Let’s be perfectly honest. Ruins are awesome. There is something utterly fascinating about the idea…
From customer service chatbots to virtual personal assistants to self-driving vehicles, artificial intelligence and machine learning is fast becoming a mainstream solution to everyday challenges. So, it was only a matter of time before scientists started enlisting its endless uses in the study of climate change and energy consumption. Could this be a good thing?
After Google acquired the artificial intelligence laboratory DeepMind in 2014, it committed funding to the study of global warming and energy consumption. By applying their machine learning to Google data centers, DeepMind managed to reduce the energy consumption used for cooling Google’s machines up to 40% by using a system of neural networks (computer programs inspired by the brain).
In terms of emission reduction, this was not an insignificant achievement. Improving the efficiency of energy consumption is one of the first steps in tackling the issues of global warming, if not the primary concern of climate change scientists. Indeed, DeepMind’s success at reducing the energy used by Google’s vast power-hungry data centers even led to discussions with the National Grid plc about using their algorithms to balance the U.K.’s energy supply and demand. By using technology developed by DeepMind, artificial intelligence could learn how to predict demands placed on the network, which would better inform the National Grid about how to manage them and reduce energy wastage.
But it is not just the use of algorithms to limit energy consumption that artificial intelligence can be used for. The prediction of changes in weather systems because of climate change has historically been a tricky business. While the use of Numerical Weather Prediction has been around since the 1950s, its reliability is, at best, 50-50; meaning that errors, inefficiencies, and lack of quality data have led to challenges when it comes to predicting large-scale weather events with enough time to alert people in the affected area. However, the use of artificial intelligence is increasingly leading to more impactful predictions with meteorologists able to predict heavy storms and tropical cyclones with the data.
A team of researchers at the Lawrence Berkeley National Laboratory is training a neural network to study patterns in extreme weather so that it can make long-term predictions on the impact of climate change in the future. The Deep Learning for Climate project has been studying the patterns of extreme weather for the last three years in an attempt to better understand how climate might change in the future. Where humans struggle to effectively predict the complex interaction between global trends and local conditions, the use of artificial intelligence could lead to determining whether the recent spate of hurricanes was caused by global warming.
There are difficulties associated with the use of deep learning algorithms for climate change predictions, however, as it is still very much in its early stages of development. While AI does excel over humans when it comes to making decisions with incomplete information, it is also very bad at explaining how it arrived at its solutions. Although artificial intelligence has recently proven its capabilities in beating professional poker players, it still has yet to learn how to accurately handle scenarios that neural networks have not yet been trained for, which could cause difficulties when it comes to predicting previously unknown climate change events. Creating the perfect player in a strategy tournament is one thing, but developing technology that could anticipate potentially catastrophic events is quite another. After all, algorithms are only as effective as the training they have received; meaning that the initial labeling of data is crucial to the success of climate change predictions.
Despite the teething problems, it is difficult not to see the benefits of artificial intelligence when it comes to climate change and energy consumption. The applications for it are endless: Utility companies can plan for repairs in storm-affected areas while agricultural businesses can decide which crop would better survive during that season, big consumers of energy can also discover ways to reduce and improve their consumption, and wide-scale disasters can be reacted to faster, keeping more people safe. By using this technology, both mankind and the environment may end up benefiting from such a major step forward.