Salesforce Tackle AI’s Environmental Issues With 6-Point Plan

The discourse around the rapid rise of artificial intelligence has seeped into our daily lives; both the convenience and the concerns surrounding AI have sparked ongoing debates about its future, from how to harness it successfully to the ethical boundaries being pushed by AI pioneers. However, one thing that is rarely considered is the environmental impact.

Running machine learning algorithms is expensive and requires a lot of energy, so much so that it makes up 2-3% of annual global power consumption, which is, according to Boston Consulting Group, likely to triple by 2030. As a consequence, we could see a significant rise in e-waste, endangering the environment, not to mention the overall health of the human race.

To mitigate this issue, Salesforce is lobbying for a set of regulations that will require companies to disclose their emissions data and AI standards. Alongside this, they are implementing their own six-point plan to support sustainable AI use with their Einstein 1 Studio. Here’s everything you need to know…

How Does AI Impact the Environment?

As mentioned, a lot of power is required to run these large language models efficiently. Inevitably, as more power is used, the risk to the environment becomes more and more prominent.

A study conducted by researchers at the University of Massachusetts observed just how much energy is being consumed to run these models. The results found that once training was complete for a large language model, it would have omitted around 626,000 pounds of carbon dioxide. For context, this works out the same as 300 round-trip flights between San Francisco and New York. 

Regularly releasing this level of greenhouse gas into the atmosphere could potentially have an impact on the current global warming issue.

Here are a few other important environmental considerations to keep in mind:

• The use of water: Believe it or not, a considerable amount of water is required to help AI run smoothly. As these high-power machines are prone to overheating, they need constant hydration to keep them cool. Freshwater resources are limited, and using vast amounts of water to cool AI systems could contribute to water scarcity in certain regions. On top of this, the water used for cooling needs to be pumped, treated, and transported – all of which require energy.
• E-Waste: The popularity of personal tech devices (computers, phones, etc.) means that discarded hardware is a significant issue, and AI is only going to worsen that. E-waste contains extremely hazardous chemicals such as lead and mercury. The World Economic Forum predicts that over 120 million metric tons of e-waste will have been generated by 2050 unless proper e-waste management is implemented.
• Resource Extraction: Resources such as metals, minerals, and oil are all needed to create AI tools – The way these are sourced is capable of causing pollution issues. As more AI-powered devices are created, more resources are needed which subsequently harms animals, plants, and the environment.

What Is Salesforce Doing About It?

In light of this, Salesforce was quick to announce its intention to foster sustainable AI practices going forward. Previously vocal about its pursuit of environmentally friendly practices, Salesforce even added “climate” to its public policy priority list back in 2021. So it comes as no surprise to see the tech giant continue the dialogue around the growing environmental issues we’re seeing with AI.

Salesforce’s Sustainable AI Policy Priorities say: “In the face of the escalating climate crisis, AI has the potential to both exacerbate and alleviate environmental challenges. Policy frameworks must strike a balance between maximizing AI’s benefits while mitigating its environmental impact.” 

This detailed framework highlights how Salesforce plans to mitigate the impact of AI on the environment. The guidelines outline six priorities for incentive programs and new regulations, which is split into two categories. Here’s what it looks like…

Measuring and Managing Environmental Impacts

Here are three key ways that Salesforce are approaching a sustainable future for the use of AI:

1. Require AI emissions disclosure for all: Companies are to release the energy efficiency and carbon footprint associated with their AI development using standardized metrics (kWh/tCO2e).
2. Acknowledge environmental impact as a risk factor: When determining the risks of an AI system, companies should consider the overall environmental impact to be pivotal. This could include when a system uses too much energy during its training process.
3. Establish efficiency standards for high-risk AI systems: For models that are considered high-risk, companies should consider these steps:
   • Optimize the model for efficiency e.g. the size of the model.
   • Utilize efficient hardware to run these models.
   • Leverage clean energy and low-carbon grids.

Accelerating Climate Innovation With AI

Salesforce also wants to encourage positive AI use by ensuring these three factors:

1. Enhance climate modeling, data analysis, and predictions: Create and maintain open-source environmental data and forecast databases that are freely accessible to train AI models going forward.
2. Encourage public-private partnerships: Foster a working relationship between academia, government, and private sectors to accelerate the environmental use of AI. These learnings and conversations can then be shared in the open-source database.
3. Spur Understanding and Innovation of Sustainable AI Use: Implement grants and tax incentives to promote the environmental application of AI.

Alongside this, Salesforce is continually supporting companies with research as part of their Salesforce Accelerator – AI for Impact initiative. With $2 million in funding, product donations, and expert pro bono consulting, Salesforce is helping five selected nonprofits harness AI to tackle climate challenges and drive positive change.

Selected organizations include: 

• WattTime: An environmental tech nonprofit working to reduce carbon emissions by enabling devices to use clean energy.
• One Concern: A company focused on using AI for disaster response and resilience planning, helping communities prepare for and mitigate the effects of natural disasters.
• Project Drawdown: A nonprofit working on solutions to climate change, providing research and analysis to help drive effective climate action.
• AI for Climate: A nonprofit organization that leverages AI to address climate challenges, particularly in developing countries.
• Climate Policy Radar: A platform that uses AI and machine learning to track climate policies and their impact globally, providing insights to help drive effective climate action.

Summary

As the discussion surrounding AI and its environmental impact grows, Salesforce’s clear stance will help position the company as a leader in sustainable AI development. By acknowledging the environmental costs associated with AI and actively working to mitigate them through initiatives like AI for Impact, Salesforce is demonstrating its long-term commitment to both innovation and sustainability.

Be sure to leave your thoughts in the comments below!