The sustainability paradox of AI and climate change

AI is facing a sustainability paradox.

Sure, AI can enable wildlife conservation and predict potentially devastating weather patterns better than humans. But the technology is also a significant contributor to climate change, producing immense amounts of heat and carbon. The process of developing the technology that some think will save the planet is actually a process that’s incredibly harmful to the planet.

Thus, a paradox. Artificial intelligence will need to solve the environmental problems its own production is contributing to one day. On its own, it can’t do so, and it’s nearly nonsensical to think that simply developing more AI is a solution.

What would help – and what already has – is applying AI to existing technologies with proven track records of success. AI can’t solve the issue of traffic on its own, but add it to traffic lights and you can reduce congestion by lowering the number of idling vehicles on the road. Artificial intelligence can’t remove litter from the ocean, but it can sift through photographs of the seas and detect debris for clean-up. When used in tandem with existing tools, AI’s potential and impact are almost limitless.

So, to make the future more sustainable, an integrated approach is best: continue to use known, successful technologies but enhance them with emerging ones like AI.

The problem with AI’s energy consumption

AI requires a colossal amount of energy to keep its systems running. It’s developed and run in data centers, which require large amounts of energy to keep their digital infrastructure up and running, produce a lot of heat in the process, and then require even more energy and water to run the climate control systems needed to cool them down.

“Data centers have been in a race for quite some time to reduce their environmental impact, particularly concerning energy and water consumption,” said Dassault Systèmes High Tech Solution Experience Director Jean-Marc Gaufrès. And, while improvements have been made, “this race will never end.”

Efforts are underway to make data centers more sustainable. Companies are just beginning to invest in renewable energy sources: Google recently announced they’re expanding their efforts in this area by partnering with a nuclear power start-up that will provide them with sustainable modular reactors to power their data centers. Despite efforts like these, the pace of AI development – and the draw of ever-increasing consumer adoption – often outstrips these advancements, presenting an ongoing challenge for the climate.

How does AI negatively impact climate change?

Beyond energy consumption, AI indirectly contributes to climate change in several ways. When it comes to data centers, the problem, according to Gaufrès, is three-fold.

●     The energy resources needed for data centers are still largely based on what’s readily available and plentiful in supply: fossil fuels.

●     The cooling processes in those data centers require both significant energy and water to prevent the necessary components from overheating.

●     Those electronic components can only be used for so long before they reach capacity and need to be replaced, so keeping the data centers running doesn’t just require a lot of energy and water; they create a significant amount of waste, too.

The quest for more data and better AI is driving the expansion of all these digital infrastructures, perpetuating the cycle of increased energy usage and environmental impact.

The accidental environmental output from innocuous individual input

As consumers, we’re often shielded from the consequences of seemingly innocuous interactions with generative AI tools like Microsoft’s CoPilot or OpenAI’s ChatGPT. We don’t necessarily know it, but our conversations with chatbots can be harmful to the environment.

Using AI requires significant computing power and energy no matter what the query is. So, asking those tools to tell you if two plus two actually equals four leads to significant energy consumption and expenditure, even though it’s a simple ask.

“We’re using generative AI for tasks they’re not meant for,” said Sasha Luccioni, an AI and climate researcher, in a recent Tech Won’t Save Us podcast. “For tasks, especially like question and answer…you don’t need to generate things. You need to extract things.”

Using AI inefficiently exacerbates its environmental impact and its effect on climate change, leading to unnecessary energy consumption and emissions.

Artificial intelligence and climate change: Making a positive environmental impact

While artificial intelligence can easily be blamed for its negative environmental consequences, it also has the potential to be a problem solver. It can revolutionize how we combat climate change and protect our environment. By analyzing vast datasets, AI can identify patterns and correlations that human eyes might miss.

In agriculture, AI can optimize irrigation schedules, reducing water waste and enhancing crop yields. AI-powered drones monitor and protect forests, detecting deforestation and illegal logging activities.

In the transportation sector, AI can analyze real-time traffic data to optimize signal timings, alleviate congestion and lower emissions from idling vehicles.

In the construction industry, AI algorithms can select materials and architectural layouts that minimize heat absorption and maximize natural light usage, leading to less reliance on artificial lighting and heating systems, which can contribute to urban heat island effect.

AI can also enable us as individuals to make sustainable choices, like with applications that can track a user’s environmental impact and provide tips for eco-friendly habits or through smart home systems that can adjust energy usage based on user behavior, lowering electricity bills and carbon footprints.

A technological tandem: AI & virtual twins

With all this potential, it’s clear that AI holds great promise for environmental sustainability, it’s still a developing technology. AI’s prospects for success can increase exponentially when paired with existing technologies with proven track records, like virtual twins.

On their own, digital twins – and in their newest evolution as virtual twins – have been enabling sustainable innovation for decades by enabling companies to virtualize their operations, from product design to material procurement to scalable manufacturing and cross-continental delivery. They’ve also helped companies prioritize sustainability as a key component of not only their business operations but their product output, too.

Virtual twins have been leveraged by thousands of companies across dozens of sectors to achieve climate-related success, particularly in the field of transportation. When hosted on the 3DEXPERIENCE platform, this kind of collaborative software, which often includes modeling and simulation programs, has enabled car and airplane manufacturers to reduce the weight of their vehicles and planes, thus also lowering fuel consumption. Even when progress is small – think a pound or two – its effect at scale produces notable results. 

Virtual twins, with their dependability and proven success, can be a necessary tool, then, in leveraging AI to its fullest potential.

Tackling climate change with machine learning: LLMaS

AI is quickly becoming a necessary component for businesses to succeed in any industry or sector. Developing it or purchasing AI tools outright, though, can be costly both in terms of capital and environmental impact. Earlier this year, Dassault Systèmes partnered with Mistral AI to launch a product that solves this conundrum: a large language model as a service, or LLMaS.

Providing an LLM as a service means that one company – not hundreds or thousands – needs to develop and train the model. That cuts down on the emissions released from data centers by reducing the need for them overall. Hosting the LLMaS on Dassault Systèmes’ sovereign 3DEXPERIENCE cloud platform makes it accessible to the company’s 350,000+ customers. Partnerships like these signal a greater commitment to sustainably developing and making AI available on a wide scale.

Will AI save the environment?

By itself, AI in its current format can’t reverse the climate crisis or save Earth. In and of itself, it’s problematic, but when its issues are addressed, it has the potential to improve the environment and better life for all of us.

“We’re in this kind of eternal balance,” said Dassault Systèmes North America Sustainability Director Stan Piper. “We need to decide if the problem I am solving with AI is offset or overwhelmed by the cost AI demands?”

The journey toward a more sustainable future hinges on recognizing the duality of AI. While this technology demonstrates remarkable potential to enhance our efforts in combating climate change—through innovations in agriculture, transportation, and construction—it is essential to acknowledge that its true power lies in synergy with existing, proven technologies.

By integrating AI into established tools like virtual twins, we can magnify the effectiveness of our sustainability initiatives, ensuring that we harness both the new and the old for a greener tomorrow. This collaborative approach not only addresses the pressing challenges of energy consumption and waste but also paves the way for a more resilient and environmentally friendly society. Embracing this integration will ultimately be key to unlocking AI’s full potential while mitigating its environmental drawbacks.