How AI Architectures are Improving Sustainability

Article by Dr Anwita Maiti, Ethical AI Researcher, deepkapha AI Lab, The Netherlands

AI architectures work on how to shape AI that would be more flexible and feasible by combining the factors of information and infrastructure. Any AI application has two aspects- technical and societal. For AI to become more popular and people-friendly, where they would give their approval of trust, it is always a criterion to see the final test of how the effect is being felt- socially, economically and ecologically. With rising awareness about climate change and a foreboding grievous ecological abrasion, AI architectures have a lot of responsibility to shoulder in terms of sustainability.

Architectures need to be built by keeping frameworks in mind. In order to bridge the technical and societal aspects of AI, one must work on the still de facto and the growing gap between science and policy. Socially beneficial AI frameworks have indeed been made in recent years, but it is still predominantly in the de jure state. In order to translate theory into praxis and see a visible and impressive positive social and ecological difference, we should always keep in mind to give equal weightage to both science and policy.

To gear toward sustainability, AI has to overcome a few challenges like:

  1. Overreliance on machine learning data- that should actually be renewed periodically
  2. Be prepared to be relevant under the pressure of uncertain human behavioural responses
  3. Endangerment of cybersecurity
  4. Unpreparedness to deal with sudden interventionist strategies

To address these issues, we need to implement more cost-effective design thinking, emphasise sociological and psychological AI impacts in the short run, and consider environmental safety in the long run.

Aspects of Sustainable AI

The concept of ‘Sustainable Development’ was first initiated in 1987 by the World Commission on Environment and Development through its Brundtland Report. While talking about sustainability- we usually mention the domains of environment, society and economy, all of which rely on the final straw of how political parties in power execute policies.

Coming to sustainable AI, let us look into the following aspects:

  1. Environmental Sustainability. Carbon footprint is a major area of concern when it comes to the environmental and ecological implications of AI. Big companies like Google and Amazon have aimed to reduce their carbon footprint by 2030 and use more renewable energy instead. In collaboration with AI enterprises like BCG GAMMA and, and Haverford College, Montreal Institute for Learning Algorithms (MILA) has designed a tool called ‘CodeCarbon’ that can be merged into the Python codebase and detects how much carbon dioxide is emitted while training machine learning models. Larger language models usually run on high electricity; for example, the MILA carbon footprint could be considered in terms of reducing carbon footprint. Energy could be optimised by combining machine learning, natural language processing and computer vision. In agriculture, AI could efficiently reduce fertiliser and water usage, and, when it comes to energy, AI could predict weather changes from beforehand and develop more cost-effective and lesser carbon-emitting models based on renewable energy. When we think about it, this AI-based weather prediction could, in turn, reduce water usage as well. In facility management, AI could predict the amount of energy required based on the number of people living in a building, hence leading to optimisation.
  • Climate Change. We are already contributing a giant leap toward reducing our carbon footprint through solar and wind energy usage. AI applications could build more energy-efficient buildings where electrical devices like light, fans or air conditioning which are not in use could be automatically switched off. Electrical automobiles also reduce carbon emissions. AI apps that have introduced shared cabs contribute towards easing traffic and curtailing emissions. By using tree-planting drones, we could aid afforestation, and by studying maps, graphs and photographs, AI could look into the distressing concern of sea level rising and glacier melting. AI could also curtail releases from steel and cement factories by making low-carbon materials made by the faster and more extensive combination of chemical compounds; otherwise, a researcher or a scientist would have a longer time to reach.
  • Smart Cities. The United Nations has noted that by 2050, 68 % of the world’s population will be living in urban areas. The thought that naturally dawns is how well equipped are we to host this population in so much limited space that comprises just cities. With the help of smart technologies, 5G connectivity and the usage of IoT (Internet of Things) technology, AI can form an idea about the wants and needs of dwellers. With regard to public safety, CCTV cameras have been a great help in detecting crimes and predicting threats. AI works with real-time data and helps out people from falling prey to unpredictable traffic by providing them with information about where and at what time they can park their vehicles. With regard to waste management, smart bins notify authorities when they get filled up. As has already been discussed, sensors can predict the energy required in a building depending on the number of people residing. They can detect which areas the habitants move most towards during what time of the day and which areas they do not enter, thereby optimising energy usage.

Conclusion: Challenges and Promises of a Hopeful Change

AI architectures are already paving the way toward making the earth more sustainable, but what we have to achieve, as has already been mentioned, is to bridge the gap between science and policymaking. Considering social, political, economic and cultural factors, we must keep in mind to introduce AI applications in public by warmly introducing people to them and explaining their benefits and how they work. Technology should not be imposed on people, but maybe the government and local bodies should take responsibility for initiating AI. The sudden implementation of AI applications without citizens’ consent might result in adverse social and behavioural reactions, creating a sense of disharmony in society.

Coming to the positives, as we can already see, AI is helping the environment by working on lesser carbon emissions, reduction of carbon footprint, and seeking solutions to control climate change. Within the public- it is fostering security and playing a major role in smart cities by easing people’s lives through apps. In the industrial sector, robots are helping in regulating waste management and energy consumption as they are better able to detect optimal needs with fair accuracy. Finally, AI has become an indispensable part of smart cities, where we need its functioning as we are challenged with hosting and giving quality lives to a gradually intensifying population within proportionally much lesser space.

About the Author.

Anwita Maiti is an AI Ethics Researcher at deepkapha AI Research Lab. She hails from a background in Humanities and Social Sciences. She has obtained a PhD. in Cultural Studies.

Skip to content