Friday, December 16, 2016

Mind over matter: Her interest in the brain led her to computer science

"Your brain can store all of your experiences at some level. It doesn’t store every detail," says Emory computer scientist Avani Wildani. "If we can get a better idea of how information storage and reductionism work in the brain, perhaps we could translate that to a computer framework." (Photos by Ann Borden, Emory Photo/Video)

By Carol Clark

Avani Wildani, an Emory assistant professor of Mathematics and Computer Science, got her first computer when she was three years old. A TI-99 from Texas Instruments, it was the first 16-bit personal computer and known for its eccentricities.

“It spoke to you in this very robotic voice,” Wildani says. “It was a really hot thing at the time for the computer to talk to you.”

Wildani was born in Mechanicsburg, Pennsylvania, to immigrants from the state of Gujarat in India. The oldest of three children in a close-knit family, she did not start speaking English until she began school.

Biology fascinated her from an early age, particularly mysteries of how the brain works. “I’m interested in how we think about things,” she says, “and what information means for the mind.”

One focus of her current research is the looming problem of information storage: The disconnect between our growing ability to gather digital information – on everything from human genetics to deep space – and our limited technology and resources for keeping and managing all of that data.

Wildani originally aspired to be a physician, like her father. “He steered me into a more sedentary direction,” she says.

Her father realized that she needed a job where she could sit because she was born with osteogenesis imperfecta, or brittle bone disease. The disorder causes bones to break easily and Wildani endured hundreds of breaks and countless medical procedures as a child.

In September, she was hit by a car while crossing an Atlanta pedestrian crosswalk in her wheelchair, an incident that fractured her legs in seven places. She continued to work while recovering in her Druid Hills home.

Light streams into the modernistic home through large, plate-glass windows, which look out onto wooded terrain. Evidence of Wildani’s many hobbies is strewn about the house – a full set of drums, a telescope that she built from scratch, her pencil sketches, photos of her kayaking adventures. As she settles in for an interview in her home office, she offers a visitor a tray of Indian pastry sweets.

Below are some excerpts from the interview.

Q. What was it like to spend so much time in hospitals as a child?

Wildani: It wasn’t that bad. It would take me about 20 to 30 minutes a day to do my school homework, and I could still keep up. So the rest of the time I got to lie around, eat ice cream, and read. In fourth grade, I got through all of Nancy Drew.

It was probably harder on my parents. As a child, I just took it for granted that my parents would be at the hospital at 3 am with Indian food because I didn’t like American food. My mother would make me treats like khandvi – chickpea flour mixed with lemon and yogurt, rolled out and fried. It’s super good! Gujartis are like the Southerners of India: We fry everything.

Playing computer games "allows you to try something and fail in a safe way," says Wildani, who also plays the drums.

Q. Did you also spend a lot of time on a computer?

Wildani: Yes, I started learning about programming as soon as I learned to read, at age 3. I used LOGO, an educational programming language for children. It’s all about drawing pictures. In fact, the first time I taught Python to college students I had them integrate it with LOGO to make flowers for their mothers on Mother’s Day. People really loved that assignment and ran with it. They did all sorts of amazing animations.

Growing up, I also enjoyed computer games. I played a lot of King’s Quest, an early strategy game designed by Roberta Williams, who is one of my inspirations. King’s Quest was full of puzzles and the puzzles were hard. It teaches perseverance.

Q. So you don’t consider computer games a waste of time? 

Wildani: No, not at all. They inspire a lot of people. I would say that everybody I knew during my college years played a lot of computer games as a kid. Even if you don’t go into computer science or technology for a career, computer games allow you to try something and fail in a safe way. You don’t die and you keep getting better the more you try. It also teaches team building. I think playing computer games is in some ways like playing a sport. I would be interested in studies that look at how the motor cortex changes if you play a lot of computer games.

Q. Maybe all those computer games you played account for the can-do spirit you embody in your real life. 

Wildani: I’m not afraid to try things. I fail sometimes, but I have a really supportive family. I have a pretty solid mat to fall on when I screw up.

Q. How does your interest in the human brain relate to your research into computer data storage? 

Wildani: Your brain can store all of your experiences at some level. It doesn’t store every detail. In fact, it forgets things on purpose, pruning data and keeping what you need to make sense of the world, stringing together related ideas and forming memories that it can then retrieve. One theory is that the hippocampus acts like a kind of card catalogue where the brain can look up the place where it has stored different pieces of data.

The brain does all this using very little power, only around 60 watts, and with a high fault rate. Neurons are dying all the time and synapses are breaking down. In fact, stroke patients can lose a significant part of their hippocampus and still function.

If we can get a better idea of how information storage and reductionism work in the brain, perhaps we could translate that to a computer framework. We’re using vision classification as a bio-model to run simulations and try to figure out how the brain makes tradeoffs between network size, power and reliability.

"We’re losing information about hidden things that we don’t already know to look for," Wildani says. "That’s why it’s critical now to come up with new ways to store large volumes of data."

Q. Why is it important to understand this? 

Wildani: People are creating entire digital identities of family photos, emails, receipts – all kinds of data. Some of it is data that people want to keep and some of it is just stuff that gets saved unintentionally. You can think of all our personal collections of data like junk-filled closets that we store in “clouds.”

But what many people don’t realize is that the clouds are actually physical computer-data centers. And they require space, electricity and people to maintain them and to run them.

The centers house the physical storage devices for all the data. Disks, for instance, save data using tiny dots, called bits. But there is a limit to how close these dots can be, based on physics. They can only go to a certain resolution and no lower. Flash drives don’t rely on that sort of density, but they are not as reliable. And a disk only lasts about five years. So you have to have a plan to move data around between different devices.

We make a lot more information than we have anywhere to put it. Right now, Internet traffic is annually generating zetabytes of information and that’s projected to go up to yottabytes by 2020. [A yottabyte equals 1,000,000,000,000,000,000,000,000 bytes.]

We have this huge mess of data collected across time. How do we decide what’s important and what’s not? How long should we keep it? Who will pay for keeping it over time? There are things we archive with intention, like family photos, and things that we archive without really meaning to, like random receipts and emails.

Q. Could it just be a matter of everyone throwing out their digital “junk” and making space? 

Wildani: The issue is much bigger than that. We can’t always know what is going to be important in the future. The tapes of the early “Dr. Who” episodes, for example, got written over years ago because the BBC didn’t think they were important.

In the future, if you get sick with something, it might be useful to pull up the medical records of your great grandparents. Or maybe not. You might also want to store a genome analysis of your family for the future.

My favorite example of “junk” data that became useful is a photograph that was taken in 1900 of a random patch of sky. Decades later, cosmologists found this photo in a drawer and saw that it contained a pre-supernova. By comparing it to modern-day photos of the same patch of sky, they were able to compute a more exact rate for the expansion of the universe.

We have much more powerful telescopes today but not enough storage to save all the data these instruments are capturing. The same is the case for CERN, home of the Large Hadron Collider, which is gathering large datasets on sub-atomic particles. CERN is only able to save the data that is being experimented on now.

We’re losing information about hidden things that we don’t already know to look for. That’s why it’s critical now to come up with new ways to store large volumes of data.

Q. How do you feel about the scarcity of women in computer science? 

Wildani: When I was an undergraduate, I would have felt more comfortable if there had been more women in the room when I was asking a basic, technical question. But I had no choice. I was the only female in most of my computer science classes. And the default assumption for women is that you’re not competent. You have to prove yourself.

The traditional image of a computer science major is a guy with a beard who drinks a lot of Mountain Dew and spends a lot of time ranting on online forums. There are definitely some people who fit that personality type, but also many people who don’t. And some of them are so far from it that they don’t feel welcome and they leave.

I knew a woman in graduate school who was great at computer science, great at math, but she was also tall and blonde. She got tired of dealing with harassment and people who thought she was the receptionist. Eventually, she decided to go into a different field.

And yet, we need more people and different perspectives in computer science. The salaries for computer engineering are high because there are not enough people to fill the jobs.

I developed a web site (Project Hypatia) to highlight important research by women in computer science. I think it’s important to make it clear that women are not just asking for equality, but leading teams, doing impressive science and getting results.

BRAIN grant to fund study of how the mind learns

No comments:

Post a Comment