Overview
- The University of Michigan announced in September 2023 that it would provide every student, faculty member, and staff with access to a generative AI assistant called "UM GPT." This AI assistant was designed to be a customized tool based on large language models like OpenAI's GPT-4, specifically fine-tuned for the university's needs. The initiative is part of Michigan’s broader effort to integrate AI into its academic and administrative infrastructure, enhancing learning, teaching, and research. (Sources: https://its.umich.edu/computing/ai/gpt-in-depth; https://news.umich.edu/u-m-debuts-generative-ai-services-for-campus/)
- The University of Michigan's decision to provide each student with an AI assistant is a forward-thinking move, aimed at enhancing the learning experience, providing personalized support, and improving student productivity. These AI tools can assist with everything from academic guidance and tutoring to mental health support, which is expected to significantly improve educational outcomes and student well-being.
- The University of Michigan (UM) is one of the largest energy consumers in the state of Michigan, with its energy usage comparable to that of a small city. The University of Michigan consumes about 2.3 trillion BTUs of energy annually across its Ann Arbor campus. This includes electricity, heating, cooling, and fuel used in campus buildings, research labs, and other facilities. For comparison, the city of Ann Arbor consumes approximately 4.6 trillion BTUs of energy annually, including residential, commercial, and industrial usage for 125,000 residents. Sources: https://ocs.umich.edu/resources/sustainability-data/building-energy-data/; https://css.umich.edu/publications/research-publications/css-factsheets-u-m-energy-consumption-and-conservation)
- AI systems, especially large-scale implementations, require substantial computational power. These systems depend on energy-intensive data centers, both on and off campus, to run complex algorithms and models in real-time. The energy consumption associated with these systems includes not just the operation of AI assistants, but also data storage, processing, and continuous learning models, which can put a strain on institutional energy resources.
- The University of Michigan has ambitious sustainability goals, including carbon neutrality by 2040. However, the energy-intensive nature of AI could complicate efforts to reduce the university’s overall carbon footprint. With AI usage growing exponentially, there’s a need to balance educational innovation with environmental impact. The increased energy demand from AI services may require additional infrastructure, which could increase the university's reliance on non-renewable energy sources. (Source: https://record.umich.edu/articles/u-m-publishes-second-annual-climate-action-report/)
Transcription
David Fair: This is 89 one WEMU. And for all of the various kinds of questions and conversations that artificial intelligence poses, there is one that perhaps is under-discussed, and that's the environmental impacts of A.I. I'm David Fair, and welcome to this week's edition of Issues of the Environment. The University of Michigan is among those leading the way in generative A.I. In 2023, the school announced it would provide every student, faculty member and staff member with access to an AI assistant called UM-GPT. It is most certainly a forward thinking initiative. It is also a program that will consume an abundance of energy with the potential to impact the U of M sustainability goals. Dr. Ravi Pendse and his colleagues are well-aware of that and working proactively to address the issues. Dr. Pendse is Vice President for Information Technology and Chief Information Officer at the U of M and our guest on Issues of the Environment. Thank you for carving out time today! We appreciate it!
Dr. Ravi Pendse: Thank you, David! It's a privilege and honor to be with you on your program!
David Fair: With all of the concerns that do surround the use of A.I., were environmental impacts among the worries before the roll out of UM-GPT?
Dr. Ravi Pendse: Absolutely, David. We considered a lot of different issues. But one of the things I want to point out is that, in my humble opinion, generative A.I. has the potential to become one of the most impactful technologies of this century and, frankly, a force for positive disruption if used responsibly, if used ethically, if used heartfully.
David Fair: Those are big ifs!
Dr. Ravi Pendse: I know. Absolutely. And who better than Michigan, as I say, in dealing with these big ifs. And one of the key issues when we think about responsible use of this technology is consideration for the environmental impact. And we definitely pay attention to it. We take that into account. And that's why, as we explored this rollout, we chose to partner with organizations that had clarity around environmental goals when it relates to data center use, the high performance computing use, that is very much necessary for implementation.
David Fair: Well, the U of M says its campus in Ann Arbor is already one of the biggest five consumers of energy in the state. Annually, it uses about half what the entire city of Ann Arbor uses. That includes residential, commercial and industrial energy consumption. A.I. is, as we know now reliant on huge data centers, which, in turn, is reliant on a vast amount of energy. So, have you determined if this initiative is going to add to the U-M's energy draw in the city of Ann Arbor?
Dr. Ravi Pendse: Since our environment is actually based in a cloud with one of the vendors, it is not going to directly impact our data centers or city of Ann Arbor per se, but it will definitely be counted as part of our scope three emission analysis. So, no direct impact per se in Ann Arbor or increasing the energy consumption of our data centers that currently reside at University of Michigan. Because all of this environment is set up in cloud with one of the vendors.
David Fair: Issues of the Environment and our conversation with Dr. Ravi Pendse continues on 89 one WEMU. Dr. Pendse is Vice President for Information Technology and Chief Information Officer at the University of Michigan. The University of Michigan has some rather ambitious sustainability goals. So, how was that priority included in the planning and implementation of the UM-GPT program?
Dr. Ravi Pendse: The way we looked at it is, when the question of generative A.I. implementation at Michigan came to the forefront, we really had two choices. One was not do anything about it and let it sit back and watch what the world does because the technology is here to stay. And it is something, in my opinion, our students and our faculty and staff should have opportunities to leverage, use and learn from. So, as we had discussions, via a large faculty committee that included also staff and students, there were many conversations in those discussions around various issues, including, and I talked about responsible use and there is concern and bias and is concerned on what is called a hallucination. There is, of course, concern around environment and sustainability. So, all those factors were discussed, and a calculated decision was made that it is the right thing to leverage and engage this technology and be part of the ecosystem, where we have these conversations openly and influence the decisions not only that occurred at Michigan, but beyond Michigan: working with the vendor partners, working with government agencies, having these conversations. These are important issues, and sustainability is very important to Michigan. And, personally, it's very important to me. So, we always keep it at the forefront as we think about not just this technology. But any technology that you roll out in the end, you have to look at the complete piece of pie. We cannot just focus just on sustainability, just on bias, just on A.I. All of these conversations have to be on the table together.
David Fair: As we become more reliant on A.I. and it becomes more integrated into all aspects of campus life and our overall lives, the initiative will continue to expand. So, as you were working to create this, implement, and roll it out, was the infrastructure done in such a way that it will be able to accommodate the technological advances to come in the next few years and beyond?
Dr. Ravi Pendse: Absolutely! I think that's a great question. So, any time we do any kind of rollout, that is enterprise-wide, meaning university-wide, we always take into account the scalability aspect. But how can it scale? Does our environment scale? And as more applications to your point come on board, is everything that we are built able to handle the number of users and use cases? So, the answer is yes. And even in those conversations, along with the environment, we paid attention to a variety of different issues.
David Fair: You mentioned the inclusive nature of how this program was put together and rolled out. Are these conversations going to continue to be as inclusive as the program grows and the use of generative A.I. continues to expand?
Dr. Ravi Pendse: They absolutely have to be. And I'm very proud to say that University of Michigan has some incredible faculty members, great staff members. And our students are absolutely brilliant. And they keep us on our toes by asking these very important questions. So, it's always good to have these conversations ongoing. I jokingly always say, "I do not want five engineers sitting around the table making all decisions related to A.I., since I'm one of the engineers." I'd rather have diversity at the table. So, I want that colleague from the philosophy department at the table. I want that student from the English department of the People. So, it is our community that makes us stronger and makes us accountable, and I welcome it!
David Fair: Once again, we're talking with the vice president for information technology and chief information officer at the University of Michigan on 89 one WEMU. Dr. Ravi Pendse is our guest on Issues of the Environment. Well, you mentioned partners. The University of Michigan chose Microsoft as one of its partners for the UM-GPT Initiative, in part because of its corporate commitment to becoming zero waste and carbon negative by 2030. Now, I'm not asking you to speak for Microsoft in any way, but you are deeply ingrained in this project and understand how difficult those goals and ambitions are to achieve. Will the U of M go as Microsoft goes when it comes to sustainability?
Dr. Ravi Pendse: I think that's a great question. But, again, the way we have built our platform is built in such a way that it's flexible. So, let's just assume for argument's sake, and I'm not speaking for Microsoft, I generally have confidence that they will meet their goals because, as I recall correctly, they have been carbon neutral since like 2010 or 2012. I don't remember the exact date. So, my expectation is that they'll meet goals. But let's assume, for argument's sake, that, somewhere along the way, our paths diverge. And we don't necessarily agree that we have the ability to simply, for lack of a better word, disconnect or discontinue that relationship and bring on a different partner that perhaps is in more alignment with our sustainability goals. I believe that our partnership will continue to grow, and good things will come out of it, as far as environment is concerned. But we have been always very thoughtful on how we have designed this entire platform to make it essentially independent of whichever vendor partners we are working with.
David Fair: Well, the world of politics seems to touch most aspects of our lives one way or another. This issue is certainly not exempt. The Michigan House last month passed a measure that would exempt large data centers from equipment use taxes through the year 2050 in order to attract more industry and jobs to the state. You've mentioned that you U-Ms program is going to be conducted in the cloud. But environmental advocates remain unconvinced that the tradeoff is worth it, saying it could make the state's sustainability and climate goals unachievable. How or would that legislation impact your initiative if it were to become law?
Dr. Ravi Pendse: Obviously, I'm not a politician, and I'm happy I'm not a politician, frankly speaking. I like being an engineer. And so, while we, of course, will follow whatever the state line would be, if it becomes a law. But my expectation is the A.I. technology--the generative technology--is here to stay. And this is just my opinion. But I do feel that any person who's going to college anywhere in the country should have some exposure, some insight, some experience with these A.I. tools because we are going to need a lot of A.I. literate people who understand A.I--graduates--to support our country, so that we continue to maintain that advantage we have across the world in innovation and entrepreneurship. So, that's something that's going to be needed. And what we need to do is stay focused on our core values. And, at Michigan, our core values are about equity, about inclusivity, about supporting our sustainability goals. And I'm very confident that working with the entire Michigan team will be able to stay true to our core values.
David Fair: Well, it's going to be interesting to watch as it grows and evolves, and that will give us an opportunity to have another conversation in the future. But thank you so much for the time today! I appreciate it!
Dr. Ravi Pendse: Thank you, David! And I welcome the opportunity to visit with you again!
David Fair: That is Dr. Ravi Pendse. He is Vice President for Information technology and Chief Information Officer at the University of Michigan and our guest on Issues of the Environment. For more information on our discussion and the UM-GPT initiative, pay a visit to our website when you get a moment, and we'll get you everywhere you need to go. Issues of the Environment is produced in partnership with the office of the Washtenaw County Water Resources Commissioner. And you hear it every Wednesday. I'm David Fair, and this is your community NPR station, 89 one WEMU FM, Ypsilanti.
Non-commercial, fact based reporting is made possible by your financial support. Make your donation to WEMU today to keep your community NPR station thriving.
Like 89.1 WEMU on Facebook and follow us on Twitter
Contact WEMU News at 734.487.3363 or email us at studio@wemu.org