On March 4th, 2025, Columbia University gathered experts from a wide range of fields to explore big questions in the rapidly evolving field of the first artificial intelligence Columbia AI Summit. Covering topics ranging from healthcare, business, policy, science, engineering and the humanities, the summit provided a 360-degree view on the transformational impact of AI on society.
Featuring researchers from climate schools, afternoon sessions, From Chaos to Code: How AI can tame the climate crisisdiscussed how AI is emerging as a powerful tool for climate science, disaster preparedness, and building resilience across interconnected systems. Read the highlights of the session or watch the video below.
speaker
Introduction Notes: David SanderrowInternational Public Service School
» View all Columbia AI Summit Videos
The power of AI
“There are various conversions [AI] Possible profits. What I’m particularly excited about is the innovation in materials… 150 years ago, when Thomas Edison invented a modern light bulb, he spent months and months, months and months physically testing different types of elements and materials, running electricity to see how much light and heat is produced. And he finally came up with what he thought was the best solution. Today, you can simulate these interactions in 1 second with the latest AI tools. And that ability gives us two advantages. One of these is that you can test materials that don’t actually exist. And if they appear to be beneficial in a particular application, we can manufacture them and see if they work. And you can make a quicker choice. There is a huge potential for AI to help accelerate the pace of energy innovation, and I think we need to know how to mobilize all sorts of tools to do that. ” –David Sanderrow
“What’s happening now is AI [weather prediction] Basically, a model that didn’t exist like it was five years ago would be just as good as it was, or definitely better than anything else. [traditional] Physics-based model. AI models explicitly know more or less about atmospheric physics. They don’t know about energy or momentum conservation. Like all other AI, they are trained on data. By giving them lots of historical weather data, and knowing what happened in the past, and then seeing what happened later, they find patterns in the mystical ways machine learning and AI do. And, like with other AI, training is very expensive, but as David said, the actual execution of the model in real time was very cheap, they became really, really powerful from anywhere, in a short time. ” –Adam Sobel
“I think equality is important. If you have problems with your power system in the past, and this is what happened in 2021 during the Winter Storm URI, then power operators will not be able to provide enough energy and need to reduce demand. During that time, I think Texas will shut down around 30% of the demand (because we lost many generations) and continue to run the rest of the system. But when you think of energy systems that are essentially branched network structures, the easiest way to reduce demand is to cut off branches. But usually or historically, it is a less fortunate community that lives in branched areas, and usually They have less infrastructure than they have modern infrastructure. On the other hand, new deployments of batteries and solar power have seen rich areas deploying microgrids. They need to rethink how to shut down demand in emergencies and take into account that some families and buildings can supply themselves through microgrids, but sometimes they are at a disadvantage to the community. In these cases, AI makes recommendations to system operators, makes quick decisions, and sometimes even simulates different scenarios. These are the areas that I think AI already shows a lot of promise.” –Bolun Xu
“As a behavioral ecologist, I’m very interested in how individual organisms react to environmental changes. And in order to start actual predictions, I think we need to understand variation at the individual level so that we can predict which groups and which ecosystems are trying to do in response to climate change and land use changes. And I think that’s here. [species identification] Technology is very useful. Images from camera traps and recording pods are used to move from individual species recognition to individual identification. You can use images to capture individual zebras. I think this was the first species because there are most barcodes like fingerprints. It provides a much better projection of the population, the ecosystem, and how organisms react to change. ” –Dustin Reubenstein
“This is a truly unique opportunity with the help of AI to establish a methodology to determine the optimal adaptation and resilient strategies of infrastructure to climate and climate-related hazards. This issue is very challenging, interdisciplinary, and involves contributions from engineering, physical sciences, social sciences and many other fields. And the only way to approach it is through the novel. Monte Carlo Approach including Stochastic optimization. AI helps to integrate all these different areas that contribute to the ultimate solution to the problem. ” –George Deodutis
The challenges of AI
“We all know that we predict hurricanes, especially as far as climate is concerned. It is very difficult for hurricanes to happen over the next five, ten, two decades, what is the frequency and the expected intensity. And so far, there are several models, stochastic models that explain this uncertainty. However, these models assume a stationary climate, which is an unchanged climate. And now, we know that the climate is clearly changing, so these stochastic models for extreme events will change in time. This makes the problem even more difficult. Therefore, we rely on an artificial intelligence approach to establish that we can quantify this evolutionary probability of the occurrence and intensity of these extreme events in the near future.”George Deodutis
“When you look at the food system, you see a lot of inequality. Small farmers, Indigenous peoples, and so many shepherds, ecosystem managers, our food system is often left to be better data and decision-making in this race. I think we need to be very careful, especially how we adopt these technologies. So this data democratization around AI will be very important.”Jessica Fanzo
“AI is moving so fast, so people may use things they don’t understand, and I think there’s a tension between public and private. There’s a movement from the White House to privatize the weather services that we’ve talked about for years, but now there’s a real threat that it’s going to happen… New [AI models] It comes from the private sector, but it still relies entirely on the public sector infrastructure of the underlying data and the physical model that does a lot of background work. So there’s a huge tension here and I think it’s the real danger of breaking things down with the infrastructure we rely on to keep people safe. ” –Adam Sobel
“I think about 30% of California’s daily production capacity comes from huge batteries now. And it’s fair to say that many of these batteries are currently being operated by AI. Of course, in human surveillance… [what the AI is doing]worries power system operators. It’s really about understanding transparency and how to adjust this AI, and I think it’s a real concern for many power system operators now. ” –Bolun Xu
“I know that models rely on the data we train. I think most people here are talking about one species of human being. What happens when you train a model of one species or one species, cut the corners, and use it to project it with other species? I think that’s a real concern. If the training data is from something different, how much can you refine the model so that you can work with other ecosystems and other species?” –Dustin Reubenstein
The Columbia AI Summit was held by Columbia AIa new initiative aimed at promoting Colombia’s work on artificial intelligence, including courses, curriculums, events, and digital tools. Columbia AI is Data Science Research Institute, Columbia Engineering,and Vice President of Research.
*Highlights are clearly edited
