Artificial intelligence (AI) could one day be deployed to discriminate against home loan applicants, monitor computer users, and circulate counterfeits to distort reality on a massive scale. Or it could be “a stable platform that we can build on to reach for the stars,” the White House science chief told a Terps audience Wednesday.
Generative AI “is the most powerful and important technology of our time,” Arati Prabhakar, head of the White House Office of Science and Technology Policy, told a packed room at the Samuel Riggs IV Alumni Center. She called on the public to harness it for good – by improving drug design, providing better weather forecasts and bridging the education gap, for example.
Prabhakar was the featured guest at the latest Distinguished Presidential Forum, part of University of Maryland President Darryll J. Pines’ “Great Challenges of Our Time” course for freshmen, which presented a range of scientists, writers, businessmen and activists since Pines introduced it four years ago.
Prabhakar, who began his role in 2022, advised President Joe Biden on the impact of developments in cybersecurity, quantum science and many other high-tech topics on the United States and its relations with its allies and opponents.
During her speech, she also focused on two challenges, besides AI, that will impact society in the coming decades: health – ranging from curing deadly diseases to fighting pandemics through the guarantee of drinking water – and climate change. These topics are “at the heart of the University of Maryland community and all of the research, innovation and outreach that happens here on our Do Good campus,” Pines said as he welcomed Prabhakar to the stage.
An electrical engineer by training, Prabhakar previously headed the Defense Advanced Research Projects Agency (DARPA), the Pentagon branch responsible for sometimes far-flung technological research and development of futuristic weapons systems. (Pines himself was responsible for the DARPA program from 2003 to 2006.)
During a question-and-answer session informed by queries from faculty and students, Pines asked him to delve deeper into the ethical parameters of AI. He shared the anecdote of a colleague, a person of color, who was unable to open a bank account because a chatbot couldn’t process her passport. “These are the unintended consequences of AI,” Pines said.
(UMD launches institute focused on ethical AI development)
Faulty facial recognition, including law enforcement errors that result in unjust arrests, are “unbearable,” Prabhakar acknowledged. But she compared its misuse to beneficial applications, using TSA airport pre-screening as an example. This process is voluntary and restricted, and the images are immediately deleted.
“Now go through these lines,” she said. “We have achieved efficiency because we have the right technology. This is where we must go.