Artificial intelligence and ecology: a surprising pairing, isn't it? However, artificial intelligence (AI) has now become a revolutionary tool to facilitate, accelerate, and render secure our daily lives. Nevertheless, we will not discuss AI’s role as a simple fraud detector, or as an unbeatable competitor in a game of chess, but as an unprecedented solution to meet the climate and environmental challenges of tomorrow. For several decades now, artificial intelligence has become our best ‘Trojan horse’, our fidus Achates when dealing with the major issues of our era: the deterioration of biodiversity, climate change, socio-economic inequalities, and poor energy management. Yet, behind this impressive technological innovation hides a certain number of limitations that should not be disregarded.
What is AI ?
Throughout contemporary society, we are witnessing the beginnings of a technological revolution that is profoundly transforming our daily lives and altering every facet of our existence - whether it be our social, personal or professional life. Through the highly efficient performance of its algorithms, artificial intelligence unknowingly controls 87% of the daily activities carried out by the average human being. Despite the ominous presence of AI, when hearing about this innovation, it is not always easy to understand its inner workings, leading us to lose grasp of the impact its actions possess in our modern societies.
Artificial intelligence can be defined in the following way: it corresponds to a constellation of different technologies, which work together to allow machines to perceive, understand, behave, and learn at levels of intelligence comparable to those of humans.
Our journey with AI began in 1954, when some of the brightest minds in mathematics, computing, and physics gathered at Dartmouth University to discuss the potential for computers to think and behave like humans. These ideas ultimately proliferated in academia and industry, promising breakthroughs in every sector of society.
Fast forward to today, and computing and datasets have become available in every direction. This greatly facilitates the use of AI, with algorithms being capable of analyzing examples in order to infer their own rules and patterns about what they perceive.
Since the dawn of the 21st century, the performance of computer systems paired with the availability of data have enabled teams of researchers and pioneers (most notably Yann LeCun, Yoshua Bengio, and Geoffrey Hinton) to give birth to phenomena such as machine learning, deep learning, reinforcement learning, and predictive analytics.
Nevertheless, the meteoric rise of AI has surged alongside growing anxieties about disturbing scientific predictions of climate change. Having become one of the most pressing issues of our time, the environmental question (global warming, depletion of resources, deterioration of biodiversity, etc) is at the heart of political and social debates. As this requires rapid action spanning many communities, approaches, and tools, all solutions are sought to slow down the frightful figures and allow the smoothest possible ecological transition.
Artificial Intelligence has been proposed as one such solution, with significant opportunities to accelerate climate action via applications such as forecasting solar power production, optimizing building heating and cooling systems, pinpointing deforestation from satellite imagery, and analyzing corporate financial disclosures for climate-relevant information.
At the same time, AI is a general-purpose technology with many applications across society, which means it has also been applied in ways that impede climate action both through immediate effects and broader systemic effects.
Nature & Biodiversity Conservation
My discussion with Dr. Emmanuel Letouzé (Researcher à Harvard, MIT, UC Berkeley, Director of Data-Pop Alliance)
"Artificial intelligence is both an instrument and an inspiration to improve the way societies around the world contribute to the fight for our planet."
An additional way in which AI contributes to the fight against climate change is through the preservation of biodiversity, as well as the minimization of natural hazards and depleting resources.
Indeed, we are collecting ever-growing amounts of satellite data about the Earth’s climatic conditions, which allow us to analyze the patterns of weather evolution, hurricanes, floods, and various other natural disasters.
For instance, researchers at Stanford University have used deep learning to develop seismic algorithms that are capable of predicting earthquakes at over 86% accuracy - saving thousands of lives on a regular basis.
Industry giants from Silicon Valley are actively using AI to anticipate floods throughout eastern Asia, by creating digital models of the affected terrain; thus enabling these environmental benefits to be adaptive and long-lasting. In effect, as mentioned by Dr. Emmanuel Letouzé :
"In such a turbulent society, it is necessary to prioritize flexibility in order to democratize AI for all ecosystems."
AI also helps in preventing the extinction of endangered plants and animals, with object detection and recognition applications permitting us to identify waste materials in the ocean, locate poachers, and assist wildlife conservation authorities to keep their population under observation.
As well as prevention, AI also ensures the effective treatment and response to environmental risks. By allowing us to understand this social system, AI optimizes solutions such as where to dispatch first responders, which hospitals to put on high alert, and where to send relief supplies. Over the last five years alone, 120 thousand lives have been saved thanks to these applications.
Although the full potential of artificial intelligence has not yet been uncovered,
the use of these technologies in natural disaster prediction and treatment, combined with the conservation of biodiversity undeniably opens up a vast number of possibilities for supporting governments in their understanding and perception of environmental issues. By appropriating these scientific tools, we are now able to monitor the Earth’s evolution more precisely than ever, enabling us to respond appropriately and minimize human impacts.
Optimization of Energy and Efficiency
My discussion with Dr. Antoine Bordes (Research Director at Meta AI, Facebook, CNRS)
“The biggest challenge for renewable energy is storage; that is precisely the problem that AI can solve.”
In effect, AI can improve algorithms for electricity planning and storage as well as microgrid management in areas where there are decentralized systems. AI can improve the operations of renewable energy generators (like wind turbines, and solar panels) and can locate methane leaks in natural gas pipelines. AI is also being used to accelerate the discovery of new energetic materials, such as those used in photovoltaic cells, batteries, and electrofuels.
“There is a tremendous amount of research going into what are called energy cycles; this basically translates to how to reuse energy in a ‘virtuous circle’ that affects every aspect of the environment.”
AI enables significant reductions in electricity emissions, across a wide range of applications. In order to balance power grids effectively and enable the integration of large amounts of renewable energy, it is essential to forecast both electricity supply and demand - a function that AI can provide.
Furthermore, AI possesses the power to increase the efficiency of energy use in buildings and urban environments. Within smart buildings, it can optimize large constructions such as heating and lighting to save energy. For city-wide optimization, AI can be used in soft sensor systems and data mining. It can also help cities in their waste management, by reducing methane emissions associated with landfill and wastewater.
AI also improves decarbonizing transport in several manners. Most notably, it can improve estimates of the use of means of transport and their infrastructure. AI is also able to optimize freight routing and planning and can increase the use of low-carbon options such as trains. To advance EV (electric vehicle) adoption, AI can optimize charging and slot protocols - it serves as a guide to inform the public of innovations in the design of new generation batteries and fuels.
“The whole difficulty of energy problems in AI is to find the optimal chemical structure, it is fundamental. […] But finding it is like looking for a needle in a haystack: it’s the AI that tells us how to do it effectively.”
AI is increasingly being put to use in the discovery of materials such as catalysts, which can reduce the energy requirements of certain chemical processes. Artificial intelligence systems can also help optimize recycling processes and waste sorting for energy-intensive materials such as aluminum and steel which, in turn, helps reduce emissions associated with mining and material processing.
Limitations
My discussion with Dr. Emmanuel Bacry (Research Director at Ecole Polytechnique, ENS, CNRS)
“AI requires a lot of data, mostly personal. Our freedoms are therefore slaves to AI, which poses a lot of dangers in the future and prevents the healthy development of these technologies.”
However, despite the benefits in the realms of energy, biodiversity, and social development, the deployment of AI technologies could also trigger widescale environmental harm - if we don’t use them responsibly.
For instance, a plethora of decarbonization experts considers AI computational resources to have a high cost, thus pushing research largely into the private sector.
As well as this, the storage and processing of data needed to fully train a large algorithm can consume huge amounts of energy - as much as 300 thousand kg of carbon dioxide, according to studies carried out at MIT. That is the equivalent of nearly five times the lifecycle emissions of an American car.
Some estimates even say that computing will account for up to 8% of the world's total power demand by 2030, raising fears this could lead to the burning of more fossil fuels.
Consequently, AI could lead to the potential logging of people's energy use, which throws up major privacy concerns about what we can consider the ability to "backtrace data to individuals". In effect, Dr. Emmanuel Bacry considers that :
“Our data does not belong to us. With the leaps and bounds of AI, certain aspects of society are improved, but there are still more elements that are put at risk on an individual level. So we all have to be vigilant, all the time.”
Thus, we must not be completely blinded by the immense predictive and analytical power that AI offers us. Indeed, the low interpretability of results - as well as the colossal energy consumption and carbon emission - make the line between AI’s benefits and destruction thin for the future. However, that's no reason to be discouraged; on the contrary, without feeling powerless in the face of these dangers, we must remain responsible when developing and using AI.
A More Responsible AI for a More Pure Planet
Although there is no definite answer as to the manner in which AI technologies will evolve, one thing is certain: as a society, we must embed responsible AI principles into the design of initiatives and innovation structures, which includes fostering the inclusion of participants from civil society, the Global South, and marginalized groups.
As the use of AI grows rapidly across society, it is imperative that leading institutions and governments be proactive in helping shape these developments with climate action in mind.
Within individual countries, meaningful action on these initiatives will require collaborations among multiple branches in addition to participation from civil society, academia, and the private sector. It is also paramount to foster climate-cognizant impact assessments of AI via the collection of relevant data and by establishing standard measurement and reporting frameworks.
An example of an establishment taking prominent action is the Institute for Human-Centered Artificial Intelligence, a research center at Stanford University, which has proposed green AI principles requiring full transparency on a tech company's carbon, energy, and environmental impact as well as the integration of tech and climate regulation.
On the technical side, future years will see exponential technological growth supported by Moore’s law enabling researchers and engineers to continue developing and expanding the impact scope of AI. New breakthroughs in the realms of Deep Learning and Artificial Neural Networks will heighten the effectiveness of environmental tools. According to Demis Hassabis (Founder & CEO of Google DeepMind) :
“We are on the cusp of a new era of scientific discovery, where we’ll be able to use AI and Machine Learning to help in areas of science”
Will AI manage to conclude our planet's perilous odyssey, or is earthly prosperity just an Ithaca, an out-of-reach illusion?
Written by Mickael Naouri
Bibliography
1: INTRODUCTION to the STANFORD AI INDEX REPORT 2022.
2: Chan, Andy. “Artificial Intelligence and the Future of Work".” TED, 1 May 2019,
www.ted.com/talks/andy_chan_artificial_intelligence_and_the_future_of_work.
3: Tegmark, Max. “How to Get Empowered, Not Overpowered, by AI.” Www.ted.com, 1528, www.ted.com/talks/max_tegmark_how_to_get_empowered_not_overpowered_by_ai/ transcript?language=en.
4: Chui, Michael. “Applying Artificial Intelligence for Social Good.” McKinsey & Company, 2018,
www.mckinsey.com/featured-insights/artificial-intelligence/applying-artificialintelligence-for-social-good.
5: ibid
6: ibid
7: De, Préface, and Myriam Maestroni. INTELLIGENCE ARTIFICIELLE.
8: Life 3.0 : Being Human in the Age of Artificial Intelligence. London, Penguin Books, 2018.
9: Hao, Karen. “Training a Single AI Model Can Emit as Much Carbon as Five Cars in Their Lifetimes.” MIT Technology Review, 6 June 2019, www.technologyreview.com/2019/06/06/239031/training-a-single-ai-model-can-emit-as-much-carbon-as-five-cars-intheir-lifetimes/.
10: Silvestro, Daniele, et al. “Improving Biodiversity Protection through Artificial Intelligence.” Nature Sustainability, 24 Mar. 2022, pp. 1–10, www.nature.com/articles/ s41893-022-00851-6, 10.1038/s41893-022-00851-6.
11: ““AI for Climate Action” German Pavilion@COP26.” Www.youtube.com, www.youtube.com/watch?v=3mXPclqLdeo.
12: Fjeld, Jessica, et al. “Principled Artificial Intelligence: Mapping Consensus in Ethical and Rights-Based Approaches to Principles for AI.” SSRN Electronic Journal, 2020, 10.2139/ ssrn.3518482.
13: “Artificial Intelligence: Can It Help Achieve Environmental Sustainability?” Earth.org - Past | Present | Future, earth.org/data_visualization/ai-can-it-help-achieve-environmentalsustainable/#:~:text=AI%20has%20the%20potential%20to.
14: Stein, Amy L. “Artificial Intelligence and Climate Change.” Papers.ssrn.com, 2 Aug. 2020, papers.ssrn.com/sol3/papers.cfm?abstract_id=3665760.
15: “Shrinking Deep Learning’s Carbon Footprint.” MIT News | Massachusetts Institute of Technology, news.mit.edu/2020/shrinking-deep-learning-carbon-footprint-0807.
16: ibid
17: Malone, Thomas. ARTIFICIAL INTELLIGENCE and the FUTURE of WORK. 17 Dec. 2020.
18: “Training a Single AI Model Can Emit as Much Carbon as Five Cars in Their Lifetimes.” MIT Technology Review, 6 June 2019, www.technologyreview.com/2019/06/06/239031/ training-a-single-ai-model-can-emit-as-much-carbon-as-five-cars-in-their-lifetimes/.
19: “Environmental Intelligence: Applications of AI to Climate Change, Sustainability, and Environmental Health.” Stanford HAI, hai.stanford.edu/news/environmental-intelligenceapplications-ai-climate-change-sustainability-and-environmental.
20: Galaz, Victor, et al. “Artificial Intelligence, Systemic Risks, and Sustainability.” Technology in Society, vol. 67, Nov. 2021, p. 101741, 10.1016/j.techsoc.2021.101741. Accessed 19 Oct. 2021.
21: “Demis Hassabis, Academy Class of 2017, Full Interview.” Www.youtube.com, www.youtube.com/watch?v=1X7Koxx4qJE. Accessed 9 May 2022.
Comments