Bold, Urgent Proposals for Using AI in Climate Action

Invest in Programs to Collect and Process Data

In many parts of Asia, particularly those most vulnerable to climate disaster, the data required to build machine-learning applications is simply not available, or at least not in machine-readable formats.

Many governments lack the technical infrastructure and resources to regularly collect and update climate-relevant data.

Where available, it is often fragmented and scattered across government departments, with few protocols or standards for data-sharing.

Urvashi Aneja of Digital Futures Lab working at her computer.

Climate models cannot be readily transposed across diverse locations as they are unlikely to capture the specificities of local climate change impacts or how these impacts are refracted through a local lens.

Investing in programs to collect and process data in low-and-medium-income countries is vital if AI is to be effective in a broad and equitable climate change fight.

Invest in Adaptation

Asia faces a number of specific climate challenges. Sea levels in oceans surrounding Asia are rising slightly faster than the global mean.

High Mountain Asia, including the Himalayas and the Tibetan Plateau, contains the largest volume of ice outside of the polar region, and the rate of glacier retreat is accelerating.

Finally, increased intensity and frequency of extreme weather events, from drought to rainstorms and mudslides, have severe impacts on food and livelihoods as well as the spread of disease.

Much AI-related research and investment is concentrated in climate mitigation solutions. The severity and frequency of climate disasters in Asia demands a focus on adaptation solutions as well.

Man recording data from a solar farm.

Incentivize Data Sharing

Much of the data useful for advancing the climate change fight is proprietary and held by private companies, many of which are reluctant to share because of competition, security, and reputational reasons.

Data collection and curation is hugely expensive, laborious, and complex, making it harder for smaller or newer players to enter the market, and easier for established companies to leverage network effects and create data moats.

Data collection by private companies is also concentrated in commercially lucrative domains, such as energy efficiency and intelligent mobility, tending to exclude areas and people most vulnerable to climate disasters.

Mechanisms to incentivize data sharing by private actors are urgently needed.

Many regional governments are creating open data platforms. But simply opening data is not enough. Evidence from other sectors shows that stakeholders need the knowledge, capacity, and incentives to use these platforms for social good.

New domain and problem-specific coalitions must be intentionally curated. Key stakeholders, including governments, private players, climate and data scientists, and civil society organizations, must work together to identify data gaps, pool data resources, and co-develop solutions.

Drone used for farming to collect plant data and increase crop yield.

Center Climate Justice

How can we make sure AI uses are equitable and accountable to local communities?

The conventional way of thinking about data infrastructures is to use FAIR guiding principles, making data Findable, Accessible, Interoperable, and Reusable.

But this focuses on the technical qualities of data access. We must tie access to a clear social purpose and values. A good reference point is the CARE principles formulated by Indigenous data rights movements that emphasize values of Collective purpose, Accountability, Responsibility, and Ethics.

Balance Cost With Impact

AI production has massive environmental costs. The process of training one AI model, for instance, can emit more than 626,000 pounds of carbon dioxide equivalent—nearly five times the lifetime emissions of the average American car (and that includes the manufacture of the car itself.)

Training GPT-3 alone required 185,000 gallons (700,000 liters) of water, and a typical user’s interaction with ChatGPT is equivalent to emptying a sizable bottle of fresh water onto the ground.

Disposal of AI computers hardware causes air and groundwater pollution and most disposal happens in underdeveloped countries.

At the same time, we haven’t yet developed standards for calculating AI’s impact on climate action. Working collaboratively, coalitions must develop metrics to evaluate the efficacy of emerging AI-based interventions, with transparency as a core condition for private sector data access.

If AI is to meaningfully contribute to addressing the climate crisis, building new products will not be enough. The first step? We need to fundamentally redesign the systems and logics through which AI is developed and deployed.