The application of artificial intelligence has quickly entered various sectors, ranging from academia to social media to medicine. But as technology rapidly advances and adapts to multiple fields, research on the ethical and successful use of AI has also picked up pace. In the College of Letters & Science, researchers have used funding from the National Science Foundation (NSF) to investigate themes, including data privacy and large language models, computational AI and physical AI systems (i.e., robots). With the prospect of AI being used to house private information or working alongside humans, it’s necessary to understand and address the possible risks it poses. For these three L&S researchers, their work through NSF is not only figuring out how AI can be used to progress research, but also how it can be applied in society with care for the impact on people.
Is AI Trustworthy?
Artificial Intelligence is gradually being integrated into critical fields, including medicine, banking and finance. However, as AI and machine learning models are increasingly used to analyze and access large datasets that may contain private data or information about individuals, researchers at UW–Madison are focused on building trustworthy AI.
“When you go to your doctor or your bank, you build trust in them because you know they’re not going to share private data and are going to follow the right practices,” says Professor Somesh Jha, Lubar Professor of Computer Sciences. “So, think of trustworthy AI as basically the same principles, but now applied to AI, which is extremely important because AI is now being used in a lot of sensitive situations.”
Jha says that without trustworthy artificial intelligence, AI can have negative effects on several sectors of society.
AI in Particle Physics
While artificial intelligence has been the subject of research at UW–Madison, it has also aided research in other areas, such as particle physics. Lu Lu, an assistant professor in physics and researcher with the IceCube Neutrino Observatory, uses a neutrino detector buried under Antarctica to detect the most energetic particles. This information provides answers to some of the most pressing mysteries in particle physics, from exploding stars to black holes and gamma-ray bursts.
“Because particle physics follows universal laws, an AI that learns to predict particle interactions can generalize broadly,” Lu says. “Instead of doing billions of calculations per particle shower, which is super computationally heavy and time consuming, why can’t we just use AI?”
Robots in the Workforce
Robots are embodied artificial intelligence, as Bilge Mutlu — a professor of computer sciences, psychology and industrial engineering — puts it. He means that they are physical systems aided by an AI system to exist and navigate in the real world. And as AI becomes more advanced and integrated into the workforce, Mutlu is working with experts across fields, including computer sciences, economics, law, engineering and psychology, to mitigate potential consequences.
“How do we better inform the design of these technologies to consider the social, policy-related and economic factors in those workplaces?” Mutlu asks. “We might look at what the effects of the automation look like. Is it shifting jobs from companies to elsewhere? Or is it moving lower-skilled workers to be higher-skilled workers? Or replacing those workers?”