing of the past.
“There are vibrations everywhere.”
This is the first thing we were told upon entering the Koser Lab, where two graduate research assistants told us about their current project to convert vibrations in the environment into energy to power small devices. In this sense, they said, the sensors are similar to solar cells, where the sun supplies the energy in a self-sustaining system.
The lab is overseen by associate professor of electrical engineering Hur Koser, but this particular project is primarily the work of two graduate students. One of them, Bozidar Marinkovic GRD ’11, put his hands on the wall and asked us to do the same. “Feel for the vibrations,” he said.
“If we put our small device on the wall of this building, it will power itself up,” he explained. “Once it gathers enough energy, you can use it to sense temperature, light, the presence of people, the closing or opening of a door and more. This information can then be stored or distributed among a wireless network of sensors.”
The device that captures the energy is a small chip, not much larger than the tip of your finger. When it is finished, it will be about 1 millimeter thick. The chip is made of silicon, which is why it is known as “smart sand.”
One of the primary applications of the smart sand will be to assess the structural integrity of a particular building. Based on the vibrations in different walls around a building, the sand will be able to detect which parts are under the most stress, or if any area is particularly weak. The machines are smaller than most and operate without the use of batteries, making them easier to maintain.
“Once you put it in the wall, you don’t have to worry about it anymore,” he said. His colleague, graduate student Muhammet Uncuer GRD ’12 added that the technology is “green” as well, since battery use and disposal are nonissues.
But the sand has human applications too.
For example, if placed inside a football player’s helmet, the sensors could monitor traumatic injury. When the player is tackled or hit in the head, the sand would register the frequency of the vibrations in the helmet and wirelessly transmit that information to doctors, who could then determine if the player suffered from a concussion. The device has not been designed for this purpose yet, said Marinkovic, but it would be relatively simple to rework the geometry, making neat work of diagnosis. Smart sand might just make humans smarter, too.
KEEPING AN EYE OUT
The remote-controlled car of the future may not be the kind you’re familiar with. For one thing, you’ll be able to ride in it.
That is, if Eugenio Culurciello, associate professor of electrical engineering, is able to realize the potential of his latest project, an efficient supercomputer modeled after the human visual system. Known as NeuFlow, it mimics the neural network of the human eye to interpret its surroundings. One of the possible implementations of this idea would be in a car that could run on its own, by recognizing the road, other cars, people, stoplights and more.
Berin Martini, one of the lab’s employees, said the system will be made small enough to fit in everyday devices such as a telephone. At the same time, it will be cheap enough to mass-produce.
His favorite application of the supercomputer is its potential use in assisted living. With the increasing elderly population, we want to make sure that people are able to stay in their homes and be independent for a long as possible, he said. For example, many elderly people need to be monitored in case they fall down.
The old people could be monitored by cameras in their homes or sensors in their clothing, he said, but this usually causes problems. For one thing, most people are uncomfortable having cameras in their homes because they fear that their privacy will be violated. For another, attaching sensors to clothing is risky because the elderly person may forget to wear the sensor on any given day.
The solution, Martini said, is his lab’s system, which would process the data from a video stream of elderly people’s movements to figure out if anyone has fallen. No video leaves the home if the person is just walking around, but the moment he or she falls, an outside source is called in to help.
In the future, Martini speculated that people might be able to use the hardware to create autonomous surveillance gear. For instance, he said, firemen might have small, remote-controlled vehicles that would be able to navigate through collapsed buildings based on sensory data. The vehicles would have a supercomputer built in, and so would be able to act independently.
As our meeting ended, he quipped, “We’re one step closer to Skynet.”
NOT MONKEYING AROUND
Sometimes looking into the past can inform our futures. At least, that’s the guiding principle at Yale’s monkey lab.
Technically called the Comparative Cognition Laboratory (or CapLab), the center is home to 10 brown capuchin monkeys who regularly undergo noninvasive experiments in physical coordination and cognitive functions. Capuchins are one of the most intelligent of all monkey species, and so they are helpful tools for determining how our own brains work.
Although it is a center for serious scientific research into the primitive roots of the human psyche, the lab also embraces the inherent cuteness of the test subjects: A possibly confused capuchin peers off the page at the top of the website, and the lab has taken great strides to make the monkey names as adorable as possible.
Felix Leiter is the lab’s current alpha male. According to the CapLab website, he is the son of a high-ranking female in his old group and “has carried on the proud tradition.” The website explains that he has a “pretty face,” a large body, is very watchful and also excels at throwing and catching. If his name sounds familiar to you, you’re probably a movie buff: Mr. Leiter shares a moniker with James Bond’s CIA counterpart, first portrayed by Jack Lord in “Dr. No.”
All 10 monkeys are named after Bond characters, because, according to the website, “capuchins are a rather cunning species.”
But the CapLab is not just a center for giving monkeys humorous names.
The primary duties of the researchers and undergraduates who work at the lab are to conduct tests on the monkeys to try and ascertain which cognitive abilities they do and do not share with humans, CapLab Director Laurie R. Santos said.
Most cognitive research has focused on identifying where the so-called “superior” human abilities (such as our capability to identify causal relationships) came from, Santos said. But while the CapLab is continuing this branch of inquiry, the research has taken a turn since 2006.
THE DARK SIDE
“Recently, our lab has taken a slightly different approach,” she said. “We’re looking at things that we’re more worried about in humans: economic errors, prejudice, stereotyping and rationalization.”
Santos explained that her lab is trying to determine whether these negative characteristics are also shared with the monkeys. If they are, it could be an indication that we inherited these qualities from our common ancestor.
In 2006, the first stages of the economic downturn helped to inspire the research into our darker sides.
“We were watching all of the people on Wall Street make really bad decisions, and we know really little about where this stuff is coming from,” she said. “There were bad loans, and bad credit card offers, but it seemed to me that people were making these errors in the face of negative evidence.”
Given these logical incongruities, Santos said that she hypothesized that our cognitive strategies might be further down the evolutionary chain than previously imagined.
In order to find the root of at least some of our more evil tendencies, CapLab researchers have trained the capuchins the basics of money.
But the capuchin conception of capital is a little different than our own. The monkeys use small tokens that they trade with researchers in exchange for food. By altering the price of food and changing the ways that trade occurs, the CapLab has developed a study on the way monkeys function in a trade economy.
Experiments can include training the monkeys to perform simple physical tasks in exchange for food or tokens, or simply testing if monkeys share our sentiments about the marginal utility of money, said Jenna Poggi ’13, a former assistant in the lab.
But it’s when times are tough, and an enterprising capuchin is short on tokens, that things get interesting.
“There is one monkey who will [try to trade] you its tail if it does not have anything else,” Poggi said.
Poggi worked in the lab last semester as part of PSYC 371: “Laboratory in Animal Cognition,” a course that combines working with the capuchins and class time. The course is open to all undergraduates regardless of major, but it is required if you want to work in the lab.
“I miss them a lot,” Poggi reflected on last semester’s work in the lab. “They each have their own personality.”
She added that there was one monkey who was clearly the “dumb one” and another who was particularly annoying, but she declined to disclose their names so as to not offend anyone.
THE RACE CARD
But although the capuchins may be sensitive to name-calling, they are not above bad behavior. Santos said that her lab’s newest venture is exploring whether monkeys can be prejudiced or create stereotypes.
This may sound like what we would call racism, but Santos explained that the scope of the project went beyond this narrow definition. Instead, the experiments were created to test if the monkeys naturally associate individual things with groups. She said the story began with a commonly held belief in cognitive science that humans cannot help but perceive difference across groups, whether that means “Yalie or Harvard student, or something physical like race.”
Santos said that this vein of research specifically tests if the capuchins can differentiate between humans and themselves, as well as if they create factions within their own group. The preliminary results are sobering: Researchers have shown that the monkeys demonstrate some of the same prejudicial predilections as humans.
“It shows that [negative] things like the origins of prejudice might be older than we previously thought,” Santos said.
These findings will be published in the February issue of Journal of Personality and Social Psychology.
Once we had completed our journey through the laboratories of those who use human behavior to advance our understanding of both our relatives and the technology we create, we were left with the sinking realization that both of these groups have the know-how to to one day supplant human hegemony. Enjoy the WEEKEND.