AI can be both a force for good in helping towards a sustainable environment, but also a cause for problems in increasing the production of greenhouse gasses from the processing required. Our work focuses on using AI for a more sustainable world, but also in reducing the greenhouse footprint of our AI work.

Examples include:

Monitoring of Dolphin Populations

If we know the number of dolphins in an environment we can assess how 'healty' the environment is as dolphins are at the top of the foodchain. This may seem a simple task but as we can only see dolphins when they come to the surface this makes it hard. We need to photograph the dolphins and then try to identify if they are the same dolphin that we photographed earlier. In this way we can estimate population sizes. We can also look at the sounds that the dolphins make and see if these are unique to individual dolphins and can hence be used in the same way. 

Reducing energy consumption in the electriciy Grid

Can we apply machine learning to decide when to turn on electricity generation sources? Many green energy sources are less controllable than more carbon producing sources such as gas powered turbines. By using machine learning we can reduce the need for more carbon producing sources whilst increasing the use of green sources such as wind and electricity. We can also work out better stratergies for carbon taxing energy production to increase the uptake of green power.

Reducing energy usage in large computer systems

We work on reducing the energy usage of large computing systems and the Cloud by better selection of where work should be deployed. We also work in predicting the execution time of Deep Learning training and prediction. This allows people to decide if it makes sense to perform the training or if deploying deep learning onto power-restricted devices whether this will work.