During the last holiday season, millions of people worldwide experienced what it was like to let ChatGPT write an essay, debug code, or dispense sage advice. They also witnessed firsthand just how smoothly ChatGPT could make up information that it didn’t know.
Artificial Intelligence (AI) is rapidly claiming its place in everything we do: it drives cars and fighter planes; it detects cancer in radiological images and black holes in telescope images; it helps us choose which songs to listen to; it helps government offices decide who gets welfare; it even designs and runs experiments, even faster than scientists are able to.
As a campus-wide organization to support data science and AI research, one focus area of the Michigan Institute for Data Science (MIDAS) is to enable the use of data science and AI to accelerate scientific discovery. During the Winter 2023 semester, MIDAS faculty affiliates, postdocs, and campus collaborators organized colloquia on “Automated Research Workflows”, “Data Justice, AI and Design” (jointly with Taubman College and the Center for Ethics, Society and Computing), and “AI in Science and Engineering.” The fourth and final one of this semester, “Implementing AI in Health,” will take place on April 17 as a joint event of MIDAS and the Department of Learning Health Sciences. These events brought together experts from around the country and U-M faculty to stimulate research ideas and collaboration.
Yet, this rapid advancement of AI is also challenging some of our fundamental beliefs. What is unique about being human, when machines can write novels, compose music, and beat the chess world champion? To address these complicated questions, MIDAS, LSA and the AI Lab organized a panel discussion in February on how AI tools challenge traditional thinking about classroom learning. On April 19, the second installment, hosted jointly with the School of Music, Theatre and Dance, focused on generative AI, music composition, the meaning of creativity, and the challenges AI poses to intellectual property.
The fascination about AI also coexists with uneasiness. The increasingly complex AI tools, sometimes in combination with flawed data, can give rise to harm in many ways, including the magnification of disinformation or surveillance capitalism, that will perpetuate or amplify existing biases and injustice. The concerns about the ethical and responsible use of AI are also not limited to human research. Ecologists need to consider whether their datasets are inclusive of endangered species. Jet engine designers need to know that their AI-engineered solutions are safe for humans. How can academic researchers address these concerns as AI developers and users? Another focus area of MIDAS is exactly to promote responsible data science and AI.
The 2023 Future Leaders Summit recently took place, with the theme of “Responsible data science and AI.” This MIDAS annual event brings together outstanding PhD students and postdocs from major research universities, midwest universities, and minority-serving universities, with the majority of the attendees being women and under-represented minorities. Trainees listened to vision talks from mentors, gave research presentations, attended career mentoring sessions, and networked with peers. The focus on responsible data science and AI helps to prepare these trainees, who will be the next generation of academic leaders in data science and AI, to maximize the positive impact of data and AI, while preserving the ideals and values of our society.
The Summit included a mini-symposium with the following speakers:
- H. V. Jagadish, University of Michigan, “Equity in data science.”
- Ellie Sakhaee, Microsoft, “Building a culture of Responsible AI (and what it means for researchers)”
- Tanya Berger-Wolf, the Ohio State University, “Human-machine partnership for conservation: AI and humans combatting extinction together”
- Andrew J. Connolly, University of Washington, “From interstellar rocks to dark energy: building data science across research communities.”
On May 16, 2023, MIDAS hosted From Theory to Practice: Building Ethical and Trustworthy AI, a forum jointly organized with Rocket Companies featuring speakers from academia, industry and government presenting on the ethical use of AI and its regulation.
“Humans constantly grapple with how new technological advances fit into our moral and ethical framework. But the massive amount of data in the world today and how quickly AI is getting more powerful makes this the critical moment for ethical data science and AI.” Says Dr. H. V. Jagadish, MIDAS Director. “We look forward to working with all researchers to tackle this challenge together.”