4 Big AI Sustainability Prospects, and One Big Problem
With the hype swirling around AI’s generative language possibilities, some of the less discussed benefits could drive innovation for a sustainable future. But there’s a snag that needs a solution.
![green tree with deep roots green tree with deep roots](https://eu-images.contentstack.com/v3/assets/blt69509c9116440be8/bltd74ed8247bd7c4d4/64bfdfe25f5c9cbddb20d44e/AI_Sustainable_Slide_1_Vladimir_Razguliaev_-alamy.jpg?width=700&auto=webp&quality=80&disable=upscale)
Vladimir Razguliaev via Alamy Stock
When more than 1,000 tech leaders signed an open letter calling for a pause on fast-moving advancements in AI early this month, they laid bare the ethical dilemma posed by language learning models and possible dire workforce impacts. But often lost in the talk of dystopian outcomes is the impact the new technology could have on the near-universal drive for a more sustainable future.
According to a report by PwC and Microsoft, AI’s role could be a major factor in future sustainability efforts. “If harnessed in the right way, emerging technologies, with AI at the vanguard, could be transformational in efforts to tackle some of the world’s most pressing environmental challenges,” the report says. “Examples include AI-infused clean distributed energy grids, to smart urban mobility, precision agriculture, sustainable supply chains, environmental monitoring and enforcement, and enhanced weather and disaster prediction and response.”
In the following slides, InformationWeek looks at ways AI could prove to be crucial for effective sustainability solutions. (The last slide focuses on one big caveat to all the promise AI holds for a sustainable future).
Fast-moving changes in the energy-sector will be crucial to effective sustainability efforts. And experts say AI will be a key factor in making those changes achievable.
A report by the World Economic Forum says AI has the potential to contribute significantly to accelerate the “global energy transition” by identifying patterns, improving system performance, and predicting outcomes of complex situations. The report, co-written by BloombergNEF and the German Energy Agency, reviews AI adoption and possibilities in the energy sector.
“Digital technologies -- particularly AI -- are key enablers for this transition and have the potential to deliver the energy sector’s climate goals more rapidly and at lower cost,” the report says.
The report found AI can create value for the energy transition, saying that for every 1% additional efficiency in demand creates $1.3 trillion in value between 2020 and 2050 due to reduced investment needs.
“AI is already making its mark on many parts of society and the economy,” Roberto Bocca, head of energy at World Economic Forum, said in a release. “In energy, we are only seeing the beginning of what AI can do to speed up the transition to the low-emissions, ultra-efficient, and interconnected energy systems we need tomorrow . . . Governments and companies can collectively create a real tipping point in using AI for a faster energy transition.”
While AI’s use in the energy sector is promising, the report authors say more investment and work is needed. “One of the key findings from our workshops was that whilst we could identify many tangible opportunities for AI in the energy transition, there was a real need for a set of common guiding principles to make these opportunities scalable,” Jon Moore, chief executive officer at BloombergNEF, said in a release.
Our farmers fuel the world’s ability to feed populations, but that life-giving effort has a big impact on the environment.
An AI in agriculture article from ITRex Group details how AI technology can be used in soil preparation, sowing, fertilizing, irrigation, crop protection, harvesting and more to create efficient and sustainable farming into the future.
“By analyzing operational data and highlighting process inefficiencies, artificial intelligence finds ways for agribusinesses to increase yields without using any extra resources” according to the article. “Automating crop . . . farming with drones eliminates human error and streamlines those processes to allow an increase in both quantity and quality of labor.”
These efforts create cost reductions and profit increases as they ease impacts on the environment. “AI solutions spot wasteful resource consumption patterns and suggest optimization scenarios,” the article stated. “Given that AI maximizes yields without requiring farms to use any additional resources and cuts costs associated with various stages of agricultural processes, an increase in profits follows as a natural and pleasant consequence of employing AI in agriculture.”
The biggest use case opportunity in agriculture, the article suggested, is in crop protection. Advances in AI solutions for herbicide and pesticide optimization, pest attack prediction, crop health monitoring, intruder identification, and weather prediction will all have a massive upside for farmers’ ability to run sustainable operations. These technologies are already being deployed in areas around the world and will continue to proliferate.
As the global COVID-19 pandemic wreaked havoc, the general population was suddenly immersed in what was once an esoteric concept for many: the supply chain. The global concert of logistics, manufacturing, and transport of goods suddenly became a part of daily life as important products -- as crucial as semiconductors and as mundane as toiletries -- were in short supply.
In the short term, ongoing efforts have eased some of the supply constraints on industries worldwide. Leaders are turning their attention to how AI can create optimizations that would alleviate supply chain snarls in the future. Along with that promise of streamlined and resilient supply chain comes the bonus of more sustainable industries.
According to a report by McKinsey & Company, AI will impact the future of supply chain in the areas of marketing, sales, procurement, planning, logistics and distribution, and production. “The good news is that AI-based solutions are available and accessible to help companies achieve next-level performance in supply chain management,” the report says. “Solution features include demand-forecasting models, end-to-end transparency, integrated business planning, dynamic planning optimization and automation of the physical flow . . . Successfully implementing AI-enabled supply-chain management has enabled early adopters to improve logistics costs by 15%, inventory levels by 35%, and service levels by 65%, compared with slower-moving competitors.”
AI has made a huge splash in the transportation industry -- from autonomous driving to traffic management solutions, the technology is racing to optimize our travel needs. And optimized travel means less energy consumption.
According to Market Data Forecast, the global AI in transportation market is expected to reach $3.8 billion by 2026 with an annual growth rate of 15.8%. AI in transportation is a key component of the “sustainable city” concept that seeks to optimize transport and energy use in major urban areas globally, according to the report writers.
“The transportation sector in particular has been the largest consumer of AI technologies,” the report said. “Rapid use of learning technology automation for product innovation, such as autonomous vehicles, parking and lane changing aids, and smart energy systems are driving AI strongly in the transportation market.”
It seems we’ve reached a precarious moment in history when environmental researchers are asking us to consume less energy for the health of the planet, while our rapidly advancing technology is demanding more and more. AI consumes a massive amount of compute power, and the manufacture of increasingly powerful machines itself requires energy, natural resources, and generates carbon emissions.
Google and University of California Berkeley researchers reported that artificial intelligence ate up 10-15% of the company’s annual electricity consumption in 2021 -- about 2.3 terawatt hours. The researchers noted that training OpenAI’s GPT-3 specifically gobbled up 1.287 gigawatt hours -- the same as 120 US homes consume in an entire year (according to 2021 figures). Yet while much is made of the training phase, the Google researchers cited NVIDIA and Amazon's 2019 reports report that roughly 90% of the machine learning electricity usage is spent on inference, not training. A widely referenced 2019 paper from University of Massachusetts Amherst researchers states that training a single AI model can produce as much carbon as five cars over their lifetimes.
While these anecdotal examples provided by willing companies and institutions offer a peek into the issue of AI power consumption, the total impact remains unknown. Researchers say more transparency is needed to understand necessary mitigation strategies to reduce AI’s carbon footprint. Some strategies are already being employed. “The choice of [deep neural network], datacenter, and processor can reduce the carbon footprint,” the Google-backed report stated. “We believe ML . . . requiring large computational resources should make energy consumption and CO2e explicit when practical.”
Google has an aggressive goal of moving to 100% carbon-free energy sources by 2030.
David Patterson, one of the report’s writers and a software engineer at Google, says reducing AI’s carbon footprint is achievable. “We believe that technology can and must play a crucial role in facilitating the transition to a lower-carbon world, and AI can play a key role. We are using four key practices that reduce the carbon (and energy) footprint of ML workloads by large margins,” he tells InformationWeek.
Ricardo Vinuesa, associate professor at Sweden’s KTH Royal Institute of Technology and lead faculty at its Climate Action Centre, tells InformationWeek that while AI has the potential to be a sustainability ally, action on the part of businesses and governments is crucial. “Overall, I would say the message is positive, but with some caveats that require global coordination,” Vinuesa says.
A paper Vinuesa published in January 2020 along with other researchers shows AI has the potential to achieve 79% of 169 sustainability goals, while inhibiting achievements of 35% of targets. “This in principle sends a positive message on the usage of AI, although one needs to be careful because only one target being negatively affected can counterbalance all the positives.”
The complexity of AI’s sustainability balancing act calls for cooperation between nations, academia and regulators, Vinuesa says. “I think the first step is on responsibility of the AI users and researchers. Focus should not be on larger and more expensive models, but rather on better models that can more effectively capture the physics of the phenomena under study,” he says.
He says the EU’s proposed AI Act -- which would create laws around AI “risk categories” -- is a good start. “Regulation, and more importantly, the involvement of experts in regulation development, are essential,” he says.
Check Out Other InformationWeek Slideshows
It seems we’ve reached a precarious moment in history when environmental researchers are asking us to consume less energy for the health of the planet, while our rapidly advancing technology is demanding more and more. AI consumes a massive amount of compute power, and the manufacture of increasingly powerful machines itself requires energy, natural resources, and generates carbon emissions.
Google and University of California Berkeley researchers reported that artificial intelligence ate up 10-15% of the company’s annual electricity consumption in 2021 -- about 2.3 terawatt hours. The researchers noted that training OpenAI’s GPT-3 specifically gobbled up 1.287 gigawatt hours -- the same as 120 US homes consume in an entire year (according to 2021 figures). Yet while much is made of the training phase, the Google researchers cited NVIDIA and Amazon's 2019 reports report that roughly 90% of the machine learning electricity usage is spent on inference, not training. A widely referenced 2019 paper from University of Massachusetts Amherst researchers states that training a single AI model can produce as much carbon as five cars over their lifetimes.
While these anecdotal examples provided by willing companies and institutions offer a peek into the issue of AI power consumption, the total impact remains unknown. Researchers say more transparency is needed to understand necessary mitigation strategies to reduce AI’s carbon footprint. Some strategies are already being employed. “The choice of [deep neural network], datacenter, and processor can reduce the carbon footprint,” the Google-backed report stated. “We believe ML . . . requiring large computational resources should make energy consumption and CO2e explicit when practical.”
Google has an aggressive goal of moving to 100% carbon-free energy sources by 2030.
David Patterson, one of the report’s writers and a software engineer at Google, says reducing AI’s carbon footprint is achievable. “We believe that technology can and must play a crucial role in facilitating the transition to a lower-carbon world, and AI can play a key role. We are using four key practices that reduce the carbon (and energy) footprint of ML workloads by large margins,” he tells InformationWeek.
Ricardo Vinuesa, associate professor at Sweden’s KTH Royal Institute of Technology and lead faculty at its Climate Action Centre, tells InformationWeek that while AI has the potential to be a sustainability ally, action on the part of businesses and governments is crucial. “Overall, I would say the message is positive, but with some caveats that require global coordination,” Vinuesa says.
A paper Vinuesa published in January 2020 along with other researchers shows AI has the potential to achieve 79% of 169 sustainability goals, while inhibiting achievements of 35% of targets. “This in principle sends a positive message on the usage of AI, although one needs to be careful because only one target being negatively affected can counterbalance all the positives.”
The complexity of AI’s sustainability balancing act calls for cooperation between nations, academia and regulators, Vinuesa says. “I think the first step is on responsibility of the AI users and researchers. Focus should not be on larger and more expensive models, but rather on better models that can more effectively capture the physics of the phenomena under study,” he says.
He says the EU’s proposed AI Act -- which would create laws around AI “risk categories” -- is a good start. “Regulation, and more importantly, the involvement of experts in regulation development, are essential,” he says.
Check Out Other InformationWeek Slideshows
When more than 1,000 tech leaders signed an open letter calling for a pause on fast-moving advancements in AI early this month, they laid bare the ethical dilemma posed by language learning models and possible dire workforce impacts. But often lost in the talk of dystopian outcomes is the impact the new technology could have on the near-universal drive for a more sustainable future.
According to a 2019 report by PwC and Microsoft, AI’s role could be a major factor in future sustainability efforts. “If harnessed in the right way, emerging technologies, with AI at the vanguard, could be transformational in efforts to tackle some of the world’s most pressing environmental challenges,” the report says. “Examples include AI-infused clean distributed energy grids, to smart urban mobility, precision agriculture, sustainable supply chains, environmental monitoring and enforcement, and enhanced weather and disaster prediction and response.”
In the following slides, InformationWeek looks at ways AI could prove to be crucial for effective sustainability solutions. (The last slide focuses on one big caveat to all the promise AI holds for a sustainable future).
About the Author(s)
You May Also Like