Tangelic Talks – Episode 04
What Impact Can AI Have on the Planet? Finding Fair Solutions with Dr. Shaloei Ren
8 minutes to read
In this episode of Tangelic Talks, we talk with Dr. Ren to unpack the complex relationship between artificial intelligence (AI) and climate change. Dr. Ren shared insights into how AI can be a powerful tool for advancing climate solutions, from optimizing renewable energy systems and improving energy efficiency to aiding in precision agriculture and climate modeling. However, we also explored the other side of the coin: the environmental costs of AI itself.
The Double-Edged Sword: AI’s Role in Sustainability
AI can both contribute to and mitigate climate change. On one hand, large foundational models significantly increase carbon emissions and energy consumption. On the other, smaller, specialized AI models are already providing solutions, such as:
Integrating Renewable Energy: AI helps optimize power grids by predicting energy demand and managing battery storage.
Air Pollution Forecasting: AI-powered models predict air quality and carbon intensity, allowing for proactive pollution control.
Optimizing Climate Models: AI speeds up climate simulations, making it easier to assess long-term climate risks and responses.
Dr. Ren believes that in the next 5-10 years, AI will become more energy-efficient, but in the short term, its overall impact on climate remains negative.
The Role of AI in the Global South
While AI benefits can reach communities worldwide, the environmental burdens are not equally distributed. Data centers are often located in low-income or marginalized areas, leading to local pollution and water stress. However, AI can support the Global South by providing:
Water Resource Management: AI-driven tools help allocate water in drought-prone regions.
Energy Planning: AI assists in designing solar and battery storage solutions for off-grid areas.
Policy Support: AI helps draft regulations and optimize resource distribution.
Dr. Ren emphasizes that ensuring equitable AI benefits requires deliberate planning and investment in sustainable infrastructure.
Looking Ahead: Can AI Be a Climate Ally?
Dr. Ren remains optimistic that AI will become more sustainable. He compares AI’s current energy inefficiency to early automobiles—while initial fuel efficiency was poor, technological advancements eventually improved it. Similarly, ongoing AI research focuses on increasing efficiency, reducing carbon footprints, and optimizing resource use.
Key strategies for making AI more climate-friendly include:
Smarter Data Center Placement: Locating data centers in regions with clean energy sources and lower water stress.
AI Workload Management: Shifting computing tasks to off-peak hours and more sustainable locations.
Standardized Reporting: Implementing uniform environmental impact metrics across AI companies.
Regulatory Collaboration: Governments and industry leaders must work together to create policies that balance innovation with sustainability.
Dr. Ren encourages AI developers, policymakers, and the public to take an active role in shaping AI’s future. “We need to design AI systems with sustainability in mind,” he says. “By making conscious choices today, we can ensure that AI serves as a tool for climate solutions rather than an environmental liability.”
Thought Provoking Q&A Session with Dr. Ren
AI faces fundamental challenges in achieving complete safety and fairness because it is trained on data distributions, optimizing for average or high-probability performance. Eliminating bias entirely is nearly impossible, though mitigation strategies exist. One approach is adjusting training loss functions to prioritize disadvantaged groups, giving them higher weight during training. However, when deployed in real-world scenarios, AI encounters unseen data, leading to errors or hallucinations. While researchers are actively developing techniques for better alignment and safety, fully eliminating these challenges remains difficult.
Ensuring high-quality, accurate, and representative data for climate-related AI applications is a significant challenge. High-quality data is essential for reliable AI models, but there is a growing concern that we are running out of new, high-quality datasets. Organizations like OpenAI are even exploring synthetic data to supplement training, but there are risks—poor-quality input can lead to degraded model performance, often summarized as "garbage in, garbage out."
However, climate modeling has an advantage over other AI applications, such as language or image generation. Unlike language, which lacks a clear governing framework, climate science is grounded in well-established physical laws. Decades of climate research provide a strong foundation for AI models, and integrating physics-informed machine learning can help improve accuracy. The key is to continue leveraging scientific knowledge, combining high-quality observational data with AI, and ensuring rigorous validation methods to maintain reliability in climate applications.
AI data centers have a significant environmental impact, and their location and workload management play a crucial role in mitigating harm. Historically, companies have built data centers near population hubs, but with improved bandwidth, location flexibility has increased. However, many still prioritize cost savings, tax incentives, and minimal regulation over sustainability, often placing data centers in areas where they disproportionately affect local communities and ecosystems.
Different regions have varying carbon intensities, water availability, and pollution levels. If these factors aren’t considered, data centers can exacerbate environmental harm—for example, excessive water use in arid regions like Arizona or increased air pollution in places like Pennsylvania and Ohio.
One way to address this is through better workload balancing. Even after a data center is built, AI workloads can be shifted in real time to locations with cleaner energy sources or more water-efficient infrastructure. This flexibility presents an opportunity to reduce AI’s environmental footprint.
Additionally, policymakers and businesses must develop clear, quantifiable metrics to measure AI’s environmental impact. While carbon emissions are a key factor, their direct impact is difficult to quantify. A more comprehensive approach—considering water usage, air pollution, and regional energy sources—can provide a clearer picture and guide sustainable AI deployment.
There’s a concept called the social cost of carbon, which essentially puts a price on the damage caused by carbon emissions. Under the Biden administration, this cost was set at around $60 per ton of carbon, and the EPA recently raised it to over $200 per ton. But during the Trump administration, it was as low as $3 to $5 per ton. That massive variation shows how much of a political number it really is—it’s not something we can fully rely on to drive meaningful change.
On the other hand, health costs associated with emissions are much easier to quantify. Take PM 2.5 pollution, for example—it has a direct, well-documented impact on human health. There’s over a century of atmospheric and public health research showing exactly how air pollution affects people, and we can measure that impact in real dollars. If we focus on reducing health impacts, we’ll also be reducing carbon emissions as a byproduct.
What we really need is a clear, tangible metric—something that makes climate action more practical and actionable. Relying solely on carbon emission weight is tricky because it doesn’t tell us how much reduction is enough. Should we just stop breathing? Obviously not. But just like when you buy a car, you see both the price and the benefits. Right now, we don’t have a real “price tag” for carbon emissions, and that makes it harder to incentivize meaningful action.
There are some interesting studies that suggest we should frame climate change as a public health crisis. Making the impacts more tangible could help people take the issue more seriously. In my own research, we’ve already quantified the effects of air pollution, which, like carbon emissions, comes from burning fossil fuels. If we focus on reducing the health impacts of pollution, we’ll also be cutting carbon emissions as a result.
The key difference is that health impacts have a clear metric. We can make science-based decisions instead of just saying, we need to cut carbon emissions without a concrete way to measure the benefits. People need to understand the trade-offs—if they’re paying more on their energy bill, what’s the direct benefit? The problem is that carbon emissions are measured in weight, while financial costs are in dollars. Converting between the two is difficult and often subjective.
I spoke with someone from HP at a climate change workshop who told me their framework helps reduce both energy costs and carbon emissions. But when I asked how they decide the price of carbon emissions, they said it depends entirely on their customers. If a customer values carbon reduction, they set a higher price. If they don’t, they lower it. That subjectivity makes it really difficult to establish a universal standard, and that’s where the private sector and politics start to play a huge role.
I’m not sure we should push for full regulation on environmental impacts just yet, although Europe is taking a stronger stance. They have an AI Act, which is much more comprehensive compared to the US. At this stage, I think we should focus on standardized reporting. For example, when companies report water consumption, there are different standards in place, and they don’t even use the same units—some use million liters, others use gallons. It’s an easy conversion, but it still creates unnecessary friction when comparing numbers.
Additionally, many companies don't report on air pollutants, which are harmful to people’s health. They might report on backup generators and carbon emissions, but they don't address emissions that directly affect health. We need a standardized framework for reporting environmental impacts more holistically. Once we have that, we can think about implementing some minimum regulations to ensure accountability. However, the goal of regulation should be to help the industry grow in a healthy and sustainable way, rather than to stifle it.
My personal take is that at this point, implementing a uniform, standardized reporting framework—one that’s more holistic—could be a step we can take without hindering the growth of the industry. It would allow us to keep companies in check and ensure they address their own environmental impacts first.
Looking at Europe, they already have many regulations in place for AI. However, when it comes to AI development, they’re a bit behind the US and China. That said, they’re ahead of both in terms of regulations, which I don’t think is a bad thing. AI is still a great tool with a lot of potential, but it does have environmental impacts, and it’s growing fast.
I think the right balance is somewhere in the middle—too restrictive on one side or too lenient on the other isn’t ideal. Finding that middle ground is probably the best way forward. But it’s complicated, and we’re still figuring it out.
Dr. Shaloei Ren
Associate Professor of Electrical and Computer Engineering at the University of California.

Dr. Shaloei Ren is an Associate Professor of Electrical and Computer Engineering at the University of California, Riverside. His research strives to build socially and environmentally responsible computing, with a broad emphasis on AI, sustainability, and more recently health-informed AI that minimizes the public health impact of AI and leverages AI to improve public health. His work has influenced AI policies adopted by many international organizations and governments such as the United Nations, UNESCO and WHO. He is a recipient of the NSF CAREER Award (2015) and several paper awards, including at ACM e-Energy (2024, 2016) and IEEE ICC (2016). He earned his Ph.D. from the University of California, Los Angeles.