Celebrating its 55th anniversary in 2024, the MacRobert Award, the UK’s longest running and most prestigious prize for engineering innovation, continues to attract a high calibre of submissions each year.
With a £50,000 prize, gold medal and opportunity to sit amongst an illustrious group of winners including Rolls Royce, Raspberry Pi and Ceres Power, the MacRobert judging panel has the unenviable task of selecting just three finalists each year to be publicly recognised for their extraordinary innovations.
Once an organisation is shortlisted as a finalist, the MacRobert Award judges take part in site visits to meet each team, explore the facilities, and get under the skin of each innovation. Following the recent announcement of the 2024 MacRobert Award winner, we look back at this year’s judges’ visits to each of our 2024 finalists and discover more about what made each innovation so exceptional.
Google DeepMind
The MacRobert Award judges’ visit began on a sunny May morning outside of London Kings Cross St Pancras station. Their aim? To visit Google DeepMind to learn more about their GraphCast technology.
GraphCast was shortlisted for its use of AI to provide significantly faster and more precise weather forecasts, utilising machine-learning techniques to surpass traditional numerical weather prediction methods.
Following a tour of the DeepMind offices, the judges settled in for the first presentation of the day, led by Peter Battaglia, Research Scientist, and Meire Fortunado, Staff Research Scientist. When assessing the extent to which a shortlisted technology can be considered innovative, MacRobert Award judges want to see the innovation in action and judge with their own eyes.
Their wishes were granted when Peter introduced the session with a live demonstration of GraphCast in action, providing a highly localised, accurate forecast in under 45 seconds; we shouldn’t have been surprised that the imminent London forecast was for rain. The speed and efficiency of this demonstration hit home to the judges the extent to which GraphCast outperforms traditional forecasting technologies, which comparatively take over an hour, via a supercomputer.
Peter explained that traditional numerical weather prediction relies on solving a set of complex equations and uses increased resources to improve forecast accuracy. However, it cannot directly use historical weather data to improve the underlying model. Enter – GraphCast. Their machine-learning-based method is trained directly from historical data, which as Meire states “represents a paradigm shift in weather forecasting.”
The second presentation of the day addressed one of the key criteria of the award - the societal benefits the innovation will provide. The impact of improved weather forecasting was obvious; by outperforming leading models in predicting extreme weather events through cyclone tracking and extreme temperature forecasting, GraphCast enables earlier warnings and more informed decision-making from governments. This translates to potentially saved lives, reduced property damage, and faster response times to forecasted extreme weather events.
Beyond the benefits it can bring to disaster relief efforts, the judges learnt of the widespread implications for both environmental sustainability (more accurate predictions present opportunities to maximise renewable energy sources such as solar and wind) and the support for more precise agricultural practices.
Perhaps not so well known prior to the visit is that GraphCast’s technology is open source, thus allowing more widespread access to high-performance weather forecasting. Institutions and researchers without supercomputers can now leverage GraphCast, potentially leading to localised weather models.
As the visit concluded, the MacRobert Award judges were left in no doubt as to the essential nature of GraphCast’s technology and the importance of this groundbreaking work to better understand how we collect, analyse, and respond to weather patterns.
What did the judges think?
Better weather forecasting will be increasingly crucial for predicting climate changes and enabling timely interventions. During our visit to Google DeepMind, we were impressed by GraphCast’s ability to deliver highly localised, accurate forecasts in under 45 seconds—significantly faster than traditional methods. The live demonstration we had truly had the 'wow' factor.
GraphCast’s innovation not only enhances accuracy but also provides critical early warnings for extreme weather events, potentially saving lives and reducing damage. The open-source nature of the technology ensures broader access, allowing institutions worldwide to benefit. This groundbreaking work will play a vital role in shaping our response to climate change.
Dr Steve Allpress FREng, MacRobert Award judge
About the MacRobert Award
The MacRobert Award is the UK’s longest running and most prestigious prize for engineering innovation. The winning team is presented with a £50,000 prize, a gold medal, national publicity, and a weekend away at Douneside House in the heart of the MacRobert Trusts estate in Aberdeenshire.
Nominations for the 2025 MacRobert Award are now open until 31 January 2025. Please contact Patrick Woodcock, MacRobert Award manager, on 02077 660 630 or [email protected] should you wish to discuss putting forward a nomination.