This bite-sized article focuses on environmental modelling with AI. We’ve discussed the risks and dangers of data services in previous articles which you can check out here. This is a very rich topic, so we definitely recommend looking at the reviews and resources we’ve cited here for more information!
Climate modelling is a huge task for policymakers. When it’s done right, and with enough complexity to come close to the real environmental system, it can add huge accuracy and credibility to policy recommendations. The pattern matching required for this is – quite literally – what learning models (AI) are made for, so it’s no surprise that environmental modelling has become a prime application.
From as early as 2009, reviews showed that neural networks, a type of AI model, were used to understand and project vegetation, fish, and microbial distribution . Skip to 2019 and the equivalent review states clearly why these methods are so appropriate . Part of this is their ability to “know what they don’t know” – to give an uncertainty to every prediction they make. Recalling our article on the IPCC (our first article!), this is very important for systems as complex as the environment.
As Yale Environment has put it, “deep learning might be a solution for areas of the climate picture for which we don’t understand the physics.”  Deep learning picks up patterns in data with an ‘intuition’ that computers normally cannot have. It doesn’t need rules and equations to tell it how to predict the environmental future – this would limit it to only what we know (although there is always risk in leaving them unsupervised – they can be blatantly wrong sometimes ).
The use and abuse of information technology poses a great immediate threat to the environment, and AI and data is center-stage in this. But as research money and commercial investments are poured into this technology, it is crucial that fighting climate change gets a share of the benefits.