AI Rethink: ChatGPT is Thirsty
The travel industry—and corporate travel programs—are eyeing the upsides of AI-powered sustainability innovation, but what about its intensive energy and water use?
AI technology could be the key to unlocking clearer pathways to meeting sustainability goals, yet it could also unlock a growing source of environmental harm.
Regardless of which door it opens—or both—AI technology will transform and dominate the global economy in the coming years. PwC research projects that by 2030, AI will be contributing up to $15.7 trillion toward the global economy, which is more than the economies of India and China are putting out combined. Broadly speaking, AI used specifically for environmental applications could boost global gross domestic product by as much as 4.4 percent while contributing to a decrease in global greenhouse emissions between 1.5 and 4 percent by 2030, according to PwC analysis.
Corporate travel stands to be a part of those reductions.
The Lady?
Indeed, AI and machine learning technology already is proving its value in corporate travel sustainability in terms of data management, said Olivier Benoit, global SVP at BCD Travel's consulting arm Advito.
"If you want to decarbonize a program, you have to measure things, but measuring has always been a challenge," Benoit said. "Machine learning provides us an easy way to implement a solution and be really good at consolidating data sources."
This is particularly true on the hotel side, where data can come in a wide variety of formats and from a wide variety of sources. Anna Feinberg, CEO and co-founder of Alō Index and former BNY Mellon global travel manager, said her company has built a model internally that can help to verify data that comes in including 10K reports, receipts or photography of landscaping. "Those types of things can help to build a complete and verifiable data set," she said.
Advito is using a machine-learning solution for hotel matching, which has enabled it to reach a matching rate of 96 percent in a matter of hours, compared with the manual rate of 80 percent over a couple of days, Benoit said.
AI will not only assist in the gathering of data for sustainability but also in its presentation and use.
"The biggest challenge has always been that we can speak in numbers, but it doesn't always speak in facts," said Cobb & Hutch Consulting's Karen Hutchings, who previously was global head of travel, meetings and events at EY. "We may say that something is X number of kilos of carbon, but what does that kilo of carbon actually mean? What AI gives you is access to much more understandable data."
As data can help build sustainability scenarios, like showing the potential impact of policy changes and the best pathways to reducing the environmental impact, AI will "greatly accelerate those opportunities," Clarasight CEO Adam Braun said. It will give managers a chance to "chat" with their data and ask it questions. To that same end, it will enable travelers to chat with their policies, asking questions around sustainability, Benoit said.
The data portion, of course, is a very small portion of AI's potential to benefit global carbon efforts in the corporate travel industry. There are also the direct benefits that suppliers could realize: Airlines using AI to map out the most efficient flight paths to save fuel, for example, or hotels using smart room technology to ensure energy is not being wasted on lights or temperature controls when guests are not in the room, reducing their carbon footprints as parts of corporate travel programs.
As the technology gets more advanced, so do the possibilities.
"It's more assumptions, but I'm an optimist, so I believe this technology will support research and development to improve the way to process sustainable aviation fuel, which is critical to de-carbonize the air industry," Benoit said. "We could expect this AI will support the development of different types of SAF using more advanced technology."
But getting there also will come at an environmental cost, and the size of that cost is becoming better understood as well.
"Alongside the incredible promise and benefits of AI, we recognize the resource intensity of these applications and the need to address the environmental impact from every angle."
Or the Tiger?
With the advent of any new technology comes a discussion of the resources needed to support it, be it the significant energy required for cryptocurrency mining or the rare earth metals required to make smartphones run. Generative AI technology is no exception, and some data shows its impact could be substantial.
The massive data centers needed to support big tech's cloud infrastructure already is a significant source of carbon emissions, electricity and water use, particularly in the cooling needed to keep servers from overheating. A report from Columbia University's Climate School last year said data centers around the world account for between 2.5 percent and 3.7 percent of global greenhouse gas emissions—that's a bigger percentage than the airline industry.
Large language model technology like ChatGPT requires large amounts of power to operate, so as their use expands, that percentage stands to grow. Per the Columbia report, data centers could account for 1.86 percent of global electricity demand by 2030—compared with between 0.9 to 1.3 percent in 2021.
That doesn't take into account the water needed to run the servers as well. A University of California Riverside study showed that running between 20 and 50 queries in ChatGPT uses about 17 ounces of water, which is lost in the form of steam emissions.
"Alongside the incredible promise and benefits of AI, we recognize the resource intensity of these applications and the need to address the environmental impact from every angle."
FCM Extension head and Shep founder Daniel Senyard said, as a part of FCM's AI Center of Excellence, he had his own eye-opening experience when he was speaking at an event and was preceded by Stephanie Hare, a researcher and technology ethics expert who outlined the energy consumption of AI technology—particularly with water.
"We've heard the next war is going to be over water, and we may get to the point where these decisions need to be made over where the water is going: to communities or to support the tech infrastructure that keeps the world running," Senyard said.
Finding the Right Door
Tech companies are well aware of the growing need for data center energy and say they are working to reduce that impact. Microsoft, for example, in a recent blog post said it is "deeply invested" in improving energy and water efficiency at data centers and for its cloud services.
"Alongside the incredible promise and benefits of AI, we recognize the resource intensity of these applications and the need to address the environmental impact from every angle," Microsoft chief sustainability officer Melanie Nakagawa said in the blog post. "In line with our commitment to responsible AI and our ambitious sustainability commitments, we’re determined to tackle this challenge so the world can harness the full benefits of AI."
The International Organization for Standardization, in fact, is working on developing criteria around "sustainable AI," including with electricity and water use. "Often, these aspects are hidden from the end users, and the objective of the report is to make these things visible to the end users and to inform them of the choices they make," said Harm Ellens, the project's editor.
As progress is made on that front, companies implementing AI to solve sustainability issues will have to find that balance. Simple data management projects, for example, "can be done internally and on a small scale without having to leverage gigantic, energy-burning technology," Alō Index's Feinberg said.
There are plenty of other reasons to be cautious and purposeful when implementing AI technology as other potential land mines are sorted out, such as data security and reducing biases and "hallucinations." As tech companies sort out and hopefully reduce that impact, consider that one other motivation for caution.
"While AI can be super useful, everything looks like a nail when you have a hammer," Senyard said. "People want to throw AI at everything, and there is a sustainability cost."
Built with Shorthand