Although the Code Green series wrapped up in 2025, Digital Futures Lab has continued working on critical questions at the AI and climate intersection. In our latest research, we deep dive into one of our favourite issues in the series - the socio-environmental impacts of material AI infrastructures.

Data centres have captured the public imagination in recent years - from announcements by major technology companies about orbital AI infrastructure to claims that vast concrete campuses will power the next wave of economic growth. At the same time, public scrutiny has intensified. Across the world, grassroots organisations and local communities are raising concerns about ecological, health and social costs of these facilities. In parallel, the industry itself is adapting, working to safeguard investments against escalating climate risks.

In many developing economies, data centres are emerging as critical infrastructure. Yet their rapid expansion presents complex trade-offs. For countries like India, which are relatively new to the hyperscale data centre race, important questions remain: What direction will this growth take? And at what cost?

Globally, the surge in data centre construction is closely tied to the rise of artificial intelligence (AI). In India, this expansion is linked to public policy ambitions that position AI as a tool to address development challenges - from public health to education, rather than primarily focusing on enterprise or agentic AI applications. However, governments also seem to rely on foreign investments and partnerships for data centres even as technology companies report significant increases in energy consumption and emissions, highlighting the resource intensity of AI-driven growth.

In late 2025, supported by the Green Screen Coalition, DFL undertook a short-term research project to better understand India’s evolving data centre landscape. Our study mapped data centre clusters across the country and examined their proximity to energy and water resources. We found concentrations in Tier 1 cities - India’s established economic hubs, alongside a gradual shift toward Tier 2 and Tier 3 cities in pursuit of lower operational costs.

India’s data centre hubs

At the same time, we observed vacancies in several existing facilities, prompting questions about the possibility of a looming data centre glut. As state governments roll out incentives, including tax benefits and electricity subsidies, we also ask: how are states measuring returns on this investment? While data centres are capital-intensive, they generate relatively limited direct employment. What, then, constitutes meaningful economic benefit?

Our research further connects renewable energy ambitions with land and infrastructure policy. We note growing partnerships between domestic energy providers and global technology firms seeking to expand their data centre presence in India. Yet, despite these developments, balanced public discourse on India’s resource constraints and what they mean for the long-term sustainability of data centre growth remains limited. We are already witnessing early signs of local resistance, including concerns raised by human rights defenders in Andhra Pradesh regarding water stress linked to proposed data centre developments.

We believe India stands at a critical juncture. Enabling inclusive digital growth requires grappling with the material realities of infrastructure - energy, water, land and jobs, among others, and aligning technological ambition with ecological and social sustainability.

Our report concludes with several pressing questions:

• Must the data centres of the future resemble the large, resource-intensive facilities of today? Could smaller, more resource-efficient models render current designs obsolete?

• How can local governance authorities, often at the frontlines of infrastructure negotiations, be better equipped to safeguard community interests?

• How can industry and policymakers build stronger bridges to advance equitable digital futures?

We invite researchers, policymakers, industry leaders, civil society organisations, and funders to engage with us on these questions. We are actively seeking partnerships to deepen this work and shape a more sustainable and inclusive trajectory for material digital infrastructure in India.

To learn more, we encourage you to read the full report at this link and to get in touch to explore collaboration opportunities. You can reach us at dona@digitalfutureslab.in

A call back to our previous work on AI and climate! Check out our research and public engagement series.

Credits

Writing - Dona Mathew | Images - Meredith Stinger

Dear Reader,

Welcome to part two of our newsletter wrap-up!

In our last issue, we encouraged investors to adopt a whole-of-system lens in funding AI-for-Climate solutions, recognising the network effects of investing in technologies that need resource-intensive infrastructures. In this final issue, we end on an optimistic note, and highlight some of the key areas of AI intervention for climate action from the series.

In July this year, we saw AI for climate action heating up on the international stage. The Technology Executive Committee (TEC) of the United Nations Framework Convention on Climate Change (UNFCCC) published a technical paper on the opportunities and challenges of adopting AI for climate change mitigation and adaptation in developing countries.

In the paper, they address issues of digital divide, capacity-building needs, climate data gaps, regulatory needs and AI’s resource intensity as key challenges. Highlighting promising use cases, the paper also offers case studies for early warning systems, earth observation, environmental monitoring, climate modelling, energy management, and community education, to name a few. The vast array of emerging use cases are an indication of AI becoming increasingly ubiquitous in aiding climate action - and its promises, if adopted responsibly.

To highlight some of the likely uses of AI for climate action that emerged from our series, we dive into the expertise of thought leaders who joined us in our podcast episodes as well as our learnings from research papers.

Expert Talk: Seeding R&D Signposts for the Future

Throughout the series, experts have helped us cut through the hype of AI, sharing instances where technology has worked, identifying some key intervention areas:

Diverse data for decision-making

Hanyuan ‘Karen’ Wang and ChengHe Guan spoke about AI’s powerful ability to analyse and make sense of large amounts of data. In a world where information is more readily available, using machine learning to understand patterns in diverse and complex climate data can generate insights for policy-makers and technology developers. In sectors like conservation, Eleanor Slade and VV Robin gave examples of how AI innovations in biodiversity, like bioacoustics, are improving species identification and protection.

AI and human behaviour

Alpan Raval and Raj Cherubal highlighted AI’s potential in nudging towards climate-positive behaviour, noting the premise that a lot of environmental harms are driven by individual consumption patterns and complex decision-making landscapes. However, it also raises thorny questions around preserving autonomy and who shoulders accountability for behaviour harmful to the planet - individuals or polluting corporations?

Tracking and transparency in environmental impacts

In many issues of the newsletter, we saw AI’s role in assessing environmental project risks and taking sustainable paths towards ecologically-intensive activities like mining. However, as Luis Felipe Murillo reminded us, it is important to weigh these benefits against AI’s own environmental and human costs.

Sector-specific innovation

Sherif Elsayed-Ali and Priya Donti spoke about the value of a sector-specific lens to AI innovation, be it for AI to reduce emissions in sectors like the cement industry or to mitigate AI’s energy costs. “It's not very sexy,” Sherif told us, but focusing on optimisation for limited, essential use cases in high-emitting sectors like cement manufacturing may yield the most positive results for the planet.

Turn towards knowledge/physics-based models

The increasing creep of AI into the climate sector risks displacing fundamental science-based knowledge in favour of data-driven models. To avoid this marginalisation of scientific expertise and to foster innovation grounded in established scientific principles, we are seeing innovation in physics-based models that combine the best of both for solutions in the energy sector, for instance.

Tools for Collective Action

Cathy Richards and Izni Zahidi shared their experiences on how access to GeoAI tools and education is democratising mapmaking and enabling the inclusion of public social values in urban planning. They showcase how a thoughtful approach to using geospatial data can improve people’s lives and environmental outcomes.

Recap

In our previous issue, we highlighted the tensions in funding climate AI solutions, noting that funding considerations need to account for human, economic and ecological impacts of technology and its physical infrastructures. We reiterate our key takeaways here:

1. Community-driven interventions that centre local needs and non-tech alternatives are key to avoiding power imbalances between users and technology providers.

2. Strategies to account for cultural norms, technical usability, accessibility of language, explainability and transparency mechanisms are needed to implement AI responsibly.

3. Intentional investment in dialogue, infrastructure maintenance and capacity-building is essential to ensure the long-term sustainability of projects.

4. Investments in critical AI and data literacy can help bridge knowledge gaps and promote the responsible use of AI.

5. A critical climate lens that challenges techno-solutionism, and asks - how does this solution improve adaptation, mitigation or resilience? - during project initiation and design is important to ensure AI’s benefits outweigh its costs in addressing climate change.

6. A whole-of-system approach to climate tech investments, including support for R&D, and collaboration in developing frameworks for assessing AI’s social, economic and environmental impacts across the supply chain, is necessary.

Around the Web

• To access all of the resources we’ve highlighted in our series, visit the CODE GREEN Library.

• Climate shocks will increasingly impact the global food supply chain - this new app reveals just how complex and precarious this interconnected system is.

• In the latest issue of Branch Magazine, Lori Regattieri maps pathways to ensure technology development and energy transitions become spaces for justice, autonomy and collective reimagining.

• For insights into the trade-offs of the rare earth mining industry feeding carbon-free energy, along the border of Thailand and Myanmar, watch this explainer.

• Learn more about the Underground Atlas, an interactive tool that maps mycorrhizal fungi (which sequester huge amounts of carbon) diversity around the world, to help scientists prioritise and protect these natural carbon sinks.

Tuning in…

A bonus podcast episode! In this finale of the season, Urvashi Aneja, Founder and Director of Digital Futures Lab spoke with the team behind the Code Green series: Dona Mathew, Meredith Stinger and Tammanna Aurora. Listen in for insights into what went on behind the scenes, the team’s core takeaways and future research directions.

A huge, warm thank you to all Code Green readers and listeners who have tuned in this past year. Your engagement has been encouraging in a landscape where we urgently need more voices. We hope you’ve learned as much throughout this series as we have!

Access the entire series on codegreen.asia

Share this with your friends who you think may be interested, or if you see any dots connecting from your work for a potential collaboration, write to us at codegreen@digitalfutureslab.in

To continue to remain in touch with our work across climate, gender, labour, foresight, subscribe to DFL’s newsletter.

Digital Futures Lab

DFL's newsletter is a monthly roundup of our ongoing work, readings, events and research outputs in the tech policy and culture space, specific to the Global South.

Credits

Writing: Dona Mathew | Illustrations: Nayantara Surendranath | Layout Design: Shivranjana Rathore | Editorial Support: Shivranjana Rathore and Meredith Stinger

Dear Reader,

Welcome to part one of our two-part newsletter wrap-up!

Since August last year, we’ve connected 21 leading experts across technology, AI policy, social science and climate-related sectors to establish a cross-cutting dialogue at the intersection of AI and climate action in Asia. We have pored over countless research articles, policy documents and news items, all in a bid to make sense of this rapidly advancing space.

This process has yielded many learnings along the way. In contextualising existing concerns like AI adoption barriers to climate issues, we’ve found some contradictions. For instance, we’ve tracked emerging R&D on ways AI can be used to reduce emissions growing alongside initiatives on ways to reduce AI’s own emissions.

Our work is reflective of the impacts of global shifts in climate governance and funding on the Asian region, such as the geopolitical tensions shaping critical mineral supply chains and impacts to the climate financing landscape. A key takeaway from our research process was how entangled global systems and supply chains are, making coordinated, equitable climate action a precarious and complex endeavour.

To take on this complexity in the climate tech space, our experts emphasised the need for a whole-of-system lens to understand both the externalities and the suitability of AI approaches for climate action. This requires going beyond traditional data and funding frameworks to think of equitable technological interventions.

In part one, we give you our key takeaways from the past year on what’s needed for responsible climate AI. We propose a Call for Action for those driving the next generation of climate solutions. In part two, we will delve further into promising areas of AI intervention in climate adaptation, mitigation and resilience.

Throughout the series, we’ve seen that while funding ‘solutions’ is essential, it is not the only area that requires investment. We encourage investors to adopt a critical and expansive view of the kinds of financial, community, and solidarity-rooted investments needed to enable systemic change in this space.

Adoption Barriers

The integration of AI in climate-related sectors is complicated by a range of issues - infrastructure limitations, data gaps, lack of data sharing and reuse standards, and inadequate inter-disciplinary expertise.

While Asian countries like China are investing in better infrastructure, intra-regional inequity is creating ‘haves and have-nots’ when it comes to power and resources for compute. Further, diverging from resource scarcity narratives framing AI discourse for the Global South, Cindy Lin suggests that ‘collective restraint’ is already happening by government scientists in Indonesia designing solutions within and in response to local contexts and community needs.

Multi-stakeholder initiatives like the Lacuna Fund have sought to address data gaps by supporting local AI solutions and climate data curation in LMICs; but recent financial constraints threaten the sustainability of some projects. Synthetic data has been proposed as an alternative where data availability is low, but more research is needed to ensure reliability of models trained on such data.

Data stewardship and governance mechanisms are essential for distributing data-driven benefits with communities who provide the data. Regional frameworks centring FAIR and CARE Principles, such as the Asian Data Sovereignty Framework exist, but further efforts are needed to translate these standards into practice across countries.

Moreover, organisations like Climate Change AI are creating communities of practice, but given the need for granular domain expertise across geographies, there is scope for more regional and global collaboration.

Call for Action: Looking beyond policy and tech infrastructure

The focus on the tangible enablers for good climate tech can, however, obfuscate some of the pressing concerns around social and ecological factors that are deep-rooted in prevalent AI-climate trajectories.

AI development and adoption, as Tom Özden-Schilling and Tamara Kneese told us, have material consequences and invisible costs across the supply chain. While conversations on AI’s environmental impacts risk setting a zero-sum discourse on AI’s role in climate action, it is crucial to examine Asia’s unique challenges and develop framings for responsible tech development.

Below is a snapshot of larger systemic issues that need to be resolved to make AI adoption inclusive, effective and intentional. As Asia battles increasing climate impacts, the need for urgent action will spur various decisions at the policy and investment level. In order to avoid reactive approaches to looming climate shocks, we propose the below forward-looking agenda. We call on philanthropies, venture capitalists, family offices, and international organisations to look beyond the hype, and recognise the power you hold in shaping our collective climate futures. These concrete pathways are a starting point in aligning investments in climate tech with resilience, adaptive and justice-focused outcomes.

1. Avoiding technology dependencies and lock-in

Top-down solutions forged from available datasets and based on providers’ assumptions about the climate problem, may erode local knowledge and diminish autonomy of community stakeholders. For example, largescale agri-business apps offering crop advisory and disease detection, alongside fertilisers and pesticides, risk creating information asymmetries, and dependencies while constraining farmer autonomy through product lock-ins. Community-driven interventions that centre local needs and non-tech alternatives are key to avoiding power imbalances between users and technology providers.

2. Bridging misaligned interests

Enabling meaningful participation between myriad stakeholder groups in the climate tech space means addressing divergent, and at times contradictory interests. While companies and developers may be motivated by ROI, researchers want to collect more data, and users may want to see tangible gains. This makes building equitable and sustainable climate solutions a very complex task. Experts emphasised the need for building shared meaning and continuity in climate action work across stakeholder groups, instead of fostering one-time interventions that disrupt community livelihoods. Research across sectors highlighted the importance of ensuring suitable participatory formats. Strategies to account for cultural norms, technical usability, accessibility of language, explainability and transparency mechanisms all impact the shared trust and collaboration needed to implement AI responsibly.

3. Building slow to avoid fail-fast approaches

Investment in participatory AI needs to extend beyond community engagement as a check-list item. Typical Silicon Valley ‘fail-fast’ approaches prioritise rapid deployment, which can have negative impacts on livelihoods and user adoption in the contexts they are deployed in. In the case of agriculture, short-lived start-up prototypes deployed without consideration of local practices and community collaboration can cost farmers an entire season’s yield, and result in distrust and non-adoption by farmers. Intentional investment in dialogue, infrastructure maintenance and capacity-building is essential to ensure the long-term sustainability of projects.

4. Dispelling climate AI hype

In addition to stakeholder capacity-building to understand the technical nuances of AI adoption, it is equally necessary for government officials, community representatives and policymakers to be able to critically assess the limits and suitability of AI for climate action, and frameworks for responsible data sharing and (re)use. Investments in critical AI and data literacy can help bridge knowledge gaps and promote the responsible use of AI.

5. Investing in climate impact reduction, and not efficiency alone

Increased energy efficiency isn’t a silver bullet for sustainability, and in technology development we must ask, what becomes invisible under such a narrow lens? An overt focus on the energy efficiency of generative AI development, for example, prioritises increased productivity at the expense of monitoring CO2 levels and water consumption and broader environmental indicators. Accelerated hardware advancement can result in increased e-waste and inaccessible costs can encourage sustained use of energy-intensive hardware. A critical climate lens that challenges techno-solutionism, and asks - how does this solution improve adaptation, mitigation or resilience? - during project initiation and design is important to ensure AI’s benefits outweigh its costs in addressing climate change.

6. Shifting away from resource-intensive growth trajectories

From EVs to AI, we’ve seen the need to decouple economic growth from resource-intensive growth trajectories and better assess AI’s full human and ecological impacts. Research has shown the increasing parallels between the AI supply chain and “green-extractivism” in which mineral supply chains reinforce historical inequities tied to colonisation. Despite promises of economic diversification and growth, labour concerns across this supply chain range from environmental impacts of mining and refining on community livelihoods, exploitative working conditions, and precarious and limited data centre jobs. A whole-of-system approach to climate tech investments, including support for R&D, and collaboration in developing frameworks for assessing AI’s social, economic and environmental impacts across the supply chain are necessary.

Around the Web

• Check out these Climate Change AI Interactive Summaries of core learnings from their paper “Tackling Climate Change with Machine Learning,” in which they highlight high impact applications for AI in climate.

• The AI & Environment Resource Hub is a curated collection of knowledge, tools, events and insights at the intersection of AI and sustainability.

• For visual guides and discussions on the planetary impacts of the AI supply chain, drop by the AI + Planetary Justice Alliance.

A huge, warm thank you to all Code Green readers and listeners who have tuned in this past year. Your engagement has been encouraging in a landscape where we urgently need more voices. We hope you’ve learned as much throughout this series as we have!

Access the entire series on codegreen.asia

Share this with your friends who you think may be interested, or if you see any dots connecting from your work for a potential collaboration, write to us at codegreen@digitalfutureslab.in

To continue to remain in touch with our work across climate, gender, labour, foresight, subscribe to DFL’s newsletter.

Credits

Writing: Dona Mathew, Meredith Stinger | Illustrations: Nayantara Surendranath | Layout Design and Editorial Support: Shivranjana Rathore

Dear Reader,

In May this year, the Indigenous Environment Network called for a moratorium on all false solutions for the climate crisis. A powerful wake-up call, they point to a troubling truth: over the last 20 years, green economy approaches that hinge on carbon markets, pricing, and offset mechanisms have proven ineffective in reducing emissions. With decades of not moving the needle, we must examine what isn’t working. Luis Murillo, in our latest podcast episode, points to the “fundamental tension” embedded in the financialisation of the climate space, “If you have capital as the engine for that space, and capital is at the source of environmental degradation, historically speaking, that led us to where we are right now.”

In this final dispatch of Code Green, we zoom in on a common thread across our series—in an increasingly tech-centric climate action discourse, what big picture are we missing? Throughout, we’ve worked to identify the potential and challenges for pathways towards safe and trustworthy Climate AI. And yet, what futures are being wrought by dominant narratives, politics, and powerful interests steering the climate-tech innovation and investment space?

Achieving net-zero in Asia will require an estimated $70 trillion investment between 2023 and 2060. However, geopolitical shifts are creating uncertainties in the climate finance landscape, with far-reaching consequences for vulnerable regions. The United States has cancelled $4 billion in funding for the UN Green Climate Fund, withdrawn a 2023 $17.6 million pledge from the UN loss and damage fund, and exited the Just Energy Transition Partnership, a multilateral platform meant to accelerate the phase-out of fossil fuels in South Africa and Asia.

According to a UNESCAP report, few countries across Asia are on track to meet GHG emissions targets. Funding remains highly concentrated in countries like China, India, and Japan. Least Developed Countries (LDCs) in the region received only 2% of international climate finance from 2021 to 2022. Beyond allocation issues, equity of fund distribution is a concern, with only 51% of Southeast Asia’s adaptation funding distributed from 2017 to 2021.

Within this complex ecosystem, investments in AI are expected to reach $110 billion by 2028. However, unequal access to AI tools, data, and resources between high-income and low-income regions poses a significant risk of deepening existing inequalities. According to the 2021 GPAI report, poorly-directed funding could lead to AI’s rapid growth becoming a distraction or a cause of societal harm. How much of this focus on AI is causing harm or distracting from essential climate funding needs?

The AI-climate intersection must avoid exacerbating inequities, particularly as attention turns to "silver bullet" solutions that may prove economically or socially unviable. Amidst intensifying climate impacts, we ask what structural changes are required to enable equitable, inclusive, and sustainable futures?

Curated Reads

Innovative mechanisms like green patents are emerging as signposts for private financing of green technologies. However, in climate-vulnerable contexts in the Global South, bottlenecks such as digital infrastructure gaps and limited data quality and availability may create additional barriers to access competitive funding sources, such as Grand Challenges. However, is a focus on financing, innovation, and efficiency of Climate AI enough to address the climate crisis?

Drivers of innovation and investment

In our analysis of the AI and climate financing domain, we came across factors that seem to be driving the adoption of AI solutions for climate on one hand and pushing for more investment in AI solutions for climate on the other.

We’ve seen a growing focus on climate solutions in AI open calls more broadly. For example, this year’s AI Action Summit announced a call for AI projects for well-being and ethical use, selecting 50 projects, of which 9 were related to climate and nature. Of these, 7 were from countries in the Global South.

In a recent assessment of grand challenges for AI, climate, and nature, the authors highlight the value of these funding models to centre climate challenges, using AI as a means to solve them, with sometimes the final output not using AI at all. However, they note that a majority of GCs emerge from the Global North, and recipients are also skewed to this region. They also note that GCs are not the only way to fund climate issues, calling for a blend of funding and open innovation models like grants, incentives, hackathons, and incubators, depending on the stage of innovation.

As we’ve highlighted in previous issues, Green AI is gaining traction as a way to centre resource efficiency in AI model development, and a combination of patenting green technologies and open-source could potentially drive private investment in this area.

Researchers are pointing to the environmental and technological potential of TinyML—ultra-low-power AI, exploring how combining patent protection with open-source collaboration can accelerate innovation in green AI. Patents offer competitive and economic advantages, with recent research from the UK and EU suggesting that they may boost VC funding probability by up to 20%, while open source fosters transparency, community input, and rapid adoption. The synergy of both is expected to drive efficient, scalable climate solutions, but findings vary by geography and investor type, cautioning against overgeneralisation.

Models like TinyReptile, compression algorithms, and benchmarking frameworks demonstrate practical progress in less resource-intensive AI. Cross-sector collaboration, talent development, open patenting, and standardisation are essential for advancing TinyML in environmental governance.

But what’s the cost of climate-tech innovation?

Global trends, with possible impacts in Asia, indicate that there is impending investment and innovation potential in climate-tech. As Hanyuan Wang notes in our new podcast episode, climate schools are emerging across the region, incubating the next generation of tech innovators. While these trends are encouraging in the context of developing solutions for the ongoing climate crisis, how impactful will this reliance on the promises of climate-tech innovation actually be?

Current discourse on climate-tech applications indicates a strong flow of ideas from the Global North to the Global South, risking technology transfers that are not contextualised to the needs of climate-vulnerable countries. For instance, carbon removal technologies have been gaining prominence in climate strategy literature. However, knowledge building is primarily happening in the Global North. The author of this study writes that carbon removal technologies must be integrated into sociotechnical systems with attention to justice, governance, and equity, particularly for the Global South. Historically marginalised, these regions face risks from exploitation tied to carbon offset markets, extractive industries, and weak regulatory capacity. Procedural equity and inclusive governance are vital to ensure fair participation in technology design and deployment. Yet, climate assessments and funding remain skewed towards Global North priorities. There is a pressing need for collaborative research, co-designed innovation, and capacity-building in the Global South to align carbon removal with local needs and uphold the Paris Agreement’s equity principles, especially CBDR-RC (Common but Differentiated Responsibilities and Respective Capacities).

For advanced technologies like AI, even as community-led, participatory climate solutions are prioritised, and funding is directed towards building the necessary data and digital infrastructures for eco-innovation, what ends do these climate solutions meet?

Curbing techno-optimism, nurturing systemic change

Climate technology entrepreneurship is essential, but on its own, insufficient to address climate change and the excessive demands on natural ecosystems that far outpace regenerative capacities. An over-reliance on technological solutions may delay necessary systemic changes, especially since their development uses more resources than efficiency gains can offset. We’ve referenced Jevons Paradox before—where increased efficiency inevitably leads to greater resource use—a concept more and more relevant to AI growth conversations. Discourses on planetary boundaries are missing from current entrepreneurial ecosystems and mission-oriented innovation policies, which are obsessed with competition and growth. Research indicates a positive correlation between countries that rank higher in digital entrepreneurship and energy consumption—they are often the biggest contributors to climate change.

Given the urgency of the climate crisis and the focus on digital interventions, entrepreneurs in Asian countries need supportive systems that provide capacity building and training in climate issues, finance to underwrite solutions that minimise resource and material dependencies, and policies to balance economic development with climate action.

Structural changes to climate action policies are integral to decouple resource-intensive technology innovation from economic growth. This study examines techno-optimism in international climate policy through two prominent climate-tech solutions: EVs and Bio-energy and Carbon Capture and Storage (BECCS). EVs, though promoted as green alternatives, rely on the extraction of raw materials like lithium, cobalt, and nickel, often sourced from Africa and South America under environmentally-destructive and exploitative labour conditions. Similarly, BECCS, though marketed as a mitigation tool, demands an estimated 300–600 million hectares for biomass production, raising serious concerns about land rights, biodiversity loss, and competition with food production.

These technologies maintain current unsustainable models, such as car dependency and industrial emissions, rather than challenging their foundations. Reliance on such innovation delays immediate action and misplaces hope in future breakthroughs. Small Island Developing States (SIDS) and other Global South actors advocate for integrating traditional ecological knowledge, community-based resource management, and stronger corporate accountability into international law. Moving beyond growth-centred development models, a more ecological and post-growth approach grounded in procedural equity, local participation, and empowerment is urgently needed to ensure that global climate governance reflects both environmental limits and social justice.

Planter Box

Post-Growth, Degrowth, and AI Minimisation

In our episode on Agriculture, Anubha Singh cautioned us against the harmful effects of fail-fast approaches, often used by tech startups to target farming communities. As the climate crisis deepens and the urgency to find solutions drives rapid growth and innovation, Michael Kwet’s call for digital de-growth feels more pertinent than ever. Responding to US climate and energy policy changes and a ramp-up of global tensions regarding AI dominance, a recent publication from the University of Virginia calls for a global governance shift to AI minimisation, encouraging us to ask, “How much of the solution is also the problem?” The author suggests that moving beyond sustainability metrics susceptible to greenwashing may mean flipping the narrative of technological progress from unlimited growth to minimisation altogether, offering 5 starting points for this shift:

1. Decrease AI Adoption: AI must be seen as a tool entangled in broader ethical, digital rights, and environmental issues.

2. Fossil-free AI Infrastructure: AI services and energy sources must decouple from fossil fuels and shift to renewables.

3. Destruction Assessment: AI must only be used when the benefits outweigh the destruction it causes.

4. Sovereignty-Focused: Minimisation agendas must promote mineral, technological, geopolitical, and Indigenous sovereignty.

5. Non-Western perspectives for the AI Agenda: Advancing the AI agenda must come from collaborative knowledge production with non-Western perspectives.

Around the Web

• A photo essay documenting the historic Silk Road in this current age of climate change and Digital Silk Road initiatives.

• Code Green podcast guest Cathy Richards recently published a Community Data Playbook of environmental data case studies and strategies for data stewards.

• Observatory of Planetary Justice Impacts of AI collects, documents, and analyses real-world cases of AI’s impacts on planetary justice throughout the entire AI supply chain.

ICYMI

In our final podcast episode, Hanyuan ‘Karen’ Wang and Luis Felipe R Murillo examine the benefits and shortcomings of AI-driven climate solutions, emphasising the importance of research and funding partnerships, community data stewardships, and recognition of AI’s negative environmental impacts.

Credits

Research and curation: Dona Mathew, Meredith Stinger, Tammanna Aurora | Illustrations: Nayantara Surendranath | Art Direction: Tammanna Aurora | Layout Design: Shivranjana Rathore

Dear Reader,

Asia's cities are vibrant, dynamic, and often full of contrasts. Images of Tokyo and Taipei, for instance, show shiny skyscrapers and high-speed rail networks co-existing with ancient temples and historical landmarks. However, these urban landscapes and their residents are increasingly under threat from the rapidly accelerating impacts of climate change — rising sea levels, urban heat island effect, and air pollution jeopardise livelihoods, biodiversity, public health, and food security.

Fast-paced urban migration is also placing significant pressure on urban governance. By 2030, it is projected that around 55% of Asia’s population will live in urban areas. Many of these challenges are complex and interlinked. Deforestation driven by urban expansion reduces vital carbon sinks, worsening climate change and degrading biodiversity. This, in turn, leads to lost ecosystem services and reduced resilience to further climate impacts.

As governments across the region prioritise the creation of climate-resilient and sustainable cities, digitalisation is emerging as a key driver in national urban strategies. The concept of 'smart cities' is central to this discourse, with advances in AI applications emerging in areas such as waste management and citizen engagement. While these innovations pave the way for more technology-driven cities, they also foster new forms of urban governance, characterised by the increasing corporatisation of city management. AI is also being leveraged to meet Sustainable Development Goal (SDG) 11 for cities and communities. A review highlights that while AI enables efficient service delivery and sustainable systems, resource demands and data biases challenge equitable, green, and inclusive development.

Modern cities are intricate webs of interconnected domains — energy, transport, health, education, law enforcement, and sanitation — each of which must function effectively to enhance the well-being of residents. Given the dual challenges of climate change and rapid urban expansion, the question remains: How sustainable will the cities of the future truly be?

Curated Reads

The ‘smart city’ and the ‘sustainable city’ have emerged as distinctive concepts envisioning pathways to green, citizen-centric, urban futures. But how much do these concepts overlap in practice? As AI becomes a central tool in transforming urban landscapes and governance, we examine the potential and contradictions of smart, data-driven, and technology-focused interventions to create sustainable future cities.

The Smart City: New Technologies, Old Dilemmas?

Smart cities utilise interconnected IoT sensors to collect, share, and analyse real-time data for urban planning and governance, from monitoring air quality to traffic management. The smart city concept first emerged in the late 2000s as a form of corporate storytelling influenced by IBM’s Smarter Cities Challenge, and has since become a mainstay of state-led public policy and urban planning. Iterations have evolved over a decade, moving from tech-centred to human-centred conceptions in global discourse. A vibrant and well-developed critique of the numerous inequities these cities of the future may perpetuate and create has emerged alongside this discourse.

The cities of today are ideal receptacles for large flows of different types of capital: financial, physical, human, social, and natural. These authors examine smart cities as a spatial manifestation of 21^st-century capitalism, characterised by the accumulation of capital in the hands of a few, complex public-private partnerships for governance and resource extraction, and ubiquitous harvesting, storage, and analysis of data from the natural and physical environment as well as from human behaviour and interactions.

A study analysing smart city mission policy documents in India reveals how national strategies often borrow heavily from global rhetoric on smart cities. These documents tend to adopt a top-down approach to urban development, using prescriptive language and abstract imagery. Citizens in these documents are represented by interchangeable stick figures, illustrating a reductionist view of urban life. Furthermore, the concept of citizenship is often tied to accountability, entrepreneurship, productivity, and business improvement, while overlooking the complex, multi-dimensional realities of urban living. In these narratives, citizens are generally portrayed as a homogenous group, expected to be both entrepreneurial and tech-savvy, reflecting an idealised vision of the smart city.

In a separate paper, authors exploring the idea of just smart cities in the Global South highlight the lack of theoretical frameworks addressing the unique challenges of these regions, emphasising the need to explore what 'the right to the smart city' means in this context. They build on the framework of 'expose, propose, and politicise' to suggest a research agenda for scholars to move beyond the current case study approach that dominates smart city literature. The authors argue that research should begin with ‘expose’ – examining who is excluded or marginalised by smart city initiatives, moving towards ‘propose’ by working with those affected to advocate for alternate visions of the smart city, and finally, ‘politicising’ these as the fundamental absence of rights which must be addressed. They call for more action-oriented research that focuses on the grassroots resistance led by citizens, NGOs, and community data intermediaries who are working on the front lines to challenge the status quo.

The Smart City — Smart Data Nexus

Data-driven planning and governance are key aspects of the promise of smart cities to surface and address residents’ needs. OECD’s report on smart city governance highlights initiatives that centralise and visualise urban data to produce real-time inputs for city planning, such as Seoul’s Open Data Plaza and Japan’s Sketch Lab. However, a lack of financial resources, data management skills, data security, and fragmented datasets across the government and private sectors hinders effective data use and sharing. In Thailand, citizen-facing apps like Traffy Fondue allow reporting by citizens on traffic conditions, and could help urban planners and local authorities address traffic issues in real-time. However, research shows that a lack of skill and capacity in local government leads to under-utilisation of data for decision-making, with data primarily being used for documentation and reporting. To foster greater transparency and empower citizens, upskilling government officials and providing data literacy programmes for citizens are critical.

According to this report, data representativeness and data quality are key concerns in equitable urban planning, and may reinforce biases and exacerbate existing social inequalities. But increased data representation isn’t a straightforward solution for those living in informal settlements, who may face increased risks of displacement by being visibilised through new forms of data collection. In cities like Hyderabad, India, for those living in informal settlements, fears of displacement have become attached to digital data collection processes. As digital technologies are more widely employed for urban governance, claims to commodified urban spaces have become more precarious, along with growing distrust in data collection processes.

This author identifies ‘data dignity’ as fundamental to smart city technologies — the notion that individuals maintain sovereignty and self-determination over information linked to their identities. Advocating for the ‘recoding’ of the smart city stack (consisting of networked sensors, data platforms, AI/ML models, and automated control systems), a system of open standards, decentralised cloud infrastructures, and community-governed oversight would rein in unchecked extraction and commodification of citizen data.

Code-Smart or Climate-Smart?

SDG 11 focuses on safely and sustainably supporting expanding urban populations through smart, resilient, and culturally-informed urban planning.

In a review of literature at the intersection of SDG-11 and smart city projects, authors argue that while environmental sustainability is a necessary aspect of achieving broader goals highlighted in SDG-11, social and economic sustainability are often prioritised over environmental sustainability in smart city initiatives. Potential externalities and environmental impacts, such as increased energy demands, are underexamined. Moreover, they highlight how critical smart city literature often speaks to environmental and social considerations in these projects in a mutually exclusive way, recommending that future research embrace an expanded sustainability framework that includes both social and environmental aspects together.

A key target (11.6) under SDG-11 is to reduce the adverse per capita environmental impact of cities, focusing on air quality, and municipal and other waste management. However, while the use of ICTs in cities has potential for emission reductions, there is a danger of a rebound effect of efficiency gains, in which expanded ICT use will drive up emissions and material impacts, curtailing environmental gains. These authors argue that this demands an expanded lifecycle approach to understand the sustainability trade-offs in smart cities. In a lifecycle analysis of smart cities and ICT development, authors highlight a gap in research empirically demonstrating increased environmental sustainability in smart cities. Rather than overtly focusing on energy efficiency, optimising transport, and waste management through technology, they recommend a turn towards low-tech urban solutions and a holistic approach that incorporates the social, economic, and environmental dimensions of urban development in smart city planning.

Towards equitable urban AI applications

Recognising a lack of sector-specific ethical frameworks for AI in urban planning, this research proposed 14 principles to guide the development of Urban AI applications. These principles address concerns like human agency, data privacy, transparency, fairness, and algorithmic bias, all crucial for equitable AI integration in cities. The framework emphasises the importance of inclusivity, ensuring AI advancements benefit marginalised communities and reduce inequalities. It also calls for adaptive governance to keep pace with rapid technological change and for ongoing interdisciplinary collaboration to address urban-specific challenges. While offering a foundational structure, the authors encourage further research and adaptation, especially concerning the socio-political, spatial, and cultural dimensions of urban AI. Future studies should focus on refining these principles, ensuring ethical AI deployment that enhances urban living while mitigating risks.

Spotlight: Multispecies Justice

How are bees experiencing the city? This is one of many questions municipalities and communities in Costa Rica’s ‘Sweet City’ are asking as they explore ways to prioritise biodiversity, multispecies interactions, and the collective design of socio-ecological urban spaces. Reflecting a growing “more-than-human” turn in nature-based solutions (NbS), emerging scholarship calls for multispecies justice, recognising the needs of humans, animals, rivers, and ecosystems while equitably distributing risks and benefits across social, spatial, economic, and environmental dimensions.

Moving beyond human-centric urban planning, a de-anthropocentric NbS approach promotes coexistence, ecosystem integration, climate education, inclusive governance, and green innovation. Examples like Singapore’s Binhai Bay Garden City, with its intelligent irrigation and real-time environmental monitoring, highlight how technology can support both human and ecological well-being.

AI can potentially aid in scaling NbS through tools like AI-generated visions of future nature-positive cities, but challenges remain. Limited data quality, opaque algorithms, and overreliance on quantitative outputs risk sidelining indigenous knowledge, social values, and transparency in decision-making. As cities in Asia grow, balancing technological advancement with ecological and cultural sensitivity is essential.

Around the Web

• Feral Atlas: Explore the connections between human-made infrastructure and the ecological transformations it affects.

• Multispecies Cities: Solarpunk Urban Futures: A collection of stories that examine our relationship with the more-than-human. (Solarpunk is a science fiction subgenre).

• City-Game Database: Try your hand at city-making and urban planning.

• Data Justice in Practice: A Guide for Impacted Communities: Actionable information for communities to implement data justice principles and practices.

• The great abandonment: what happens to the natural world when people disappear? Read on to find out an under-explored impact of urban migration.

Tuning in…

In the latest episode of Code Green, ChengHe Guan and Raj Cherubal examine the role of data and AI for creating smart and sustainable cities in Asia, sharing their distinct experiences from Shanghai and Chennai. A sneak peek here!

Note to readers!

This is the second-to-last issue of our newsletter! We've loved curating these monthly dispatches for you. As we plan the next chapter and look for ways to keep the Code Green community thriving, we’d love your input: What articles or topics resonated with you? What would you like to see more of? Drop us a note at codegreen@digitalfutureslab.in — your ideas might just inspire what’s next!

Credits

Research and curation: Dona Mathew, Meredith Stinger, Tammanna Aurora | Illustrations: Nayantara Surendranath | Art Direction: Tammanna Aurora | Layout Design: Shivranjana Rathore

Dear Reader,

Transportation is Asia’s second largest source of emissions, representing 40% of global transport emissions, equivalent to that of Europe and North America combined. This is primarily linked to fossil-fuel dependence in the region, although countries have taken many steps to decarbonise the sector and electrify fleets.

As efforts are made to combat climate change and decarbonise transportation, concerns around equity and user-friendly mobility complicate policymaking in urban transport. To address some of these concerns, leaders from 21 Asian countries signed the Aichi 2030 Declaration for achieving universally accessible, safe, affordable, low-carbon, multi-modal transport. One of the strategies in the Declaration is the adoption of smart information and communication technologies like intelligent transport systems (ITS). In these systems, machine learning (ML) is used to optimise traffic flow and navigation, minimise accidents, enhance driver experience, and in autonomous driving. But safety and privacy concerns persist, and other challenges like lack of public benchmark datasets, explainability gaps, difficulties in identifying multi-modal transport, and infrastructure limitations pose adoption barriers.

In this issue, we explore how advanced technologies like AI are being deployed for mobility in Asia. What goals are being optimised for? Is the pursuit of efficiency in the sector leading to sustainable outcomes? How are access and connectivity incorporated in AI policies for mobility?

Curated Reads

AI use in the transport sector

As commuters move from one place to another, they generate a lot of data, which can feed into AI systems. This includes location data from mobile devices, data from traffic sensors, and frequency of travel from ride-sharing apps. Dynamic GPS data, in particular, supports transportation demand forecasting, route optimisation, and traffic and safety management. However, GPS data alone lacks important context and cannot directly identify location names or transport modes, making accurate transport mode detection challenging. Traditional methods to differentiate between cycling or car travel rely on predefined routes or traffic patterns, limiting adaptability to real-world conditions such as unexpected traffic jams.

This proposed approach eliminates fixed time intervals by extracting features directly from raw GPS data and converting them into images for training a deep learning model. Using GeoLife, a five-year GPS trajectory dataset from Microsoft Asia, researchers first extract features like velocity and acceleration to generate images. Then, the deep learning model divides these images into patches to identify transport modes such as walking, biking, and driving. While achieving 92.96% accuracy, misclassification of certain modes, like cars, and unbalanced training data highlight the need for further feature analysis and improvement.

Autonomous Vehicles (AVs) in Asia

While driverless vehicles have long promised safety and sustainability, technological, regulatory and cost barriers have slowed down widespread deployment. But researchers encourage us to look beyond current AV discourse that is shaped by technological deterministic narratives and a narrow focus on ‘trolley problem’ ethical concerns to consider how AVs are deeply political, intertwined with urban environments, labour, infrastructure, and mobility.

Mostly emerging in high-income countries in Asia, such as Singapore’s Grab and Moovita, South Korea’s A2Z and China’s ApolloGo and Pony.ai robotaxis, AV rollout is still limited in size and scope due to requirements for data governance, access, infrastructure, and public acceptance. In contrast to usual goals for optimisation and efficiency, for Singapore, the highest ranking in AV readiness, AVs have become a strategic investment to address land scarcity and labour shortages across transport and delivery sectors. In other parts of Asia, AV infrastructure requirements may create inequities. Some challenges for deployment include the high cost and complexity of software and equipment, and reliance on high-speed network connection and urban communication infrastructure – all of which have an impact on safety. In South and Southeast Asia, AVs also have to account for dynamic and chaotic road conditions and transport modes like tuk-tuks, jeepneys, and rickshaws. This suggests that AV use may be restricted to areas with predictable road conditions that can afford to build and maintain this infrastructure, bringing up questions of spatial equity in urban mobility access.

In rapidly developing urban contexts, responsible steering of AVs towards shared transport models could transform mobility ecosystems, but this requires long-term, localised, multi-stakeholder planning. In case studies on potential governance approaches to multi-modal AV deployment in Ho Chi Minh City and Bangkok, authors present a comprehensive framework on how different policy pathways adapted to unique urban contexts could shape AV implementation. They argue that without regulatory guidance, AV deployment in the region could exacerbate urban sprawl, congestion, heightened emissions, and socio-spatial inequities. They recommend context-specific participatory planning between local and regional governments, communities, and mobility actors.

Mobility data justice

Modern mobility increasingly involves the generation, storage, and sharing of data, raising pertinent questions: What inequalities arise when mobility intersects with datafication? Whose movement is captured or overlooked in data collection and sharing, and what factors influence these inclusions and exclusions?

Researchers highlight ‘data and information asymmetry’ between the public and private sectors, where governments provide open transport data while businesses use it to develop and profit from mobility services. They note tensions emerge between ‘the need to be seen and represented’ versus ‘the need for autonomy and integrity’ – for example, data from app-based transport represents those who can afford it, rendering other commuters invisible.

To challenge discrimination in mobility data, they develop a mobility data justice framework, centered on:

• Distributive justice: How access to mobility and data is unequally distributed, who benefits or is harmed, and how marginalised groups reclaim control through counter-mapping and commons-based approaches.

• Procedural justice: Who designs and controls mobility data infrastructures, highlighting the exclusionary role of AI and governance systems.

• Epistemic justice: What counts as mobility-related knowledge, how data systems shape inequalities, and how alternative ways of knowing can promote fairness.

Electric Vehicles & the Nickel Pickle

AI is expected to play a crucial role in electric vehicles (EV) - from optimising energy management to predictive maintenance. But like our last issue, we find ourselves in a similar pickle when looking at the EV battery supply chain - a case of carbon-intensive processing with outsized human impacts. While most battery manufacturing happens in Japan, South Korea and China, in Southeast Asia, Vietnam and Indonesia are emerging as battery hubs, due to their nickel reserves, a key input for battery cathodes.

Indonesia’s reliance on coal-powered smelting for nickel production has made them one of the highest emitters amongst producers globally. Difficulties in disposing of industrial waste have also halted deep sea mining waste disposal. While Variable Renewable Tech and Hydro/Geothermal power are alternatives, high electricity requirements and the remote locations of Indonesia’s mining sites remain barriers to meeting heightened battery supply chain standards.

In reports from CRI and Mighty Earth, Indonesian communities surrounding the Weda Bay Industrial Park face unlawful land dispossession, unfair compensation, and disruption to indigenous livelihoods. Wildlife and crops have been impacted, and water and air pollution are causing health impacts. There is also rapid deforestation due to expanded mining activities, including in protected areas. Both reports stress the importance of private and public commitment to transparency in EV battery supply chains, stronger observation of Indonesian laws protecting indigenous lands and protected forests, and adequate compensation for those experiencing harms.

Advancing sustainable and equitable transport futures

Data justice issues, inequitable access, and compounding ethical and environmental harms complicate AI-enabled green mobility transitions in the region. In a recent study, authors argue that current sustainability efforts in the sector are insufficient as they focus on the least effective action - improving mobility efficiency. As a complex system embedded within the broader urban environment, urban mobility requires a systems-thinking approach, advanced by:

• Cross-disciplinary collaboration between engineers, planners, economists, and behavioral scientists.

• Simulation-based approaches using reusable models and datasets to predict and evaluate intervention outcomes.

• Transparency and ethical research addressing privacy concerns, unequal data access, quantifying environmental costs, and abandoning projects that cause harms.

• Public engagement by researchers to fact-check misinformation and communicate findings quickly to support evidence-based policymaking.

On our Radar: Mobility-as-a-Service (MaaS)

MaaS refers to the integration of public transit, ride-hailing, bike-sharing, and car rentals into one digital service. Popularised since 2014 as a flexible, sustainable, user-friendly mobility solution, it is still untested in localised contexts, and a lack of empirical studies and real-world applications makes it difficult to validate its benefits.

MaaS has found some success in Japan, as researchers found that a MaaS-AI on-demand shuttle significantly improved the elderly’s access to healthcare and shopping activities based on place (geographical distances) and person-based (individual’s unique behaviour) factors. However, what may work in developed economies may not be easily transferable to all contexts - insufficient physical infrastructure, entrenched institutional frameworks, lack of data standards, and fragmented operator groups pose challenges. In developing contexts, government-led rather than private models may be more suitable to ensure equitable and sustainable implementation.

Further, these authors warn that if AI-enabled MaaS relies solely on digital data reflecting real-world gender or age-based mobility access issues, it may exacerbate these inequalities. AI recommender algorithms could also prioritise convenience over equity, favouring cars over active and environmentally responsible modes like cycling. The integration of multiple data sources by private actors offering AI services poses a risk of corporate power concentration and standardised solutions. To address this, a collaborative governance model for MaaS, centered on sustainability, transparency, and accountability while balancing public interest and innovation, can foster equity in access.

MaaS as a concept is also evolving. In this vision paper, researchers envision Mobility-as-a-Resource (MaaR) - a shared resource managed by technology. Data science is essential to MaaR - providing cross-disciplinary insights and integrating knowledge from travel behaviour, traffic flow, and network science to support equitable mobility. But it is imperative that MaaR is not reliant on market-based systems alone - its implementation must be carefully managed with new policy tools and regulatory frameworks to account for digital divides and mobility inequalities, and to reduce environmental harms.

Around the Web

• Slime mould grows network just like Tokyo Rail System: A fascinating peek into how single-celled slime mould construct nutrient-channeling tubes strikingly similar to Tokyo’s railway networks.

• The atlas of sustainable city transport: Curious about sustainable transport in your city? Explore this atlas to know how close you are to public transit.

• Grab built its own map in Southeast Asia, and is now going after Google: A deep-dive into how Grab used their drivers and cameras to create hyperlocal maps in Southeast Asia.

• Transport-Our World in Data: A visualisation of global transport patterns and their environmental impacts.

Tuning in…

In the latest episode of Code Green, Huê-Tâm Jamme and Kris Villanueva-Libunao discuss mobility transitions in Asia and AI’s role in the sector. Together, they advocate for a whole-of-systems approach to mobility infrastructure, centering people’s needs and experiences in planning and policy-making for the future.

Credits

Research and curation: Dona Mathew, Meredith Stinger, Tammanna Aurora | Illustrations: Nayantara Surendranath | Art Direction: Tammanna Aurora | Layout Design: Shivranjana Rathore

In Part 2, we look at the material aspects of mineral extraction, manufacturing, and waste generation as part of the AI lifecycle. Read Part 1 here.

Dear Reader,

In a 2024 report on critical energy transition minerals, the UN Secretary-General António Guterres said, "A world powered by renewables is a world hungry for critical minerals. For developing countries, critical minerals are a critical opportunity – to create jobs, diversify economies, and dramatically boost revenues. But only if they are managed properly”.

Renewable energy components like photovoltaic solar panels, wind turbines, and lithium battery storage, while key to decarbonising the energy sector, have their own resource-intensive lifecycles that begin with mineral extraction and end with waste. Projections show that by 2040, the extraction of critical minerals and rare earth elements (REE) will have to quadruple to meet climate goals. Developments around the US-Ukraine minerals deal indicate that the future of war and peace may hinge on the ability of nations to shore up their mineral reserves.

From the frenzied demand for lithium for EV batteries, the metal mining panic has turned to the minerals required for AI’s energy demands. In 2024, commodity trader Trafigura estimated that data centres to power AI will likely require an additional 1 million metric tons of copper by 2030. Amidst tech players like Google pledging to be 100% powered by clean energy, the rush for critical minerals is on.

But who bears the costs of these transitions? The drive for minerals reveals unequal geopolitical dynamics as Global South countries become resource frontiers for extraction and refining. Mining has its own environmental, social and health risks, such as water pollution, industrial waste, worker exposure to toxic chemical processes, and biodiversity loss, impacting indigenous lands and communities. To meet mineral demand, initiatives for climate-smart mining have emerged - but critics warn that this is simply greenwashing masking continued harms of extractivism.

Given broader narratives around the geopolitics of mineral supply chains as well as the long-term environmental and local impacts of extraction and processing, policymakers and investors must consider the connection between AI’s growth and its reliance on critical minerals, instead of leaving it to market forces.

Curated Reads

We trace geopolitical challenges and the environmental and social implications of the mineral supply chain, AI for responsible and sustainable mining, and highlight potential pathways for environmental justice.

Paradoxical paths

The dual nature of AI and critical minerals - its potential to improve mining efficiency alongside its reliance on accelerated mineral extraction - presents a paradox. Net-zero means AI development depends on the clean energy transition, which is further driving the rush for critical minerals. This leaves ecological and social destruction in its wake, standing in tension with its potential role in optimising mineral extraction and processing.

In our research, we found that much of the literature urges increased investment in AI mining technologies and research, with efficiency as the bottom line. For example, to address funding barriers to meet mineral demands for the green transition, researchers in Australia found that the use of AI can help shorten the duration of mining projects, map mineral reserves, and reduce associated disaster and environmental hazards. However, they also note that AI’s potential in the mining sector is deeply linked with upskilling workers to use AI, filling acute gaps in databases to train and test AI models, and challenges in storing and transmitting data from remote and often inaccessible mining locations.

“Not in my backyard” politics

In the mineral production supply chain, the outsourcing of harmful social and environmental effects is characterised as an offloading of risk, a type of ‘not-in-my-backyard’ politics of industrialised economies. While nations (historically the Global North) fuel escalating demand, they shift production to countries (often in the Global South) with lower environmental standards, and labour and energy regulations, and hence, cheaper costs. In efforts to capture economic gains, these countries often do not have the luxury of prioritising social or environmental concerns.

Take, for example, the semiconductor industry which is driving demand for rare earth elements like gallium and germanium. When China, which accounts for 98% of the refined gallium and 68% of germanium production globally, imposed export restrictions in 2023, there were global impacts on supply chains, escalating already fever-pitch tensions between the US and China.

When supply chains are threatened, countries search elsewhere, typically where costs are low. Disparities within the Asian region also enable inequitable distribution of benefits and risks, as countries move damaging mining and processing intra-regionally to meet low-carbon economy goals.

China has led Rare Earth Element (REE) production since the 1980s, influenced in part by stronger environmental regulations in developed economies. REE processing involves highly toxic chemicals like ammonium sulfate and chloride to separate minerals. Now, as China’s domestic environmental standards are improving with clean-up projects underway, its activities have moved to less regulated zones like Myanmar’s Kachin State, where water contamination has impacted food security and presented health risks for workers and local communities.

Similarly, China now invests in bauxite smelting facilities in Indonesia which run on coal, making them energy-intensive and environmentally dirty, but less expensive.

Long-term environmental impacts

Salvador Regilme writes about ‘necro exportation’- how the Global North, which reaps AI’s benefits, transfers the environmental costs to the Global South, leading to ecosystem and community damage through resource extraction and energy consumption.

Historically, extraction and refining processes have endangered biodiversity hotspots, creating health risks and polluting water sources. For example, nickel mining is endangering the biodiversity-rich Sibuyan Islands in the Philippines. In Indonesia, producer of 1/4th of the world’s nickel supply, extraction and the construction of industrial parks for smelting have resulted in flooding, deforestation, and the loss of indigenous livelihoods.

The supercycles of mineral supply booms and busts also leave mining communities vulnerable to ‘legacy issues’ of mining: abandoned and derelict infrastructures, open pits and residual industrial waste, which haunt communities with soil, water, and air contamination. One of the UN Secretary General’s Actionable Recommendations includes a proposal for a Global Mining Legacy Fund to clean up these mines and to strengthen financial assurance mechanisms for mine closure and rehabilitation.

Locating place and identity

Research by the International Renewable Energy Agency (IRENA) indicates that 54% of energy transition minerals are situated on or adjacent to Indigenous territories, highlighting critical intersections between resource extraction and Indigenous rights.

In their comprehensive mapping of over 5,000 mineral projects critical to energy transitions, researchers in Australia and Indonesia found that strategic minerals like lithium and manganese have the highest spatial correlation with Indigenous territories. To ensure that the push for renewable energy does not lead to displacement and disruption for communities, they suggest improved sustainability assessments, stronger free, prior and informed consent processes, and the promotion of circular economy solutions to reduce reliance on newly mined critical minerals.

Mining also poses sociocultural risks to Indigenous cultural heritage through environmental degradation and social disruption of land-connected rites and traditions. In Papua New Guinea, proposed deep-sea mining threatens communities' access to traditional marine spaces integral to their spiritual practices and cultural identity. These authors advocate for expanding the meaning of ‘mining impacts’ to incorporate cultural heritage and recommend collaborative research partnerships with traditional knowledge holders, and localised impact studies in Global South contexts.

Health and labour

Artisanal and small-scale mines (ASM) are known for human rights abuses, with unsafe working conditions, lack of protective gear, and exploitation, including child labour. In Indonesia, more than 2000 ASM gold mines are spread across 30 provinces, many of them unlicensed and unregulated. While the Democratic Republic of the Congo is the world’s largest producer of cobalt, an important component of AI hardware, the Philippines and Indonesia also have significant reserves.

In a review, authors found that ASM is often more labour-intensive in Asia and Africa compared to other regions like South America. While the expansion of ASM for critical minerals could correlate with increased employment and income in these regions, without suitable working conditions and continued health and safety risks, ASM will exacerbate existing social problems. To combat this, the authors advocate that the sector be formalised with adequate regulatory oversight and policy enforcement.

Towards sustainable and responsible mining

Given the many environmental and social costs across the critical mineral supply chain, and AI’s role in driving some of this demand along with clean energy technologies, future growth trajectories hinge on decision-making that can factor in varying costs and benefits. More recent literature highlights AI’s potential to advance environmental justice considerations, centring community and planetary needs within larger extraction strategies. For example, authors suggest Geospatial AI modelling to identify prospective regions with lower ecological risks, and ML models to help create conservation strategies that minimise environmental degradation by integrating geological and ecological data.

A review of current and upcoming AI use cases in the mining industry frames mining as an endlessly complex array of trade-offs (e.g. between meeting budgetary constraints and reducing environmental harm). It proposes a multi-objective optimisation model framework that could balance competing objectives, such as profit, safety, land use and environmental impact, in a way that is transparent and accountable — so as to not amplify existing inequities, environmental degradation, and other harms endemic to the historic ‘resource curse’ dynamic experienced by Global South countries.

Spotlight: Critical minerals can be recycled

In its 2024 report, the International Energy Agency (IEA) stated that recycling is indispensable to the security and sustainability of critical minerals supply for clean energy transitions. Metal and material recovery from electronic waste (e-waste) creates a valuable secondary supply source, reducing reliance on new mines and providing supply security to importing countries.

In countries like India, innovators are tapping into the big business of waste, using machine learning to scale up the effective identification and sorting of e-waste in landfills while also protecting waste workers from the risk of contamination.

However, e-waste management in Asia is still a nascent economy. Rapid technological advancement renders equipment obsolete, generating unprecedented amounts of e-waste. According to a recent report, global e-waste is rising 5 times faster than documented e-waste recycling. Researchers stress the importance of a circular economy for sustainable management of e-waste, highlighting that in the absence of formal state policy, untreated dumping of waste and unscientific methods used for the recovery of metals can lead to contamination of air, soil, and water, creating toxic environments and public health hazards.

Data centres are also potential sites for reuse and recycling of obsolete equipment. A novel study proposes a computational power-driven material flow analysis framework to quantify the inflow, in-use stock, and end-of-service volume of AI servers in data centres. The researchers found that two actions in particular can reduce e-waste: 1) Extending the lifespan of AI servers by one year (keeping them in use longer) before replacement; and 2) Module reuse, which refers to the dismantling, renovation and reassembling of old key parts like GPUs, CPUs, and batteries for reuse in lower-performance computers. They note that coordinated efforts are needed across the AI value chain - model creation, chip production, data centre operation and waste disposal - to reduce waste footprint.

Around the Web

• In numbers: Critical mineral production, ownership, and social unrest is an interactive map that looks at four core critical minerals - lithium, copper, cobalt, and nickel.

• Researchers are constantly exploring sustainable materials for technology. In fact, Japan just sent the world’s first wooden satellite to space.

• The film Demon Minerals documents the Navajo Nation’s struggle for vital living space amongst abandoned uranium mines.

• The invisible seafaring industry that keeps the internet afloat is a good reminder that repair and maintenance cannot be an afterthought in this digital era.

Tuning in…

In the latest episode of Code Green, Tamara Kneese and Tom Özden-Schilling break down the complex landscape of mineral supply chains fueling global digital transformations and green energy transitions. Listen to an exploration of global extractivism, examining the environmental costs and labour implications associated with the infrastructures that underlie AI.

Credits

Research and curation: Dona Mathew, Meredith Stinger, Tammanna Aurora | Art Direction: Tammanna Aurora | Illustrations: Nayantara Surendranath | Layout Design: Shivranjana Rathore

Dear Reader,

At Davos this year, leaders spoke of the crucial part that technology has to play in climate adaptation, spotlighting innovations in AI-enabled early warning systems and agriculture applications.

But as we’ve highlighted in previous issues, while AI has many potential applications for climate action, these gains stand in tension with AI’s own water and energy intensive nature. Globally, approx. 1-2% of electricity is consumed by data centres, responsible for about 1% of GHG emissions. These numbers are expected to rise, with predictions that data centres will consume as much electricity as Japan by 2026. With the introduction of ‘leaner models’ like DeepSeek, this discourse is ever-evolving, prompting governments and investors alike to rethink the linkages between AI and energy demands, and highlighting the urgent need for reliable evidence on AI’s resource consumption.

Digital transformation across Asia, while promising economic development and job creation, is linked to rising emissions across the region. Reports warn that rapid AI development may be driving up reliance on fossil fuels, slowing clean energy transitions to power data centres. In the US, plans are in motion to reopen Three Mile Island to power Microsoft’s data centres, the site of the worst nuclear disaster in US history.

At the regional level, emerging data centre hubs in Malaysia, Indonesia, and India face potential water and energy shortages, intensified reliance on fossil fuels, and biodiversity loss. Singapore, which powers 60% of Asia’s data centre needs, imposed a moratorium from 2019 - 2022 on data centre expansion due to land use and grid constraints. During this pause, investment has shifted to neighbouring ASEAN countries, like Malaysia and Indonesia, with significantly lower operating, land, and labour costs, as well as government-backed incentives for foreign direct investment in data centres.

Even as Singapore’s and Malaysia’s governments lay blueprints for more energy-efficient and sustainable data centre operations, and try to prioritise domestic needs, residents in Malaysia’s Johor Bahru are increasingly worried about demands on the energy grid, high real estate costs, and water scarcity.

Curated Reads

In this section, we consider the many innovative approaches to Green AI, from data centre efficiency to hardware improvements and the nuances of making AI resource-efficient. As this is an emerging and fast-developing area of research, many of the readings we came across are pre-prints. Combined with a lack of transparency from model developers on AI’s carbon footprint and energy consumption, researchers also end up making simplistic extrapolations to arrive at resource-use estimates. So keep a close, critical eye on what you read.

Find definitions of some key terms in our climate glossary.

Green AI approaches

Green AI, a term coined by Schwartz and team in 2020, encapsulates approaches that monitor and mitigate carbon emissions and increase resource efficiency in the AI lifecycle.

In a collaborative study, researchers from Japan and Germany emphasise that training and retraining phases are major energy consumption hotspots, making it critical to analyse the computational costs of software used during these stages. Additionally, substantial environmental impacts occur during hardware manufacturing and operational phases.

To advance Green AI, researchers advocate for broader accessibility to knowledge and resources related to sustainable AI models. They recommend that AI programming frameworks not only support parallelism but also include tools such as pruning, low-rank decomposition, and knowledge distillation — methods, collectively referred to as "AI4GreenAI." While optimistic about the potential, they acknowledge that realising this vision will demand significant further research and development efforts.

Greening AI infrastructure

Data centres play a vital role in powering AI — they are the physical infrastructure that stores data, networks of computing resources, and cooling systems.

In Indonesia, researchers propose sustainability standards through the concept of Greenship Data Centre - a certification ratings tool specifically for digital infrastructure in Indonesia. The paper highlights the importance of digital twin technologies to simulate the performance and efficiency of data centres, helping operators identify inefficiencies and optimise energy usage, ultimately reducing the carbon footprint of the data centre. While it states that adopting digital twins poses data security risks and high upfront costs, it rationalises that the long-term benefits of sustainable data centres outweigh these challenges, echoing wider sentiments within the region of not missing out on technological advancements.

Hardware innovations in computing chips, such as 3D chip stacking, Neural Processing Units (NPUs), and Google’s cost-efficient AI chips Tensor Processing Units (TPUs) also promise energy and cost efficiency, faster performance, and customisation of AI tasks and edge computing. Challenges include high acquisition costs and high cooling costs to meet requirements for ideal temperatures to run properly.

There is a lot more nuance to this conversation

Recent literature highlights that a narrow energy-efficiency lens to AI lifecycles is inadequate to address issues around human-induced resource extraction. A systematic review of Green AI acknowledges the promising directions for energy-efficient solutions for AI computing such as structure simplification for deep neural networks, quantising decision tree outputs, and data-centric AI. It also found that Green AI research predominantly relies on image datasets due to their accessibility, but diversifying data types is crucial for a comprehensive understanding of the space.

But more importantly, it points out that the overt focus on energy efficiency as a core metric in this research omits monitoring of CO₂ emissions and other environmental indicators that examine the harms of computing more holistically.

The risks of greenwashing

In the race for digital transformation across Asia, the badge of ‘sustainability’ risks creating longer-term environmental inequality. For example, while current efforts towards carbon-aware computing are promising, this article argues that time-based load distribution may only displace energy use, not reduce it, and approaches that allocate tasks using geographical load balancing to clean energy locations may put undue strain on precarious grids, creating energy equity issues. Or it may increase the water consumption of data centres in water-stressed regions.

Bigger models = better performance?

The rapid adoption of LLMs for everyday tasks has also generated a lot of interest in trying to understand their environmental impacts. On one hand, larger training datasets have been equated with better performance, but on the other, bigger models require more time, and therefore more GPUs to train, leading to higher energy consumption and emissions. In a survey of 95 ML models across time and different tasks in NLP and computer vision, authors found that the carbon intensity of the energy grid used, as well as the training time of the models, were the most influential factors in determining emissions. Asking if the higher emissions related to large models meant better performance, the authors found no clear correlation between higher CO₂ emissions and better model performance.

Authors argue that the current bigger-is-better paradigm in AI is shaping research agendas towards economically and environmentally unsustainable ends while concentrating power in few hands who can afford massive computing power. They advocate for a “small is beautiful” approach to models, which balances resource efficiency with model performance. They emphasise that ML research should be transparent about computational costs, energy use, and memory requirements, using tools like CodeCarbon to track these metrics.

The rebound effect of efficiency gains

We are seeing innovations increasing energy efficiency across hardware and software emerge. However, primary drivers for investments in energy efficiency appear to be aimed at lowering operating costs and increasing energy autonomy rather than a commitment to mitigating environmental costs. So we have to ask… will the increased efficiency of hardware and models actually mitigate the environmental harms of AI compute or just ramp up production? The rebound effect of efficiency gains, in which increased efficiency can lead to increased consumption is a well-known issue in the energy sector, particularly in developing contexts. This author argues that the danger of this in AI development is very real. Cutting-edge, efficient hardware is expensive, and as model efficiency increases, we could see a resurgence of older hardware that powered the energy crises of early crypto mining.

Okay - where do we go from here?

Improving AI’s energy efficiency is an important goal, especially as AI is set to be adopted more widely for climate action. However, there is a need to re-evaluate priorities and mechanisms for measuring and making AI ‘more green.’ In an article about the Climate and Sustainability Implications of Generative AI, authors argue that we need to move beyond the narrow focus on energy and carbon efficiency in measuring AI’s impacts and adopt a lifecycle approach that examines comprehensive environmental and social impacts across stakeholder groups. They suggest a benefit-cost analysis framework for more sustainable AI at compute, application, and system levels as a starting point for Gen AI and industry community collaborations with broader stakeholder groups, such as CSOs, economists, social scientists, and supply chain groups.

Planter Box

How do we tackle the ecological polycrises we are currently facing? Michael Kwet suggests that time is not on our side — and that Digital Degrowth is the solution. He argues that under capitalism, our focus on limitless growth will never create the conditions for a just and equitable existence for all within planetary boundaries, and the drive for more is leading us to environmental/social collapse.

Taking a closer look at digital technology as a societal and economic construct; he urges us to see tech as not only shaping our day-to-day lives but also as a driver of global labour inequities. US tech supremacy controls social media, search engines, semiconductors, cloud computing systems, operating systems, business networking, office productivity software, and more. Meanwhile, the real (and exploitative) labour driving tech development is borne by the Global South; from physical risks of mineral extraction, underpaid roles of data annotators, and the mental health costs of social media content moderation.

In his book “Digital Degrowth: Technology in the Age of Survival”, he lays bare this unequal exchange between the global north and the global south — who owns the digital economy and its connection to environmental and human costs.

Around the Web

• Carbon Tracker: Track and predict energy consumption and carbon footprint of training deep learning models.

• CO2 Inference: Explore insights on energy efficiency and carbon emissions of AI models.

• Green Software Patterns: An online open-source database of software patterns. Also, refer to an environmental Impact Framework for software.

• Green Software for Practitioners: An introductory course by Linux for software practitioners to build, maintain, and run green applications.

• Inside the miracle of modern chip manufacturing: Read about the design and innovation powering your tech devices.

Part 2 of the environmental costs of the AI lifecycle will be out next month. We examine Material AI with a focus on extraction of critical and rare earth minerals, AI’s e-waste problem, and the circular economy.

Tuning in…

In the latest episode of Code Green, listen to John Cotton from the Southeast Asia Energy Transition Partnership and Priya Donti from Climate Change AI as they delve into the complexities of energy transitions and machine learning in Asia. A sneak peek here!

Credits

Research and curation: Dona Mathew, Meredith Stinger, Tammanna Aurora | Illustrations: Nayantara Surendranath | Art Direction: Tammanna Aurora | Layout Design: Shivranjana Rathore

Dear Reader,

In the face of significant global biodiversity loss, assessments highlight that Asia is the most underperforming continent in achieving global biodiversity targets due to factors such as rapid deforestation and cash crop plantations, construction of mega infrastructure projects, urbanisation, and unplanned use of natural resources.

As this crisis deepens, technological innovation is stepping up to play a critical role in conservation. A team of experts, primarily from Asia, has developed principles on Artificial Intelligence (AI) for biodiversity conservation, anticipating the growing use of AI in this space. Deep learning, computer vision, environmental DNA (eDNA), and networked sensors are seen as game-changers by global conservation experts. Leading up to last year’s United Nations Biodiversity Conference of the Parties (COP16), AI gained remarkable attention, including initiatives like the first International Summit on Indigenous Communities and AI.

However, as a field that has used digital technologies for decades, practitioners have been cautious in adopting advanced technologies for conservation, warning that technology alone will not solve all biodiversity challenges. In the last few years, across Asia, deep learning in ecological research has shifted focus from purely technical studies to addressing practical issues like land use change, biodiversity conservation, climate change, and societal impacts.

We examine existing and cutting-edge technologies in ecological research and conservation, the limitations and risks of data, and some new developments in funding models.

Spotlight: The Cali Fund

At COP16 in November 2024, we saw strides towards implementing biodiversity targets set in the 2022 COP15 Global Biodiversity Framework (KMGBF). Notably, we saw the creation of the Cali Fund, which focuses on sharing the benefits and profits from the use of Digital Sequence Information (DSI) of genetic resources to develop biotechnology products. Under the Cali Fund, large companies benefiting commercially from DSI use “should” contribute 1% of their revenue into a fund allocated to support the self-determined priorities of indigenous groups in implementing the KMGBF. However, conservationists are sceptical - there is no unified enforcement mechanism for contributions, and coordination is needed for open access to data, database interoperability, and better benefit-sharing outcomes.

June Rubis, a conservationist and part of the Global Council, ICCA Consortium, and International Indigenous Forum for Biodiversity, who was in Cali during the negotiations, told us that while the Fund is an important acknowledgement of the contributions of Indigenous Peoples, Afro-descendants, local communities, and Global South countries to biodiversity conservation, true reciprocity requires more than financial contributions — it demands the recognition and support of Indigenous governance systems and rights.

“Mechanisms like the Cali Fund must ensure that benefits are not only equitably shared but also directly controlled by Indigenous Peoples and local communities, respecting their self-determination and traditional knowledge. Without enforceable commitments and a rights-based approach, the promise of equitable benefit-sharing risks being unmet.”

Curated Reads

In this section, we look at use cases for AI in conservation, ethical tensions that arise, and the data question. Find definitions of some key terms in our climate glossary.

GeoAI - mapping and monitoring for conservation

After our last issue, we’re revisiting GeoAI to highlight its critical use for biodiversity conservation. In this article on deep learning and satellite remote sensing, the authors experiment with Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN) for spatial and temporal data, respectively, enabling tasks like mapping invasive species, monitoring vegetation, and counting animals like elephants and whales.

However, these tasks require high-quality labelled datasets and computational resources, usually inaccessible to Global South institutions. Other barriers to adopting technology include algorithm complexity, and environmental costs of computation, underscoring the need for sustainable practices and better data access in the field.

The emergence of bioacoustics and AI

Bioacoustics serves as a tool to document the changes to an ecosystem over time; recording biodiversity loss, monitoring species population, and tracking migratory patterns. It can be used to distinguish between similar species as well, proving useful in tackling the songbird trade crisis in East and Southeast Asia that threatens many species with extinction. A new deep learning-based bioacoustic tool can identify 15 morphologically similar White-eye species using bird vocalisation data to aid citizen scientists and law enforcement officials.

AI tools for marine ecosystems

To protect marine environments, innovative, noninvasive monitoring methods are essential. Traditional approaches often harm ecosystems and rely on field expertise. This review highlights the use of environmental DNA (organic genetic material like skin cells), metabarcoding, and high-resolution underwater imaging, which provide more precise and automated ways to monitor biodiversity.

Researchers compared machine learning (CatBoost) and multivariate linear models to monitor the Indo-Pacific eel in central Vietnam. The eel (Anguilla marmorata), vital for global ecosystems and fisheries, faces threats from overfishing and habitat loss, necessitating conservation efforts. They found that CatBoost achieved high accuracy and effectively estimated eel length, outperforming traditional models. These findings underscore the potential of ML tools in mapping species distribution and growth, providing valuable insights for conserving marine biodiversity and their habitats.

Sounds great, but what’s the catch?

The use of AI/ML is slowly but steadily increasing in conservation, but like other sectors, it is accompanied by risks and ethical considerations. In the case of remote sensing, York et al unpack the discursive landscape of justice and ethics in conservation technology, addressing tensions between anthropocentric and multi-species ideas of justice.

Potential risks to marginalised groups include oppressive surveillance and criminalisation of marginalised groups who might be reliant on forest produce for livelihoods. For instance, in a recent study of camera traps and drones in the Corbett Tiger Reserve in India, researchers found that cameras monitoring dry stream beds to track elephants and tigers had inadvertently captured a semi-nude woman in the area, which was later circulated on social media by temporary forest personnel.

Narratives on multi-species ideas of justice tend to focus on how technology-driven real-time monitoring can prevent illegal wildlife trade activities. The fear associated with such surveillance technologies is also considered a positive in some literature and the monitoring itself is framed as an ethical imperative, rather than a risk. However, for conservationists, it can be a dilemma to decide how much data should be shared with law enforcement officials if these technologies also capture the movement of local communities accessing forest produce in protected areas.

Data justice for biodiversity conservation

High-quality data is a critical scaffolding for developing AI applications. But the justice implications of data in conservation remain underexplored, particularly how data shapes representations, decisions, and the distribution of benefits and harms. In their work on data justice in biodiversity conservation, Pritchard et al identify the issues with global biodiversity datasets such as bias towards particular locations which have implications for how data is used and conservation resources are distributed. Those collecting and using the data are also distant from the people and environments represented in the data.

For systems that support people in understanding the use of data, recognising biases and challenging potential harms, they propose a data justice framework for conservation across five elements: data composition (what or who is represented), access and control (who benefits from data), use (how data is analysed), and consequences (outcomes of data-driven decisions).

AI for biodiversity conservation - being responsible

The Global Partnership on AI (GPAI) has put together a seminal report on the intersection of AI and biodiversity. They categorise some key risks from the use of AI:

• Impacts risks, where the use of AI does not add any value or is used as a means of greenwashing to distract from fundamental changes to business models;

• Social risks, when AI is deployed in languages local communities do not understand or is used to monitor environmental defenders; and

• Environmental risks, in how AI entrenches unsustainable business models and drives e-commerce consumerism, increasing the demand for resources and extraction with impacts on biodiversity.

Some of the key recommendations in the report include:

• Establish a community-data dialogue to identify sensitive data and explore tools like federated learning, edge computing and privacy-enhancing technologies for data use and sharing while respecting community data rights.

• Encourage philanthropic foundations to adopt an open funding model, actively supporting applications from the Global South.

• AI-for-biodiversity projects should explore offering services to the private sector as the market for biodiversity-positive business models grows.

On our radar: Interspecies Money

The lack of funding for conservation is a persistent challenge, even preventing clear pathways forward at COP16. This has led to some creative funding instruments from the private sector. For example, what if we could expand our definition of ‘money’ beyond human transactions to include how different species might contribute to or affect each other's well-being and ecosystems? The concept of Interspecies Money emerged out of this very question for Jonathan Ledgard and his company, Tehanu, becoming the first to operationalise the concept and make an ‘interspecies money’ transfer in 2024. Tehanu’s AI model assesses the needs of gorillas, providing conservation-related advice to agents using it, who receive mobile payments based on pre-determined values of the species when conservation targets are met.

Many of these interoperable systems and transfers are automated, leading sceptics to ask — who ascribes this value and how can interspecies exchange even be commodified? In this pre-print, researchers document how a digital stewardship and reward system for biodiversity conservation was implemented in a game reserve in South Africa. One of the tensions they highlight is in the often arbitrary nature of allocating monetary value to species protection - for example, if the returns from poaching or illegal trade are higher, how does one incentivise the protection of a particular species?

Around the web

Global Ecosystems Atlas: Launched at COP16, the Atlas is the first comprehensive map of the world’s ecosystems.

WildMe: Explore the impact of crowd-sourced image databases on wildlife conservation in this article documenting databases made by conservationists and citizen scientists.

Earth.fm: Immerse in a library of natural soundscapes from field recordists around the world.

Wings for Wildlife: In this compendium of 15 case studies from diverse ecosystems across India, researchers evaluate the integration of drones in conservation practices.

In this month’s podcast, listen to Eleanor Slade and VV Robin, both conservationists working across Asia, talk about how technology has improved human engagement with nature for the better but why challenges around data and financing mean we still have a long way to go in using AI for conservation.

Credits

Research and curation: Dona Mathew, Meredith Stinger, Tammanna Aurora | Illustrations: Nayantara Surendranath | Art Direction: Tammanna Aurora | Layout Design: Shivranjana Rathore

Dear Reader,

Satellite imagery, drones, IoT devices, remote sensors, and GPS-enabled devices generate rich geospatial datasets, allowing us to map the interplay of naturally occurring and human-made phenomena to a specific time and place. In an Ecologies of Entanglement interview, electrical engineer turned hardware and software developer, Agnes Cameron notes how this technology of seeing — so rich in information and zoomed-out in scale — can add to despair and dread as we witness the beginnings of environmental collapse. But this data also offers new vantage points from which to see some of humanity’s most complex problems.

The integration of AI and machine learning with Geographic Information Systems (GIS) is transforming how we analyse complex geospatial data, opening new opportunities for climate action. For example, Geospatial Foundation Models (GeoFM) are designed to process diverse datasets effectively. With improvements in access to geospatial data and tools, innovation is picking up. Initiatives like HOTOSOM sustain and use open-source mapping tools and data to aid humanitarian response.

But open source also carries risks of misuse and privacy violations. And, simply providing access to data is not enough. Geospatial data is complex and requires skills, resources, and training programs like NASA’s ARSET to promote meaningful use for environmental justice. Without guard rails to protect vulnerable groups, biased or poorly contextualised data can harm communities. Much like the paper maps of the past, contemporary geospatial data infrastructures are ways of seeing shaped by power relations, enabling states and private entities to map land and resources for governance and profit.

In this issue, we descend from the 30,000 ft view of a satellite and land on the ground in Asia, where we’ll explore pathways through GeoAI towards equitable environmental and social outcomes.

Spotlight: Disaster Response, Prediction, and Mitigation

In Asia, a region characterised by extreme weather events, disparities in the state of critical infrastructure, and inadequate access to appropriate early warning systems, the use of GeoAI has the potential to establish reliable systems of disaster response and management.

For example, Indonesia-based Peta Bencana builds flood risk maps, leveraging geospatial data, crowdsourced social media data, and validations by government agencies.

Researchers in Bangladesh propose using AI for tropical cyclone prediction. Graph Neural Networks (GNN), a deep learning model that works with graph data, are great for capturing complex interactions between atmospheric and oceanic datasets. Trained on historical satellite data and field observations, GNNs can predict cyclones from past wind flows and cyclone formation patterns. However, researchers note that GNNs are complex, difficult to train and require cross-field expertise to adopt.

The lack of credible data poses another challenge. While good data sets exist on basic rainfall, temperature, humidity, and wind speed parameters, there is a lack of granular, localised data in many regions. Data on social, economic, and health factors also remain incomplete, making it difficult to predict disease outbreaks or create efficient warning systems.

Curated Reads

In this section, we look at the data infrastructures that inform GeoAI, their role in environmental justice, and strategies for Responsible GeoAI.

Context Matters: State Power & Data Infrastructures

The digitalisation of maps is an ongoing process for many parts of Asia, as governments transfer paper-based land records to mapping software. Through a comparative analysis of One Map projects in Indonesia and Myanmar involving the consolidation of geospatial data onto digital platforms, authors embed the trajectories of these projects in unique political and historical contexts. They track the success of the project in Indonesia — attracting both foreign funding and domestic support — due to its rich history of counter-mapping in defence of indigenous lands, its established democracy, and its focus on climate mitigation in forest monitoring. In contrast, Myanmar's narrower focus on domestic policy reform, combined with bureaucratic resistance and the 2021 military coup, led to the project's failure.

Counter Mapping to document Indigenous land use

Top-down mapping initiatives have a legacy of negative impacts on community land rights and resource management practices, prompting communities to adopt their own cartographic methods such as counter-mapping. Increasingly, handheld devices such as sensors, drones, GPS trackers, and easy-to-use GeoAI software are simplifying the collection of community data (oral histories, local landmarks, and generational land use) and correlating it with available spatial data (hills, rivers, forests) - thereby democratising the process of mapmaking.

The expansion of community-led counter-mapping practices into the digital realm makes GeoAI a critical tool to protect native lands from ecological destruction. For example, Indigenous people in the Malaysian state of Sarawak, on the island of Borneo, are turning to such mapping technologies to formally establish their land rights and resist the expansion of palm oil plantations in their territories. These community-made maps have even been presented as evidence in ongoing litigation for previously undocumented native land rights to be legally recognised and thus protected for generations to come.

Critical Remote Sensing & Environmental Justice

Critical remote sensing has emerged as a counterpoint to extractive uses of satellite data in environmental justice struggles. Through affordable IoT kits, citizen scientists can collect data to establish their own evidence base. Following citizens deploying CitizenSense kits near rural fracking sites in Pennsylvania, filling gaps left by government air quality indices, authors argue that while this data may not meet regulatory standards, this diverse, frequent collection produces “just good enough data” when combined with official measurements to demand legal accountability.

An excellent guide for those interested in focused applications of remote sensing for environmental justice, this socio-technical framework identifies key open-access resources to support ground-level stakeholders, for efforts in water management, timber extraction, food production, and fossil fuels use. Authors argue that through knowledge sharing and expanded access to resources, remote sensing data infrastructures can (and should) be adapted to expose environmental injustice.

To be open or not - resolving inherent tensions

While aiming for transparency, open-access data also presents privacy and misuse risks. In this interview, Finnish experts highlight that geospatial datasets can include sensitive information about critical infrastructure, creating vulnerabilities to physical attacks. They then recommend using a risk assessment matrix and implementing a process to evaluate risks before releasing geospatial data publicly.

Purpose-oriented data releases are another way of preventing data misuse. In Norway, the International Climate and Forests Initiative (NICFI) Satellite Data Program, provides commercial satellite imagery covering the world´s tropics for non-commercial applications specifically for climate action, like reducing and reversing the decline of tropical forests. Such approaches are particularly relevant for Asia, which is increasingly seeing government-led open data initiatives.

Doing Responsible AI with geospatial data

In their systematic review, Marasinghe et al. emphasise the need for human-in-the-loop and participatory approaches in leveraging GeoAI, especially in urban settings. Given the vulnerability of Asian cities to climate change and the growing use of GeoAI for monitoring and early warning systems, the article’s proposed Responsible AI framework offers valuable guidance for policymakers and researchers.

The authors caution that GeoAI systems, which often rely on visual data to assess built environments, can produce inaccurate or unethical results if local context, such as cultural or ecological factors, is overlooked. They emphasise the importance of involving diverse stakeholders, including local communities and experts, in the design and implementation of GeoAI to ensure these tools meet the specific needs of the communities they serve. This resonates with the findings in another pre-print, where the authors highlight the value of multi-stakeholder perspectives in data labelling. Labelling, the process of categorising real-world phenomena into distinct classes or groups, is highly context-specific. Therefore, it requires regional expertise and knowledge, and is ideally coordinated with local experts.

On our Radar

Globally, we are seeing a lot of interest in using Foundational Models (FM) on multi-modal geospatial datasets for environmental monitoring. For example, Hong et al. developed SpectralGPT, a universal remote sensing FM that can handle images in ultraviolet, infrared, various sizes, resolutions, time series, and regions; useful for crop health or deforestation monitoring.

Last year, IBM and NASA jointly launched their first open-source geospatial FM for Earth observation data. Built using NASA's Harmonised Landsat and Sentinel-2 dataset, it has been applied to flood mapping, with significant accuracy using smaller samples. It also aids burn scar identification, crucial for active fire management and post-fire recovery efforts.

In a vision paper on GeoFM, researchers note that task-agnostic large language models (LLMs) can sometimes outperform task-specific, fully supervised models in zero-shot or few-shot settings. However, for other geospatial tasks involving multiple data types like street view image noise classification and remote sensing scene classification, current FMs lag behind specialised models.

Around the web

• Navigating GIS: A guide by Cathy Richards to make responsible mapping more accessible for environmental justice organisations.

• Counter Mapping: In this documentation, the Zuni People counter Western notions of place and geography by bringing their Indigenous voices to map creation.

• Indigenous Communities & Mapping: Hear Michael Jok from Borneo speak about how Indigenous communities in Malaysia are using GIS to assert land rights.

• Data Dialogues: Interested in the complexities of environmental data? Tune into this podcast series by Open Environmental Data Project for learnings & reflections from the ecosystem.

Our next podcast episode is out! Cindy Lin, an ethnographer working on critical data studies and environmental governance, and Sherif Elsayed-Ali, with a career spanning from human rights to Climate AI entrepreneurship, reflect on the current noise around the potential of AI for climate action. Highlights from the episode include:

• In the best use-case scenarios, AI could help optimise existing processes such as cement manufacturing to reduce emissions;

• Developers often overlook externalities like energy or water requirements;

• Sometimes “collective restraint” can spur innovations in ‘small’ infrastructures;

• All AI is not built equally - some consume more energy than others.

Tune in as experts break down some of the myths of Climate AI and the systems that sustain them.

Our upcoming newsletter editions look at AI applications for conservation as well as the environmental impacts of this technology. For our readers working or interested in this space - what innovations or emerging use cases are you excited about? What are you reading? Email codegreen@digitalfutureslab.in to share your ideas for our next issue.

Credits

Research and curation: Dona Mathew, Meredith Stinger, Tammanna Aurora | Illustrations: Nayantara Surendranath | Art Direction: Tammanna Aurora | Layout Design: Shivranjana Rathore

Dear Reader,

Is data the new crop? It’s a question we found ourselves faced with while digging into the tensions of embedding AI in agricultural practices. In the latest episode of the Code Green Podcast, we asked Elenita ‘Neth’ Daño and Anubha Singh for their thoughts on it which led to a discussion on whether farmers are the ones who really benefit from the cultivation of agricultural data.

“I think it's unfair to crops to be used as the analogy of how this works because everything from the seed to the benefits is antithetical to what crops mean in agriculture,” says Neth. “In real agriculture that involves labour, that involves contribution of everyone in the communities and recognising that contribution, everyone knows where the ‘data’ comes from, how it is used, how it is harvested, who benefited.”

Amidst disruptions to traditional farming practices due to extreme weather events and the spread of pests & disease, AI is emerging as a potential solution for climate-resilient agriculture, with applications ranging from AI for precision farming and early warning systems to AI-powered chatbots for farmers. 

The 350 million smallholder farmers in Asia stand to lose the most to climate change. Fluctuating yields, economic losses, and myriad health impacts are just some of the risks they face. 

There are concerns that innovations in precision agriculture and farm automation are not suited for small landholders, and adoption challenges may go unaddressed without parallel investment in farmer training, financial support, and community consultation.

On the one hand, agri-tech innovations have the potential to empower farmers by improving market access and data-informed decision-making. Agricultural data can be used to build models focussed on regenerative agriculture while providing critical data for policymakers.

On the other hand, the trend of corporate-state partnerships seen in agri-tech innovation may lead to the monopolisation of digitised agricultural value chains while the privatisation of public extension services may create information asymmetry through unfair data extraction.

Curated Reads

In this issue, we explore: What is the tech optimised for? How does it incorporate local traditional knowledge systems? What are farmers' perceptions of technology? How to design and develop AI for Agriculture that protects farmers’ autonomy and leads to better outcomes for agriculture?

Integrating local knowledge systems & scientific data for weather forecasting

Farming practises everywhere rely on the ability to read the land and weather, traditionally taking cues from seasonal patterns of birds, insects, flowering plants, cloud movements, and other changes in local landscapes to make predictions. How can this keen understanding of ecology be integrated with data-driven technologies like AI?

A recent study in Bangladesh, Guatemala, and Ghana tapped into generations of accumulated knowledge and practical farming experience to co-create a hybrid weather forecasting system that improved the legitimacy of weather forecasting services and produced tailored and actionable insights for farms. Farmers participated by using an app to enter their daily predictions based on ecological indicators and their study of their surroundings. This user-driven data generated was integrated with scientific forecasts from local meteorological services to increase the accuracy of weather forecasts (credibility), while at the same time ensuring context-relevance (salience) and user ownership and trust between stakeholders (legitimacy).

Smart Agriculture in the Phillippines: Use cases, challenges, and limitations

Expert commentary from our AI + Climate Futures work notes that smart agriculture is gaining momentum in The Philippines to combat climate change, guided by the Agriculture 4.0 agenda, which promotes digitalisation through technologies like drones, blockchain, and data analytics. Public-private partnerships with foreign governments, transnational corporations, and agri-tech startups are picking up pace. Technologies like sensors, drones, and smartphone applications assess pest infestations in major crops, while digital tools like the Philippine Rice Information System (PRiSM) are being integrated into government programs. The government is also digitalising farmland data through the Enhanced Farmland Geographic Information System (eFGIS).

However, no independent studies assess the impact of AI and digital tools on productivity for climate action. The expert-author highlights key concerns, which ring true for similar economies in the region, such as financial burdens on small-scale farmers, data ownership, and the environmental impact of digital tools. Moreover, these tools do not address practices contributing to the climate crisis - like the continued use of chemical fertilisers affecting soil and groundwater. Given Asia’s context of majority smallholder and subsistence farmers, the author argues that infrastructure and extension services are critical to making AI-based solutions accessible to marginalised farmers.

Technology for farmers, with farmers

Given the economic costs of adoption and digital capacity requirements, what motivates farmers to adopt advanced technologies like AI? For Nepal’s small-scale farmers, researchers found that factors such as perceived usefulness and ease of use as well as social factors such as technology anxiety shape farmer attitudes towards technology adoption. They emphasise the importance of government support for farmers through training and subsidies to increase trust and confidence. 

In technology development cycles, it is crucial to engage with farmers from early stages. In a study with Australian farmers, researchers argue that factors such as transparency and clarity regarding data ownership, portability, privacy, trust and liability in commercial relationships, contribute to farmers’ reluctance to engage in the widespread sharing of their farm data. Top-down approaches by agribusinesses may constrain farmer autonomy by targeting smallholder farmers as consumers rather than as primary stakeholders. Methods such as choice architecture and behavioural nudging are used to shape farmer behaviour in a way that generates and distributes value and information asymmetrically, creating dependencies and lock-ins.

However, participatory bottom-up approaches can also be hard to navigate. In a study of a failed participatory technology development initiative for assessing soil fertility, researchers found that in the Japanese context, cultural norms that emphasise politeness, modesty, and respect and factors like gender and status influence stakeholder engagement in discussion forums. In this case, as farmers’ perceptions of the technology’s usefulness wavered, so did their engagement with researchers. The authors suggest that self-evaluation models are crucial to enabling participatory formats to address discrepancies in interpretations of the usefulness of technology interventions.

Creating trustworthy AI for Agriculture applications

A farming ecosystem is a complex web of relationships between people, technologies, and the environment. In addition to stakeholder engagement, a thorough risk assessment and mitigation of systemic risks, such as data issues and unintended consequences, is important to responsibly deploy AI in agriculture. The authors of this paper propose a social-technological-environmental system (STES) framework for Responsible AI design to mitigate negative impacts on farmers’ decision-making, asking questions around the core functions of the application and attributes at risk.

But who is responsible for implementing these frameworks? Scholars at UC Davis interviewed researchers of the Artificial Intelligence Institute for Food Systems in the US and found misalignments and conflicts between standards, ethical principles, and interests of researchers, funders, and food system partners. For example, funding agencies' transparency mandates could deter potential partners from sharing data, reducing collaboration and limiting data diversity for researchers.

In the absence of clear guidelines, and when there is such a mismatch of interests, it is the researchers who must personally build trust with stakeholders. They do this by “(1) presenting AI technologies as an aid, not just a displacement of human labour and decision-making; (2) privacy of the data with which they are entrusted, (3) reliability and accuracy in the algorithms they use the data to produce; and (4) building long-term relationships of mutual trust.”

Around the Web

• Our World in Data looks at land use, food production, and crop yield over the years through interactive maps.

• Saukaash Jaa, a NatGeo-funded multimedia project explores the worries of the rice-growing farmers of Ratnagiri over the increasingly erratic southwest monsoons. 

• Farm Hack is an open-source platform run by a global network of farmers for building, fixing, and tinkering with their own tools. 

• Grow Asia’s Digital Directory is a collaborative effort to make information, finance, and market solutions accessible to investors, agribusinesses, and smallholder farmers. 

• Dive into the fascinating world of Seed Banks and their role in climate resilience and averting a food crisis.   

As we look to the future of farming, Anubha Singh reminds us that the ‘fail fast’ approaches we currently see in the agri-tech startup landscape may be doing more harm than good for farmers: 

Anubha: “Because I think one of the biggest harms these technologies do– and in these incubators as well, where I did my fieldwork, a lot of these startups were taught, ‘oh fail fast, design a prototype, test it out in the field, fail fast’. I think that ends up doing a lot of harm in real-world situations, like in agriculture, farmers end up losing their income for the whole season. And then, it accrues this entire cycle of debt and all of those things.”

Credits

Research & Curation: Dona Mathew, Meredith Stinger, Tammanna Aurora | Illustrations: Nayantara Surendranath | Art Direction: Tammanna Aurora | Layout Design & Communications: Shivranjana Rathore | Editorial: Urvashi Aneja

We'd love to hear from you as we work on our next issue on Geospatial AI for climate action - do write to us with your research and recommendations at codegreen@digitalfutureslab.in

Dear Reader,

Late last year, the International Energy Agency (IEA) published an article characterising energy and AI as the next ‘power couple’ — a call out to the increasingly complex energy grids of the future that require more advanced analytical tools to manage. Given the hype around the potential of AI, this coupling might be unsurprising, but how challenging is this integration considering the many factors that complicate energy systems and transitions?

The energy sector remains the largest global contributor to greenhouse gas (GHG) emissions. With fossil fuels like coal, oil and gas responsible for more than 75% of global GHG emissions, the transition to cleaner sources of energy is urgent. And despite the region’s renewable energy growth, rising electricity demand continues to drive up coal emissions. Digitalisation and power-hungry data centres will only add to this demand for electricity.

As Asia navigates this complex transition, the economic and human costs loom large, with energy transitions projected to cost $5.8 trillion annually in 48 developing economies alone. Phasing out coal can lead to significant job losses, as observed in China and Europe. Amidst these challenges, there is a growing push for smarter energy grids, with varying levels of automation.

While AI promises to enhance efficiency and predictability in energy management, its adoption must be tailored to Asia's unique challenges, particularly in countries where energy access is still insufficient. The path to AI adoption also needs to solve challenges around privacy, cybersecurity and lack of AI explainability.

Read on to find out more about AI’s role in sustainable energy futures for the region. 

Curated reads

Research on approaches and frameworks highlighting key challenges, solutions, and opportunities in adopting AI across the energy system. Find definitions to some key terms in our climate glossary.

Synthetic data could address core challenges to using AI in energy systems

Given the critical nature of energy infrastructure, trust is key to AI adoption for stakeholders across the energy system. While large, diverse datasets enhance the reliability of energy AI, collecting sensor data can be costly, raise privacy concerns, and may have gaps. To address these issues, researchers propose using synthetic data in the energy sector. 

Researchers suggest that synthetic data can improve Trustworthy AI by creating transparent, unbiased datasets that lead to fairer models while promoting sustainability by reducing the need for resource-heavy real-world data collection.

Using synthetic data does have its challenges, however, including in assessing its reliability and determining which real data is needed to create valuable synthetic data. For South and Southeast Asia, which have the highest number of households that do not have equitable access to energy, gaps in ground data will determine how reliable derived synthetic data is.

Solving for the efficiency & privacy trade-offs in smart meters and smart grids

As Asia moves toward grid optimisation in smart energy grids, data generation will rise, making privacy a topical concern. The use of granular data from real-time monitoring of consumers in smart meters has the potential to reveal personal details like appliance use, household size, and daily routines. While strong policy and regulatory frameworks are needed to protect privacy while pursuing energy efficiency goals, researchers have also examined ways to protect consumer privacy in smart meter usage and data.

The authors of this article undertook a survey of privacy-preserving techniques used in smart meters (SM). Some of these techniques include altering SM data through cryptography, adding random noise to the data, and managing demand-side energy storage. There are some limitations with these methods; de-anonymisation is still possible and these compute-intensive protections are expensive, requiring more research and investment to strengthen solutions.

Predictive cooling in the tropics

At the consumption stage, rising temperatures mean more AC usage, particularly in commercial buildings, straining the energy grid. In what could be a boon for sustainable and efficient cooling in tropical climates, researchers have shown how advanced deep learning models can predict next-day cooling needs for these buildings. They show the advantages of sequence-to-sequence models in knowing what data to keep and what to discard. And how, by using historical data, these data-driven models are easier to develop than more complex physics-based models. 

Some limitations of these models include privacy concerns stemming from the need to monitor building occupancy as crucial data for predicting energy use. Other challenges to overcome include gaps in data quality, quantity, and interpretability. For sustained adoption, interpretability of models is needed to increase trust and accountability in smart energy systems, as this research article on Explainable AI for energy found.

AI-enhanced building energy management in challenging urban environments

In many Asian cities, extreme environmental conditions, unstable power grids, rapid development, and limited historical data require new approaches to creating sustainable and efficient energy systems for buildings. 

This article proposes a physics-based model to leverage the best principles of both machine learning and physics for forecasting in energy systems. As traditional data-driven building energy forecasting models are hard to generalise (i.e., the ability to make accurate predictions about unknown data based on limited training data), they argue for using a physics-based model that predicts building energy usage and a lumped-parameter model to predict how buildings will handle heat. A Domain Adaptation technique, a type of Transfer Learning that uses data from similar buildings to reduce reliance on historical and labelled data, is added to tweak pre-existing models to work in new situations. 

Combining these techniques predicts future data points not included in initial measurements by converting simulated physics data into estimates for unobserved periods.

The Ethical Turn Towards AI for Energy Justice

Even as we focus on technical AI solutions for energy management, it's crucial to consider the broader societal impacts of these technologies. Given the context of inequities in energy access in Asia, adoption of AI in the sector should seek to reduce, not deepen marginalisation. To ensure equity in the energy system, researchers argue that energy justice principles should inform the selection of data, algorithms, models, and metrics in the ‘AI for Energy’ lifecycle.  

Energy justice is a framework to ensure fair and equitable access to affordable, reliable, and clean energy resources. The choice of AI techniques, such as centralised or decentralised control, can impact how communities access and control energy sources. One example is Reinforcement Learning, which can adapt to changing environments, allowing for greater control over localised data. However, poor interpretability of system behaviour affects its accountability, leaving room for future development.

On our radar: Decentralised approaches to energy systems

Decentralised renewable energy systems, such as solar microgrids, are poised to expand significantly due to their resilience to disruptions, enhanced reliability, and price stability owing to the localised, and community-governed nature of energy generation. Decentralised systems are small energy networks creating power independent of, or to supplement energy from larger grid systems. In Japan, for example, microgrids are a key feature of smart and resilient cities in disaster-prone regions. 

Here’s a look at new mechanisms and indicators of change in Asia, with a special highlight of AI’s role: 

Microgrids 

Despite increased interest and funding, microgrid projects have trouble taking off. Small microgrids struggle with demand variability and renewable energy volatility. Across India, solar microgrid projects fall into disrepair due to a lack of local maintenance capacity in communities. Failed projects are seen by residents as “fake energy,” not meeting local needs, and creating toxic waste as defunct equipment degrades over time. In Indonesia, public-private funding models fizzled over time, failing to account for short infrastructure lifespans and high maintenance costs. 

Despite these hurdles, renewable microgrids hold potential due to the low cost of solar panels, improved battery technology, and suitability in tropical climates in the Asian region. Emerging integrations of AI to address barriers in microgrid implementation are nascent yet promising and could potentially improve the sustainability and success of these projects. 

Energy communities 

Energy communities are characterised by local ownership, collective decision-making, and distributed economic and social benefits from energy sources and are shaped by localised needs and demands. For example, in Malaysia, the local workforce supports the electrification of rural areas by installing, managing, and maintaining solar plants in their villages. For the most part, however, Asian countries still have a way to go in creating a regulatory environment that permits private entities to generate, manage, and trade in energy.

With the introduction of smart grids and AI tools, communities have the ability to forecast their energy needs, categorise consumers, control loads, and examine data to gain insight into their energy consumption patterns. This bottom-up approach to community-driven governance along with local funding models could enable inclusive and just energy distribution, making the implementation of energy policy more effective. It could generate jobs and revenue as community representatives become more involved in energy management and trade in surplus energy.

Around the web 

• A series of ground reports by Scroll on the human cost of energy transitions in India.

• A cheat sheet from MIT on the major greenhouse gases driving climate change.

• Image Carbon offers a way to calculate the carbon footprint of your website along with ways to reduce it.

• Low Tech Magazine has a new book series, How to Build a Low-tech Internet, on their solar-powered website! It considers the historical context of digital technology and its impact on sustainability.

• Coldplay’s Tour Emission Update for 2024 and Massive Attack’s latest festival gig lead the charge on green-ing the music industry.

Our second podcast episode is out! Listen to Cathy Richards and Izni Zahidi as they discuss the applications of Geospatial AI for climate action, trends in democratisation of map-making, and the ethical risks arising from the use of GIS data.

Thank you for reading Code Green! We would love to hear from you — please do write to us with your feedback and suggestions at codegreen@digitalfutureslab.in

All our newsletters and podcasts can be found at codegreen.asia

Credits

Research and curation: Dona Mathew, Meredith Stinger, Tammanna Aurora | Illustrations: Nayantara Surendranath | Art Direction: Tammanna Aurora | Layout Design: Shivranjana Rathore

Dear Reader,

Hello and welcome to the first issue of Code Green, a monthly newsletter published by Digital Futures Lab in partnership with Earth Venture Foundation. 

Code Green showcases the latest interdisciplinary research and initiatives at the intersection of AI and climate action in Asia, a region at the frontline of climate impacts and digital transformation.

“Not another newsletter on climate tech!” you say. Here’s what makes us different. Our work traces the interconnectedness with which society and technology operate and develop. Here, we separate the heat from the hype around climate tech and focus on how AI innovation trajectories can align with just and equitable climate transitions in Asia. 

Each newsletter will explore a specific theme, such as the opportunities and risks of using AI for precision agriculture or the role of AI in advancing energy justice in Asia. We’re also excited to announce a ten-part podcast series to complement the newsletter featuring dialogues with researchers, industry experts, investors, technologists and community leaders. 

By tuning into Code Green, you aren’t only staying informed. You’re joining a community dedicated to driving meaningful climate action through human-centred approaches, responsible innovation and cutting-edge technology.

So, let’s get started! In this opening edition, we look at the landscape of innovation and action, highlighting key issues and areas of research.

Research Spotlight

The State of Play in Machine Learning for Climate Action

Accelerating climate change is one of humanity's greatest challenges, and current responses have been falling short. There is great excitement and anticipation around the potential for machine learning (ML) to provide powerful tools to tackle it, but what exactly is the state of play in this area? How advanced are these technologies, and in what fields can they help? 

This article by experts at Climate Change AI, a global non-profit that catalyses impactful work at the intersection of climate change and machine learning, is a great way to get acquainted with the high-impact problems where ML can fill existing gaps. It highlights potential uses of AI in a wide range of sectors, including energy systems, agriculture, transportation disaster reduction, and climate financing. In energy systems, for example, AI can forecast demand and supply and improve system monitoring. 

Using AI for climate action is not without its challenges and risks. This article by scholars from the Oxford Internet Institute highlights how AI can amplify social and ethical challenges already associated with AI more generally, such as unfair bias, discrimination, or opacity in decision-making. For example, data collected to build AI systems for managing energy demands may contain sensitive personal information and risk individual and group privacy.

Building a Supportive Ecosystem for AI in Asia

These global narratives on AI and Climate Action need to be contextualised to specific local contexts. Here, we highlight a landscape report from our recent project, AI + Climate Futures in Asia, where we mapped the maturity of the AI innovation ecosystem in the region. We found that while East Asia shows higher AI readiness compared to South Asia, both regions face common challenges like a lack of quality data and limited investments in R&D. For example, a lack of ground-level data in agriculture and concerns around cybersecurity in the power sector complicate high impact AI adoption in key sectors. 

A key recommendation from our project was the need to develop equitable data-sharing mechanisms rooted in principles of collective responsibility, reciprocity, and accountability. Failure to do so can erode local knowledge systems and community agency, as this article by a leading expert on indigenous data sovereignty highlights.

Environmental Cost of AI 

The environmental cost of developing and using AI is a significant complicating factor in the narrative around AI and climate change. Researchers at the AI startup Hugging Face and Carnegie Mellon University were among the first to calculate the carbon emissions of AI models for different tasks. They found that the most energy and carbon-intensive tasks involved the creation of new content, such as text and image generation and summarisation tasks.

Apart from AI use, the environmental costs of data centres – the physical infrastructure powering AI – must be acknowledged as countries like Singapore, China, and India rapidly expand their digital infrastructure to support AI capabilities. These data centres are both “thirsty & power hungry,” a report from China Water Risk tells us. In China alone, the water consumption of data centres could more than double by 2030, stressing already scarce water resources in the region.

Innovation Spotlight

Top-down approaches to building AI tools for climate action are likely to exclude the needs of those most vulnerable to climate change. Instead, we need a consultative approach that amplifies the agency and capacity of vulnerable communities. 

CoRE Stack, a new initiative by the teams at Gram Vaani, IIT Delhi and Palakkad, BITS Pilani, and Well Labs, attempts to do just that! They have developed a technology stack to support communities in India in understanding the socio-ecological situation of their landscape, planning for the creation and repair of various natural resource management assets, and sharing these demands with relevant government programmes and donors for action.

Around the web

• Visualise how your city may look in the future amidst changing weather patterns with Climate Zones by Derek Taylor.

• Get tools and resources to become familiar with the Climate Finance space with this wiki by Climate Change AI.

• Curious about what Responsible AI looks like in practice? Check out the new Global Index for Responsible AI. Or see the  top 10 takeaways here.

Next up on Code Green

Responsible AI in Asia’s Energy Transitions: Our next newsletter covers the use of AI to manage energy transitions in Asia. It highlights emerging innovation opportunities in the region, from supporting the transition to renewables to optimising existing systems. But we also need to address concerns about energy equity, data privacy and security. 

Precision Agriculture in Asia: Cutting-edge technologies like satellite imagery, AI-driven analytics, and IoT devices are poised to revolutionise the agricultural landscape globally. However, they could also exacerbate the gap between large agribusinesses and smaller farms. Our third newsletter explores this and other issues around precision agriculture.

Our first podcast episode is also out! Listen to leading experts at the intersection of AI and Climate Action in Asia from a policy, technical and financial lens. 

• Elina Noor, Senior Fellow, Carnegie Endowment for International Peace

• Alpan Raval, Chief Scientist AI/ML, Wadhwani AI

• Varad Pande, Partner and Director, Boston Consulting Group

Thank you for reading Code Green! If you found this newsletter useful, please subscribe and share it with your networks.

Subscribe now

We would also love to hear from you—please do share your feedback and suggestions at codegreen@digitalfutureslab.in

All our newsletters and podcasts can be found at codegreen.asia

Earth Venture Foundation is a not-for-profit initiative providing grants and sponsorship to academic and scientific initiatives that contribute to the fight against climate change.

Digital Futures Lab is an interdisciplinary research and advisory firm studying the complex interactions between technology and society. Through evidence-based research, foresight, and public engagement, we work to develop pathways toward equitable, safe, and sustainable futures.

Credits

Research and curation: Dona Mathew, Meredith Stinger, Tammanna Aurora | Illustrations: Nayantara Surendranath | Art Direction: Tammanna Aurora | Layout Design: Shivranjana Rathore