Public Engagement & Impact
Advances in artificial intelligence (AI) have the potential to revolutionize things like transportation, healthcare, and communication. But it also has potential societal implications for employment and privacy.
Three U-M faculty at the leading edge of AI research joined the April 16 Wolverine Caucus to help make sense of this fast-moving field. The Caucus is a forum held in the state capital where alumni, policymakers, and the public can exchange ideas and hear from U-M faculty experts.
Professor John Laird said AI development has accelerated since 2011, thanks to a convergence of large amounts of available data, low-cost computers, and advancements in simulated neural nets.
Huge datasets and fast processing means computers can learn to correctly identify objects, and some of the practical uses have been applied to healthcare. For example, Laird said a skin cancer screening system recently developed did as well or better than human dermatologists.
“One of the things we’ll see is a possible revolution in healthcare,” he said.
But one of the challenges, he said, is that we don’t know why computers make decisions. Pictures and road signs can be obscured to make computers make the wrong choice, which raises questions about reliability.
“One of the things that’s scary about these systems is they don’t always make decisions the way we make decisions,” he said. “And we don’t always understand why they make these decisions.”
That’s why, he said, we need continued research on AI — both its technical workings and how it will impact jobs and and privacy. Laird said that’s why policy needs to be informed by research and why people need to have access to that information.
“We have to be informed citizens and make sure our desires are met in terms of how this is incorporated in our society,” he said. “We need people to be informed and active.”
Professor Rada Mihalcea said interest in AI is growing among graduate researchers, noting that the growth in applications “shows the excitement for this area.” Collaborations between the university and industry labs help connect the research to practical applications.
She detailed the AI work going on at U-M in machine learning, computer vision, robotics, language processing, and human-computer interactions.
The idea is to create AI systems that assist people in real time, such as high-quality language translation, word transcription, wheelchairs that can navigate on their own, and vehicles that can drive themselves in uncharted areas.
“As humans, we don’t just see things, we hear things and read things,” said Mihalcea, director of the U-M AI Lab and the Language and Information Technologies group. “Ideally we want these AI systems to behave in the same way.”
Professor Walter Lasecki, director of the Crowds+Machines (CROMA) Lab, said the near future likely will involve hybrid intelligence systems that combine human and machine learning.
“People are not going away just because AI is improving,” he said.
Systems in Amazon Alexa devices are useful in that they are voice-controlled, he said, but they’re not fully interactive.
“These are not systems that hold open-ended conversations about whatever you want to talk about today,” said Lasecki. To do that, “we still need human intelligence somewhere in the system.”
That’s why he’s thinking about how to build better human/AI teams, how to best use them, and how they would work in practice.
