Share / Discuss

Deep Thinker

Making A.I. smarter through diversity

What do preventing deforestation in the Amazon, creating wearable devices for patients with Parkinson’s disease, and collecting the largest archive of Urdu texts have in common?

Deep learning—and data scientist Rachel Thomas ’05 is helping pave the way.

“It’s such a high-impact field,” she says. “There’s so much that’s possible.”

A subset of artificial intelligence, deep learning allows tools like Google Photos to organize huge libraries and Skype Translate to work in real time.

When Thomas first developed an interest in 2013, however, she found the field extremely exclusive—so much so that it led to the biases of homogeneous Silicon Valley developers being encoded in the tools themselves. Last year, FaceApp’s “Hot” filter was revealed to whiten people’s skin and make their features appear more Eurocentric, but other pervasive biases aren’t always as visible.

“Algorithms have been used to make hiring, firing, and parole decisions,” Thomas says. “It’s dangerous when they’re biased and not auditable.”

Driven to make deep learning accessible and inclusive, Thomas and her partner, Jeremy Howard, launched, a research lab working with the Data Institute at the University of San Francisco to train a new generation of developers. Last year, also established a diversity scholarship to help members of underrepresented communities learn the field.

As she’s gotten more people talking—and coding—Thomas was named one of Forbes’s “20 Incredible Women Advancing A.I. Research.”

“Being able to get deep learning into as many hands as possible will help solve problems that people working in tech may not even know about yet,” she says. “That’s what really excites me about’s mission.”