How Do Computers Think?

It's Sarah Joseph's job to figure it out

October 6, 2021

To understand Sarah Joseph, a 31-year-old computer science researcher, first you have to understand the intersection of explainable artificial intelligence (AI) and adversarial machine learning.

That’s what she does for a living at the National Security Agency, and she’s very good at describing the highly complex profession in simple terms. She begins with explainable AI:

“Say we have a photo of a cat,” she says. “As a human, I see fur, pointy ears, whiskers, a nose, and I determine it’s a cat. A machine can look at the same photo and come to the same conclusion, but could be looking at it very differently, looking at textures and other things that we didn’t look at.”

It’s Sarah’s job to understand how the machine recognized that cat.

“Machines are trying to mimic human learning based on rational models,” she says. “When you ask it a question and it gives an answer, is the answer correct? How did the model come to that conclusion? How do I improve it?”

Next is adversarial machine learning, a closely related concept with a unique twist.

Instead of working toward sharpening the machine’s thought process, she says, she approaches adversarial machine learning with the goal of making the machine come to the wrong conclusion.  

In other words, her job is to get the machine to engage in the kind of logical fallacies that so often lead us humans astray. She is, in a very real sense, learning how to make the machine misbehave.

Sarah didn’t arrive at the intersection of these two fields after some focused and tenacious childhood buried in computer science books. In fact, computers were the last thing on her middle-school mind as she fed goats and horses on a friend’s family farm.

Taking a Chance

Sarah is one of a growing number of young professionals who were home-schooled. Unlike the public image of home-schoolers being closed inside a home, Sarah belonged to a co-op of home-schooled kids who took classes together taught by parents. The experience went beyond math and science.

“I’m not really an animal person, but I had friends and they had a goat farm, and I fed the goats. They also had horses. I was pretty afraid of all those creatures at first, but it was fun.”

Sarah took a basic programming class when she reached high school, and although she enjoyed it, it didn’t lead directly to computer science. At that age, she still had other things on her mind.

“I liked cooking, so my original plan was to be a chef,” she says. “I was working at a restaurant at Hersheypark and they put me in the kitchen, and I really enjoyed it.”

She soon realized, much to her dismay, that chefs work odd hours, mostly evenings and weekends, and that didn’t fit her vision for her life. She had to change direction.

Trying something different, it turns out, is a theme in Sarah’s life. If one thing doesn’t pan out, try another, and keep trying until you find something that clicks.

“Don’t feel like you have to have your entire career vision mapped out,” she says. “It’s OK to move around and try new things even if it appears a little different from what you originally set out to do.”

She thought back to the basic programming class in high school and decided to give it a shot in college.

That led to a bachelor’s degree in computer science and a master’s degree in human-computer interaction, both from Shippensburg University. During her time as a student she decided to apply for an internship at NSA.

She didn’t know anything about the secretive cyber agency at the time, and she wasn’t confident she would be hired. But she got the internship, and when she finished grad school, she was formally accepted to NSA’s Computer Science Development Program, a three-year rotational tour that immersed her in the NSA way.

“If it sounds interesting, don’t be afraid to try it,” she says. “Don’t sacrifice new opportunities for the comfort of the familiar.”

Seeing NSA for the First Time

Once inside, she learned that NSA is far from the isolated, inwardly focused agency that some imagine. In her relatively short time in the development program she’s had the opportunity to meet and work with foreign government allies in England, New Zealand and Australia.

“It was really cool that I got to talk to them every day,” she says. “It gave me a better understanding of how we’re all human, and especially how we can all help and protect each other.”