The Promise and Peril of AI for Nature

If you’ve been tracking the latest developments with breakthrough artificial intelligence applications like ChatGPT Plus and you are feeling a little creeped out, don’t worry, you are not alone. For anyone who has played around with the earliest version of Open AI’s chatbot – known as ChatGPT-3 – or now the application’s latest and much more powerful sister, ChatGPT-4, it is a little bit what it might feel like to experience vertigo while taking a walk through the uncanny valley, where robots are acting so human-like that they are downright scary.

There is no question that we are on the precipice of a wondrous new technological age, in which breakthrough AI systems stand to deeply transform every sector of the economy, with the potential to increase living standards and unlock untold wealth. But we’ve arrived at this paradigm-shifting moment at the same time that humankind is facing existential environmental crises. In addition to anthropogenic climate change, human activity is causing a mass extinction, only the sixth in the history of life on Earth. According to the World Wildlife Fund, since 1970, the world has lost 68% of its biodiversity, causing significant consequences for ecosystems, species, and the human communities and ways of life linked to the natural world.

In the face of this paradox, the question is obvious: Can we harness these transformational AI tools to help protect the natural world? And conversely, how can we ensure that the applications of AI systems do not accelerate or worsen harms to humankind, ecosystems, and to the communities who depend on ecological preservation?

As to the first question, a growing number of use cases are demonstrating how AI-based tools can support efforts to protect and restore nature and limit the impact of environmental disruptions. Organizations large and small are developing and deploying AI to assist in natural disaster risk reduction, combat environmental crimes such as illegal logging and land ownership fraud, clean up ocean plastic pollution, and measure the carbon sequestration capacity of forests as the basis for carbon credits that support local communities. The nonprofit Climate Change AI catalogs applications at the nexus of machine learning and climate change, and AI for Climate produces case studies from around the globe.

Of particular importance for the biodiversity crisis, AI tools are being used to help protect endangered species and monitor the health of critical ecosystems. Under the project Tech4Nature Mexico, my organization C Minds, with the support of the International Union for Conservation of Nature, has brought together a consortium of government, university, private sector, and local community partners to develop and deploy AI tools to help protect biodiversity in Mexico’s ecologically-rich Yucatán Peninsula.

These systems are based on machine-learning and deep-learning algorithms trained on camera-trap images and eco-acoustic data, the sounds of an ecosystem, gathered over several months in mangrove and tropical forests of the Yucatán. So far, the project has identified and monitored 116 wildlife species, several of which experts were unaware lived in the area. One algorithm identifies individual jaguars based on their spot patterns. By enabling conservation experts to rapidly process and interpret vast amounts of collection data, these applications dramatically reduce the costs and time of wildlife monitoring, the first, essential step for conservation. They also provide a base of evidence and insights on which local communities, NGOs, and government officials can make policies and take conservation action.

But though such promising applications are emerging, the story of AI’s impact on our natural world is not all sunshine and rainbows. These tools can also generate environmental and social harms that need to be identified, mitigated, and managed. For example, in the case of wildlife conservation, inadequate data security might mean that an AI system designed to track animals could end up aiding poachers.

AI systems also have a large carbon footprint. Researchers have found the process of training a common large AI model can emit as much as 626,000 pounds of carbon dioxide equivalent, five times the emissions of a car over its lifetime. Experts estimate that emissions from the use of AI technology overall are similar to those of the aviation industry.

As with all greenhouse gas emissions and environmental degradation, the impacts are disproportionately felt by historically marginalized communities. And there are also other ways AI risks exacerbating historical injustices and global power imbalances. Powerful nations and companies, many of which are in high-income countries, lead in the development and implementation of AI systems, including those applied to nature and biodiversity in low- and lower-middle-income countries. The failure to involve and consult with local and indigenous communities in the development and deployment of AI systems that will affect their environments and livelihoods risks reinforcing existing social inequalities, neglecting community needs and interests, and worsening the digital divide.

Steps must be taken to mitigate these risks. For one, when it comes to tools that aim to promote conservation and protect biodiversity, indigenous perspectives are essential. As the guardians of many ecosystems, these communities often have unique historical knowledge and practices related to conservation and sustainability that should be valued and included in high-tech conservation efforts. Whereas Western scientific approaches tend to be reductionist and focused on quantifiable information, indigenous knowledge systems frequently take a more integrated perspective, recognizing the interconnection of all living beings and their environment. Both perspectives should be incorporated and balanced in informing AI for nature and biodiversity protection and regeneration.

In addition, stakeholders must aim to close the digital divide and promote connectivity and accessibility for AI technologies to ensure their successful integration and adoption. This means ensuring access to high-speed internet and computing resources in remote and low-income regions. Similarly, it is essential to design AI technologies with accessibility in mind, addressing issues related to affordability, usability, and availability. The larger goal should be to ensure AI technologies are accessible and beneficial to all individuals and communities, regardless of socioeconomic status or geographic location.

Finally, ethical frameworks and standards should guide the development and use of AI in environmental applications. In 2020, I had the opportunity to co-draft the global agreement on AI ethics led by UNESCO. Signed by 193 countries, the guidelines were the first of their kind to include a chapter on AI and the environment. Although this global effort was a significant step, greater international awareness and other actionable protocols for the responsible application of AI in this field are necessary. To this end, there is a great need to merge the digital rights movement with the environmental and climate justice movement.

We must remember that AI carries risks. Nor is it a silver bullet to solve the environmental crisis. But if developed and used responsibly and inclusively, AI can be a valuable aid in protecting our planet's natural wonders and promoting better practices for the well-being of all species.

Constanza Gomez Mont is the founder and principal of C Minds and AI for Climate and was the 2021-2022 AI for Humanity Chair of the World Economic Forum Global Future Council on the Future of Artificial Intelligence.