ChatGPT launched on November 30, 2022. That was basically four months ago, but given its widespread adoption and the media’s nearly endless coverage of artificial intelligence topics, it feels as though large language models have been in our daily lives for years. Meanwhile, a divide in the national discourse continues to widen. One camp worries that AI will steal jobs and endanger humanity. Another promises a digital utopia, where exponential technologies save the planet. It’s here where an interesting yet seldom discussed intersection exists: yes, AI may have capabilities to help curb the devastating effects of climate change, but it’s also a heavy contributor to those same problems. Taking some cues from companies like Salesforce, big tech can reduce these impacts.
The Role of AI and Exponential Technologies in Climate Change
Businesses like Microsoft, Google, and Salesforce are rushing to incorporate AI into their solutions. MSPs and staffing providers are seeking ways to modernize and push forward through AI’s advances in automation, business intelligence, content creation, and data analysis. Scientists believe these technologies can assist humans in conquering the barriers that are preventing us from controlling the rampant spread of climate change and detrimental emissions. But at the same time, the machines themselves are polluting.
“The Earth is facing an imminent risk of crossing tipping points in its life support systems,” the World Economic Forum reported. “When that happens, self-reinforcing cycles will kick in that could potentially lead to a ‘hothouse Earth’ state.”
The digital sector represents a key industry in this debate. It’s on track to reduce its own emissions, which represent 1.4% of the global total, and it has the opportunity to cut global emissions in half by 2030 while driving exponential growth in data performance. Some in the digital sector have also assumed the mantle of strong leaders in championing the demand for 100% renewable energy.
“Our assessment, as laid out in the Exponential Climate Action Roadmap, is that digital technologies could already help reduce global carbon emissions by up to 15% – or one-third of the 50% reduction required by 2030 – through solutions in energy, manufacturing, agriculture and land use, buildings, services, transportation and traffic management,” wrote the World Economic Forum. “This corresponds to more than the current carbon footprints of the EU and the US combined. But it is through the Fourth Industrial Revolution – particularly 5G, the Internet of Things (IoT) and artificial intelligence (AI) – that the digital sector can take the pace of change to the next level.”
Jim Bellingham, executive director of the Johns Hopkins Institute for Assured Autonomy, also believes that technology has an instrumental role to play in solving the climate crisis — particularly through AI.
“Climate data sets are enormous and take significant time to collect, analyze, and use to make informed decisions and enact actual policy change,” Bellingham explained. “Using AI to factor in elements of climate change that are constantly evolving helps us make more informed predictions about changes in the environment, so that we can deploy mitigation efforts earlier.”
“AI is one tool that provides insights into where uncertainties come from related to climate change and that can help us understand what the models are telling us, which can feed back into better observation programs, improving the models, and even using AI as part of the model system,” he added. “Another application of AI to climate change is how satellites orbiting in space are used to make observations and assess changes to the Earth. Satellites can help monitor forest fires and determine potential sources of carbon dioxide that are found in the environment.”
On the other hand, big tech is also one of the biggest producers of carbon emissions, as a study published by Electronics Hub demonstrated. Of the Big Five tech companies, Amazon emerged as the largest polluter: “Amazon produced over 16 million metric tons of CO2 in 2021–nearly 20 times the carbon footprint of Microsoft, which polluted least out of the 5. That said, Microsoft’s carbon footprint is still pretty big–at nearly 870,000 metric tons, it’s about as big a polluter as the population of Rhode Island.”
Technology’s Filthy Footprint
The revelation of crypto-assets’ impact on the environment shone a stark spotlight on the tech space in general. In September 2022, the White House issued a fact sheet on “Climate and Energy Implications of Crypto-Assets in the United States,” accompanied by an executive order to ensure the ecologically responsible development of digital assets. The boom in cryptocurrency mining offered a dire example:
“Global electricity generation for the crypto-assets with the largest market capitalizations resulted in a combined 140 ± 30 million metric tons of carbon dioxide per year (Mt CO2/y), or about 0.3% of global annual greenhouse gas emissions. Crypto-asset activity in the United States is estimated to result in approximately 25 to 50 Mt CO2/y, which is 0.4% to 0.8% of total U.S. greenhouse gas emissions. This range of emissions is similar to emissions from diesel fuel used in railroads in the United States.”
Energy Monitor’s Oliver Gordon also provided equally troubling data: “Although estimates of the total environmental footprint of all crypto assets vary, it is thought that the electricity demand of Bitcoin alone exceeded 13GW, with an associated carbon footprint of more than 65 million tonnes of CO2 (MtCO2), in 2021. That demand is just over half that of all global data centres combined and represents nearly 0.5% of global electricity consumption. The associated carbon footprint is bigger than the global CO2 reduction created by all electric vehicles. On top of that, other cryptocurrencies were estimated – prior to the Merge – to add another 50% to Bitcoin’s energy demand.”
But crypto serves as just one illustration in a broader mural of challenges. AI is likewise a resource-demanding digital asset. Training ChatGPT, according to researchers from the University of Colorado Riverside and the University of Texas Arlington, required enough water to fill a nuclear reactor’s cooling tower, as Mark DeGeurin reported in Gizmodo:
“Popular large language models (LLMs) like OpenAI’s ChatGPT and Google’s Bard are energy intensive, requiring massive server farms to provide enough data to train the powerful programs. Cooling those same data centers also makes the AI chatbots incredibly thirsty. New research suggests training for GPT-3 alone consumed 185,000 gallons (700,000 liters) of water. An average user’s conversational exchange with ChatGPT basically amounts to dumping a large bottle of fresh water out on the ground, according to the new study. Given the chatbot’s unprecedented popularity, researchers fear all those spilled bottles could take a troubling toll on water supplies, especially amid historic droughts and looming environmental uncertainty in the US.”
That’s just an overview of ChatGPT. Identical water consumption issues affect every organization developing AI and large language models.
“In 2019,” DeGeurin noted, “Google requested more than 2.3 billion gallons of water for data centers in just three states. The company currently has 14 data centers spread out across North America which it uses to power Google Search, its suite of workplace products, and more recently, its LaMDA and Bard large language models. LaMDA alone, according to the recent research paper, could require millions of liters of water to train, larger than GPT-3 because several of Google’s thirsty data centers are housed in hot states like Texas; researchers issued a caveat with this estimation, though, calling it an ‘approximate reference point.’”
Over 2.2 million U.S. residents lack water and basic indoor plumbing, based on World Economic Forum data. Another 44 million live with inadequate water systems. By 2071, with rising temperatures and droughts — influenced by human activity — 50% of the nation’s 204 freshwater basins won’t be able to meet monthly supply demands.
“AI’s hefty water demands will likely become a growing point of contention, especially if the tech is embedded into ever more sectors and services,” DeGeurin cautioned. “Data requirements for LLMs are only getting larger, which means companies will have to find ways to increase their data centers’ water efficiency.”
Following the Example of Sustainable Tech Leaders Can Help
Some tech pioneers realized early on that saving the planet and improving livelihoods meant not contributing to the same problems. Ethereum, perhaps the most popular cryptocurrency behind Bitcoin, managed to cut its electricity consumption by 99.84% — a reduction equivalent to the annual power requirements of Ireland or Austria — through an event called the Merge. It replaced the blockchain’s “proof-of-work” (PoW) mining mechanism with an alternative coined “proof-of-stake” (PoS).
With PoS, Energy Monitor’s Oliver Gordon explained, “the network does not incentivise participants, or ‘minters’, to compete on computational power to create new blocks. Instead, the block selection process is primarily based on wealth. Minters must buy some of the native currency from a cryptocurrency platform, which is then used as collateral in the ‘staking’ process – a minimum of 32 units of Ether, in Ethereum’s case. The software then randomly chooses a “staker” to create the next block for the blockchain. The larger the amount staked, the greater the chance of being chosen. While minters still need a device with sufficient storage capacity and an active internet connection, the computational power of the device is irrelevant. This negates the need for vast mining networks of electricity-gorging computers.”
Salesforce is perhaps the most recognizable tech company to promote massive sustainability initiatives. Salesforce has net zero residual emissions, achieved 100% renewable energy for its operations, and is a founding partner of 1t.org. Salesforce achieved carbon neutrality by taking the following steps:
- Committing to net zero emissions. Salesforce committed to achieving net zero emissions across its full value chain by 2040, in line with a 1.5°C future.
- Reducing emissions. Salesforce has reduced its emissions by 50% since 2012. The company has done this by investing in renewable energy, improving energy efficiency, and reducing its reliance on fossil fuels.
- Offsetting emissions. Salesforce offsets its remaining emissions by investing in projects that reduce or remove greenhouse gasses from the atmosphere. These projects include planting trees, investing in renewable energy, and improving energy efficiency in developing countries.
Salesforce's journey to carbon neutrality is an example of how businesses can take action to address climate change. The company's commitment to net zero emissions and its progress in reducing emissions are a positive step for the environment. Salesforce has also taken the following actions to reduce its environmental impact.
- Using sustainable materials. Salesforce uses sustainable materials in its products and packaging, such as recycled paper and plastic.
- Recycling and composting. Salesforce recycles and composts all of its waste, including paper, plastic, metal, and food waste.
- Conserving water. Salesforce has installed water-saving fixtures in its buildings and has implemented a water conservation program.
- Energy efficiency. Salesforce has made its buildings more energy efficient by installing energy-efficient lighting, HVAC systems, and appliances.
- Green transportation. Salesforce offers employees incentives to use public transportation, carpool, or bike to work. The company also has a fleet of electric vehicles.
How GPT Models Can Improve Sustainability
In addition to adopting the best practices of tech companies like Salesforce, researchers believe there are simple yet critical steps those involved in training large language models can take.
- Location matters. Outside temperatures affect the amount of water required to cool data centers. Choosing cooler regions for data center housing could reduce the need for more water in cooling systems.
- Time matters. Training AI models at night, when temperatures drop, could cut down on water consumption and improve water efficiency. Models could restrict full usage until water-efficient hours. Similar practices exist for off-hours appliance usage in certain European countries.
- Partners matter. Like Salesforce. AI developers should enforce transparency and sustainability plans from any third parties they rely on for colocation, data collection, public cloud hosting, and more.
Salesforce's commitment to sustainability is an example of how businesses can make a positive impact on the environment. The company's actions are helping to reduce its environmental impact and are inspiring other businesses to do the same. As the popularity of AI and large language models grows, it will remain essential to the preservation of the environment that companies make inroads to actually saving the planet through their development practices rather than just developing tools that purport to help overcome the problems to which they’re unwittingly contributing.