Artificial Intelligence (AI) is on everyone’s lips. The term was coined by John Mc Carty in 1955, since then, the technology has rapidly advanced. The technology of a machine capable to learn and think on its own has raised both hopes and worries for the future of humankind. As climate change is the biggest challenge facing the planet, what if we harnessed the potential of machine learning to bring solutions for climate action? A lot of discoveries remain to be found for us to truly understand what use we can make of it.
All orgs developing advanced AI should be regulated, including Tesla— Elon Musk (@elonmusk) February 17, 2020
The origins of a super-intelligence
To get a better understanding of how AI could help us to tackle climate change, we first need to understand what the term AI actually means. Artificial Intelligence is a field of science that points to situations in which a machine can simulate human intelligence. It can perform calculations, make interpretations or take decisions that typically require human knowledge. The latter includes the capability to learn from past experience, adapt to new situations and handle abstract ideas. Under AI, we have Symbolic Reasoning and Machine Learning. In Symbolic Reasoning, the human writes an algorithm for a machine on how to give an output based on an input. For example, in a Coca Cola factory, a sensor detects an empty bottle (data) by which it will move a robotic arm and fill the bottle with Coca Cola (output).
In Machine Learning (ML), the algorithm does not react to the input. Instead, a human writes code. Then, the machine ‘learns’ the algorithm to come to an output. With every new data, the engine will then improve its algorithm. For example, Softbank has developed its robot based on ML, Pepper can interact with humans by reading their emotions, can have a conversation and a throw a ball. This robot is already in use in companies and schools, but is still qualified as a “weak AI”. Now, we may wonder, could we utilize this advancement to develop robots to prevent food waste, or integrate sustainable living in our daily lives?
Deep Learning, a subcategory of ML, is the most sophisticated form of AI. It is inspired directly from the human brain as a neural network. Deep learning recreates the decision-making of humankind. Apart from these categories of AI, there is also a difference in specialist and generalist AI. In this article, we are talking about ‘specialists’ forms of AI, that can perform multiple tasks at the same time. For example, deep learning enables better predictions and estimates of climate change. Research led by Google, the Mila Institute, the German Aerospace Centre demonstrated the critical role that AI has in making sense of the extensive data set on Earth, understanding better climate change, and acting upon it. Down the line, it provides enhanced energy monitoring for buildings and electric cars. AI leads to sufficient energy distribution and battery management, enabling reduced carbon footprint and switching to renewable energies.
It is a “call to arms” to bring researchers together, AI is not a silver bullet against the climate crisis, but can help tackle climate change.David Rolnick, University of Pennsylvania.
A recent paper named “Tackling climate change with Machine Learning”, conducted by the most prominent researchers in the AI field, disclosed 13 areas where ML can be developed, such as energy consumption, CO2 removal, education, solar engineering and finance. The outcomes encompass monitoring deforestation, creating new low-carbon materials, and greener transportation. However, AI is still at its infancy stage and cannot solve everything.
Understanding climate change
First of all, AI can help us, in understanding climate change better. Everything from global-scale modelling to individual weather forecasting relies on a massive number of variables, which is impossible for a human brain to do on its own. The interpretation of climate data is based on climate informatics, a discipline created in 2011. It covers a wide range of topics, such as predicting extreme weather events, reconstructing past climate conditions, or the socio-economic impacts of climate change and precipitations. This can help policymakers to take action and save lives. If you as an individual or as a company want to benefit yourself from an improved AI weather forecasting, take a look here.
Nevertheless, the community of scientists and experts on climate modelling claim that models tend to agree in short-term forecasting while diverging for long-term forecasts. There is still a lot of uncertainty, as models do not even agree on how precipitations will change the future. For example, MILA researchers have created an app, to show individuals what their neighbourhood will look like with different climate change outcomes. People are able to upload pictures of bush fires or floods, to ameliorate the algorithm. This concept will be deployed across Canada.
Algorithms are not only getting better and better for specific weather events, but also for the more global changes and its consequences. An example is to predict the relationship between the measures we take and how fast we will go to the 2°C rise in global temperature. Instead of trying to write complex models based on physical laws (symbolic reasoning), the study of Ise and Oba gave global monthly temperatures of the last 30 years to a neural network. Without any other data, the neural network successfully predicts the rise and fall of warmth for the next 10 years, with an accuracy of 97%. AI could also help to understand the causes of climate change. It could, for example, based on satellite images, detect and map significant CO2 emitting sources in countries where the regulation about reporting is scarce.
Optimization of existing systems
A more concrete way in which AI can help us is in reducingCO2 emissions through the optimization of existing systems. ‘Climate Change AI’, a group of volunteers that wants to bring together AI experts and climate science specialists, identified how ML can help in different areas (e.g. electricity systems, transportation, buildings and cities, farms). For example, in electricity systems, that account for ¼ of the global CO2 emissions. Carbon Tracker is an independent financial think tank working towards the UN goal of preventing new coal plants from being built by 2020. It monitors coal plants emissions by using satellites data, and convince the finance industry that it is not profitable. Thanks to a grant from Google, Carbon Tracker is expanding the satellite imagery efforts to include gas-powered plants’ emissions and get a better sense of where air pollution is coming from. Carbon Tracker will analyze emissions for 4000 to 5000 power plants, creating the biggest data bank and make it public. This could help us in having a global perspective, on tackling carbon emissions and reducing air pollution. It will also pinpoint companies responsible for carbon emissions, and implement an emission price.
Further, then a prediction, the energy consumption can even be managed, e.g. by smart home solutions that charge your car or do your laundry when there is renewable energy available. In industry, AI can be used to optimize processes that require electricity. For example, the Google Data Centers consume 3% of global energy consumption (see also our article on datacenters). Deepmind implemented an ML system to optimize the management of all the settings to cool these datacenters, minimizing energy consumption. The ML continuously performs the optimal configuration of all the gauges and valves, based on its learnings from previous data, and like this, they could reduce the energy consumption with 40%! Microsoft has found another solution: self-sufficient underwater datacenters. Steered with AI, they are cooled by the ocean and powered by wave energy. Imagine that if similar systems are implemented in all industries, what a considerable amount of CO2 emissions could be avoided?
How to design a performant AI?
Next to optimizing existing systems, AI can assist in the design of new sustainable products and technologies. By switching from the traditional way of doing things to these new products and technologies, CO2 emissions can be avoided or even reversed. In designing a sustainable product, one has to take into account a significant number of factors like abundancy, toxicity and properties of raw materials, regulations, end-of-life options, price, … A human could not even imagine all the possibilities. Machine learning can incorporate all the constraints, go over all the possible materials and configurations and propose the most optimal combination. In this way, AirBus designed a new 3D printable aeroplane part that is not only lighter than before, reducing CO2 emissions while flying, but also stronger and using less raw materials.
Also, in research and development of technology, ML is getting more and more important. It is a promising tool in designing batteries with a longer life-span and higher energy storage capacity. It could also accelerate the research on nuclear fusion reactors, that can become a safe and carbon-free electricity production alternative. These require an intelligent experimental design as they have a large number of tunable parameters. Or if we dream bigger, and feed a deep learning algorithm with the right data about our universe, maybe it could help us to understand it better, or bring space travelling to a whole other dimension. No need to change the climate on the earth anymore. Thirty years ago we couldn’t even imagine how the internet was going to take over our lives, now, AI could maybe also make it possible to get a grip on our own future as humanity.
The challenge of the climate crisis can be viewed as an opportunity for forwarding thinkers tech entrepreneurs. However, all the solutions listed above are not “silver bullets” against climate change, it necessitates an international “brainiac” collaboration. All these examples mentioned above are just a tiny tip of the iceberg of how AI could change the tide and help us in the transition towards a sustainable, green, but maybe also unimaginable future. So, in contrast to what the Matrix predicts, AI can become one of the most essential tools in helping humanity to sustain our race. The only thing we need is a collaboration between climate specialists, engineers, AI specialists, entrepreneurs, and governments to use our collective knowledge to make it happen.
Technology is an ally of environmentalism. Plan A advocates for a data-driven approach to carbon footprint reduction. Monitoring your carbon emissions is primordial when starting a sustainable business. Join the conversation here.