Lori Pratt had the advantage of a strong role model at home: Her mother is a PhD who is very active in the women's movement and who has always supported her daughter's interest in science. Pratt has tried academia and industry, operating systems development, symbolic AI, and neural networks. She has always been interested in evolution and biological self-selection-ideas that have led her to study problem solving, learning, and questions about how new knowledge can be integrated into existing systems. She's particularly intrigued by the novel computational properties that emerge when massively parallel systems are used to solve problems. Pratt's most active current research explores the relationship between AI and neural networks - two fields that she believes have a lot to learn from each other.
Pratt discussed the insights she's gained from AI: ``Knowledge is power, and you learn better on the thresholds of what you know.'' In her doctoral research, she applied this idea to neural networks. She studied how a neural network that's been trained for one task can be mined for information to be used in a related problem. For instance, information from a neural network used for speech recognition in a general population might be used to help train a new network for a specific speaker. Or a medical diagnosis neural network built using information about patients in one geographical location might be used to help train a network for a different patient population.
Pratt's research has also explored more fundamental questions, such as how to obtain a neural network that uses a small amount of storage, how to visualize network learning dynamics, and how to construct large networks modularly.
She is just beginning a new job as an assistant professor in the math and computer science department at the Colorado School of Mines, where the computer science faculty women will outnumber the men. And two of the three women are in AI.
Suggestions: