At U-M’s Climate + AI Panel, experts explore how artificial intelligence is reshaping the fight against climate change

Climate Week blog posts were written by students from the University of Michigan. They have not been edited by the U-M Communications team.

The intersection of artificial intelligence and climate science took center stage during the University of Michigan’s first Climate Week at the Climate + AI Panel on October 1, 2025. The discussion was held in the Johnson Rooms of the Lurie Building and presented by Climate and Space Sciences and Engineering. Faculty experts and researchers came together to reflect on a crucial question: whether artificial intelligence is helping or hindering humanity’s progress toward climate solutions.
The event was moderated by Jeremy Bassis, Professor of Climate Space and Engineering, and Adriana Raudzens Bailey, also an Assistant Professor of Climate Space and Engineering who opened the session by encouraging a dialogue rather than debate. The panel featured Mosharaf Chowdhury, Associate Professor of Computer Science and Engineering; Hansi Singh, Co-founder and Chief Executive Officer of Planette; Mohammed Ombadi, Assistant Professor of Climate and Space Sciences and Engineering; Rabab Haider, Assistant Professor of Civil and Environmental Engineering; and Jing Liu, Executive Director of the Michigan Institute for Data and AI in Society.

Each panelist offered a unique perspective on how AI is transforming their field. Their work spans topics such as power grid management, extreme weather forecasting, sustainable data infrastructure, and the ethics of energy use in large-scale computing. Together, their insights revealed the many dimensions of what happens when climate research meets computational intelligence.

When asked how they use AI professionally, the panelists spoke about both the opportunities and the boundaries of the technology. Jing Liu described her early research in neuroscience and how she now leads university-wide efforts to integrate AI across research disciplines. She explained that she uses AI tools to organize complex information, such as designing training programs for faculty who teach data science.

Mohammed Ombadi reminded the audience that AI in climate studies is not new. He began using machine learning in precipitation forecasting during his doctoral studies in the 1990s. What has changed, he said, is the scale and speed of discovery. “We can now integrate machine learning with science questions, not just use it for predictions,” he noted.

Hansi Singh explained that her work focuses on translating environmental data into useful forecasts for people and organizations. “We train on physics-based models to make long-range predictions,” she said. “Once trained, these models are incredibly efficient.” Rabab Haider added that AI is accelerating research on energy systems and grid management, yet she emphasized that human expertise remains central. Transformation, she said, lies in speed and scale rather than in replacing scientific reasoning.

A major theme of the discussion was the energy footprint of artificial intelligence. Mosharaf Chowdhury, who leads Michigan’s Machine Learning Energy Initiative, explained that the only sustainable way for AI to grow is through efficiency. He questioned the current obsession with building ever-larger models and urged a shift toward smaller, purpose-built systems.

Jing Liu added that the research community must reflect on how much energy is truly needed to train these models. “At what point do we realize that smaller and more efficient models are enough?” she asked. Haider connected the issue to global equity, warning that data centers are often built in lower-income communities, repeating patterns from the industrial era. Ombadi agreed and highlighted that the water footprint and community impact of such centers must also be considered.
The panelists also discussed the importance of data quality. Chowdhury emphasized that the performance of AI models depends on the accuracy of the data they are trained on. He pointed to distributed learning approaches that keep data secure and local. Singh noted that many regions in the Global South still lack basic weather observation infrastructure. “If we want climate AI to be equitable,” she said, “we must invest in building systems and training scientists across the world.”
Ombadi described the potential of physics-informed machine learning, which integrates scientific laws into algorithms to address data limitations. The discussion made clear that data is the hidden currency of climate research and that equitable access to data is a moral and scientific necessity.
When asked whether AI is hype or transformation, the panelists agreed that it is both. Singh remarked that AI can be climate’s most powerful tool, but only if applied to problems that matter rather than to convenience. Chowdhury added that efficiency, transparency, and regulation must guide the next phase of AI innovation.
The session closed with an audience vote on whether AI’s rise is harmful, neutral, or beneficial. Most participants agreed that its overall impact is neutral or even slightly negative, given its energy demands, ethical concerns, and uneven global access. Yet the discussion also carried a sense of cautious optimism that, with intentional design and strong public oversight, AI could still be steered toward the public good.
As Climate Week continued, the panel left participants reflecting on one core message: achieving our climate goals depends not only on reducing emissions, but also on ensuring that our digital innovations do not deepen existing inequalities or environmental burdens. The future of climate and AI, the panelists agreed, will depend less on what technology can do and more on how responsibly we choose to use it.