With AI data centers making their way into Louisiana, there are many concerns being raised by LSU students on the ethics of AI usage and its impact on the environment.
To discuss some of these issues, the Student Alliance for AI Reform, formerly known as the Student Alliance for AI Regulation, partnered with Geaux Green to host a talk titled “AI’s Environmental Impact in Louisiana: Where Are We Headed?” The talk was given by Supratik Mukhopadhyay, Ph.D., a professor and researcher at LSU, this past Wednesday in the Greek Theater.
Mukhopadhyay used the talk as an opportunity to show students AI is neither flawless nor unsalvageable. There are many applications for AI programs in the improvement of research, but there are also many harms that come with the use of AI.
“It’s important to be aware of AI and how to use it to your benefit, and how to not to abuse it,” Mukhopadhyay said. “AI today has an impact on the world.”
Throughout the course of the talk, Mukhopadhyay spoke about many of his research projects that utilized AI, emphasizing how these programs created major advancements for the field of environmental science.
One research project he spoke about was his contributions to an AI program that was trained to identify and classify certain formations into categories, such as trees, buildings, bodies of water and roads, from satellite images. The focus of this project was to map above ground biomass, with a focus on tree coverage, across the nation.
“This is basically related to how much carbon is getting captured in the ground,” Mukhopadhyay said. “It’s very important for environmental scientists, because people want to know how much carbon is there.”
This research provided valuable information to environmental scientists, as this then allowed them to more accurately determine the carbon consumption of plant life in America.
“We mapped the entire land cover using AI, using the satellite images of tree cover,” Mukhopadhyay said. “That today is part of what is called DeepSat. It is part of what is called the NASA Earth Exchange.”
This research improved the field of environmental science, since the AI program was not implemented until it was trained to near perfect accuracy. Before this program, scientists would have to piece together satellite images, which are incredibly vast with some formations appearing as meer pixels. It created a quicker but still accurate solution to a daunting task.
Mukhopadhyay also spoke about a research project he worked on to help with the prediction and detection of wildfires called DeepFire, working to prevent mass devastation for the communities in high risk areas.
“We built a prediction tool that can predict wildfires with more than 90% accuracy,” Mukhopadhyay said.
To predict wildfires, Mukhopadhyay, alongside other researchers, created an AI program that uses surrounding weather data, such as the type of vegetation in the area, wind speeds and storm conditions, to determine the chances of a wildfire occurring.
The program would then produce a prediction of how long until a possible wildfire would occur. This prediction could range from as little as a few days to over a month in advance.
Once a prediction is made, the program is designed to then focus on sensors and cameras in high risk areas to detect a wildfire beginning to spread. Once it detects a fire, it can predict how the fire will spread, giving first responders a better idea on how to tackle containing the wildfire.
The AI program was also designed to track lightning strikes in high risk areas, a major contributor to wildfire outbreaks.
“We made it so it can predict where the lightning strike is going to hit,” Mukhopadhyay said. “That’s one of the most powerful tools that predicts wildfires, but prediction itself is not enough.”
Mukhopadhyay reminded the audience that predictions are just that. There is no guarantee that the program will be correct 100% of the time, but it gives people in the community time to prepare and gather resources to combat wildfires and help prevent a major catastrophe.
By talking about these advancements through AI, Mukhopadhyay was able to show that the use of AI in research is not inherently a bad thing. AI programs specifically trained for research accomplish things faster than humans can, leading to quicker developments and more advanced research that was previously not possible due to time and resource constraints.
However, Mukhopadhyay also touched on the negative impacts of AI applications, specifically when it comes to generative AI programs.
One of these impacts he touched on was the “hallucinations” AI tends to have, where it incorrectly answers questions with confidence that convinces users it is correct information.
He gave an example where a hypothetical AI was trained to only identify giraffes and horses.
“If you give it a giraffe, it’ll say giraffe,” Mukhopadhyay explained. “If you give it a horse, it’ll say horse. It’s very good at that. It’s better than humans at that. We have already achieved a level where it’s better than humans. But now, what I do is I feed it an elephant.”
The introduction of the elephant seems harmless, but it leads to the program making a major error because it lacks the insight humans have.
“Normally, when a human sees the elephant, it’s like ‘oh, it’s not a giraffe or a horse,’” Mukhopadhyay said. “But, it’s different for AI. The AI has been trained on giraffes and horses, right? It doesn’t know anything other than giraffes and horses. So, it will see if the elephant is more similar to a giraffe or a horse, and depending on that, it will say it is a horse or giraffe.”
This mistake seems minor, but in other contexts such as cancer cell identification research, it can lead to major issues, especially when these mistakes go unnoticed.
“The problem with that is it does that without giving you a warning, so you don’t even know that it received something that it has never seen before and should not have received, like an elephant,” Mukhopadhyay explained. “And then it made a mistake, and it made that mistake silently.”
He also pointed out that the data centers used for many generative AI programs pull huge amounts of energy and water, taking away from crucial resources needed for survival.
“There’s so much heat getting emitted that unless you cool the system, the system can get damaged,” Mukhopadhyay said. “So you need huge amounts of fresh water, not salt water.”
These concerns are major for residents of Louisiana, as AI data centers are headed straight for the Boot.
“AI is coming to Louisiana — one of the world’s largest data centers is coming to Louisiana,” Mukhopadhyay said. “We need to be aware of how AI impacts our lives, especially in the environment, because if the environment is adversely affected, it will adversely affect our health.”
Mukhopadhyay’s talk revealed that the use of AI is not a one-sided argument. The argument for and against the use of AI is incredibly nuanced, with strong arguments on each side. On one hand, it is allowing research in a variety of fields to improve in quality and speed, but it is also consuming resources at high rates and, at times, spews out inaccurate information.
For SAFAR, this talk was not just a meeting. It was an opportunity to share the organization’s goals and present information to students about AI uses beyond the language learning models that students typically interact with.
SAFAR Event Planner Anderson Krupala, a freshman international studies and honors tract double major, explained SAFAR’s mission in hosting these talks and advocating for AI Reform on LSU’s campus.
“We know that AI is here to stay,” Krupala said. “Especially in the academic environment, we want to help facilitate that understanding between students and faculty and have it implemented in the proper ways in the school.”
Although Krupala doesn’t believe AI should be completely removed from the academic setting, he draws a hard line on when he believes AI should not be used.
“When AI starts making your decisions for you,” Krupala said. “When you don’t listen to your own intuition, like when you go to submit a paper and you ask AI if it’s okay to submit first rather than yourself, or you go to write a paper and you let AI fill in your thoughts for you.”
SAFAR President Jude Terrell, a junior political science major, expressed that he enjoyed this event because it showed attendees the purpose of SAFAR.
“I think that this event ties in with our central philosophy,” Terrell said. “First and foremost, we want to educate people on AI issues, that way, when they make their choices regarding AI, they have the most information possible at their disposal.”
For Ian Frick, president of Geaux Green and a junior coastal environmental science major, this talk broadened his perspective on the use of AI in research.
“I would say this event really shifted my perspective in allowing me to speak more clearly about how AI can be used in positive ways in research and modeling settings for environmental situations,” Frick said. “But it is still environmentally detrimental in other ways, specifically for that generative AI.”
Frick explained how this talk has better equipped him to talk with others about AI uses and its benefits and harms.
“I feel that I can now counter large claims that are made on both sides of the argument that ‘AI is necessary,’ ‘AI is not that bad,’ or people who say ‘there are no benefits to AI,’ ‘there’s no reason we should be using it,’” Frick said. “I feel much more informed to contribute to that conversation and correct people’s biases and misconceptions than I did before this event.”
SAFAR and Geaux Green are student organizations dedicated to making LSU a better place for all who visit its campus. Whether that be through efforts to protect the environment and nature of campus or the integrity of academics, these organizations help create the community we know at LSU today.
For those looking to attend talks such as this one in the future, you can check TigerLink under the Geaux Green or SAFAR student organizations for event listings, or each of the organizations respective instagram accounts @geauxgreenlsu and @safarlsu.
