IBM and UN team up on green, equitable AI models

IBM has teamed up with the UN Development Programme (UNDP) to build AI models for sustainable, equitable energy.

UNDP and IBM spent the past two years building two new models for UNDP’s Data Futures Exchange, the UN organisation’s open source hub collecting data innovation focused on development.

Both models – one is an AI model focused on forecasting electricity access, the other a statistical model mapping clean energy equity – will be available to the general public for free via the Data Futures Exchange GeoHub, a collection of geospatial data and services. The project is powered by IBM’s watsonx, IBM Cloud and IBM Environmental Intelligence.

“Bringing together UNDP’s knowledge and global leadership in sustainable development and IBM’s breakthrough innovations in AI and hybrid cloud, we are proud to unveil solutions that demonstrate the power of technology to make a lasting, positive impact on our environment and in our communities,” said Justina Nixon-Saintil, IBM Vice President and Chief Impact Officer.

The aim is to ensure the necessary data is available to the general public as well as policymakers, journalists and more. “By making innovative models freely accessible to the public, we aim to empower leaders, organisations and community members alike with the insights to make impactful energy decisions around the world,” said Nixon-Saintil.

That access is important, IBM notes, because advanced models like the two developed for the programme are not always available to everyone because of their high costs and complexity.

Electricity Access Forecasting model

First up is the Electricity Access Forecasting model. This allows users to explore future trends in electricity access to uncover how people could be at risk of being left behind. The datasets powering the model include household-level electricity access up to subnational energy data for more than 100 countries, offering “hyper granular” estimates for electricity access down to the square-kilometer.

This model uses IBM’s Watsonx AI platform and IBM Cloud, with land use data provided by IBM Environmental Intelligence. This churns through information on population, infrastructure, urbanisation, elevation and satellite images.

“By modeling these factors to make a future forecast, the Electricity Access Forecasting model provides a distinct advantage compared to more commonly available, current-day estimates of electricity access,” the organisations said in a joint statement.

Clean Energy Equity Index

The second model is the Clean Energy Equity Index. Developed by IBM, UNDP and Stony Brook University, this statistical geospatial model considers factors including environmental, economic and social indicators. Social indicators such as education, local jobs, foreign aid, access to loans and relative wealth. The model calculates a score for clean energy equity across 53 African countries.

“This score reflects both opportunities for clean energy development as well as urgency, through the lens of equity and a just transition,” the organisations said.

The dashboard lets GeoHub users customise each factor to test the impact of different interventions on local outcomes, helping to show governments how and where to invest.

“In this dashboard, GeoHub users can also individually view and customise each environmental, economic, or social factor analysed in the model, to evaluate which factors have the greatest impact on equitable access to clean energy, empowering better decision making,” stated the organisations.

Energy solutions

The aim is to use the models to help countries and communities make the right decisions as they move towards clean, accessible energy.

“The solutions we’ve co-created provide a credible evidence base to help countries make meaningful and practical progress towards a just energy transition,” said Laurel Patterson, Head of the UNDP SDG Integration Team, UNDP Bureau for Policy and Programme Support. “Net-zero investment and people-centered development strategies are fundamental to accelerate the [sustainable development goals].”

The work comes as companies developing AI are under increasing pressure for the amount of energy that it takes to develop and run a model, with Google’s emissions leaping 48% in the last four years, reportedly largely off the back of its AI efforts.

These new tools from UNDP and IBM help show how AI technologies could be useful in providing a data-driven route to fighting climate change, improving energy sources, and ensuring access for all – meaning there’s potentially an environmental benefit to AI, even if there are costs too.