It’s been a momentous few weeks for the community involved in applied machine learning (ML), with the field being recognised by not one, but two Nobel Prizes.
Celebrating the awards
The 2024 Nobel Prize in Physics was announced on 8 October, with the award shared by John Hopfield of Princeton University and Geoffrey Hinton of the University of Toronto. The citation points to “foundational discoveries and inventions that enable machine learning with artificial neural networks”. Hopfield created a structure to store and reconstruct information that is fundamental to the operation of machine learning. Hinton invented a method to independently discover properties in data.
Just one day later, the Nobel Prize in Chemistry was awarded to David Baker of the University of Washington and the Howard Hughes Medical Institute and Demis Hassabis and John Jumper of Google DeepMind for their work on applying AI models to protein design and protein structure prediction, leading to critical insights in life science research.
At Intellegens, we celebrate and acknowledge the contribution of these pioneers, and others like them who have driven the development of ML and its applications. We would not get the enjoyment and value that we do from our work without their collective contributions. With our roots here in Cambridge, it’s also good to see the University of Cambridge noting (here and here) its connections to Hinton, Hassabis, and Jumper. The University has been an incubator for much AI talent.
What does this news tell us about our own work?
Machine learning is a proven technology – as the Nobel committee’s background article on the Physics Prize outlines, the origins of machine learning date back to the 1940s, and Hopfield’s work on networks is now forty years old. Although machine learning can sometimes be seen as a novelty, it is in fact now a very well-established set of methodologies, a standard course in university curricula, and embedded in systems and methods that are fundamental to our everyday lives.
Innovations in methodology drive success – it takes advances in methodology to create the basis for tools that, when appropriately applied, deliver the benefits of ML. Our own small contribution to this process at Intellegens has been to further develop the work of Dr Gareth Conduit and collaborators, also at the University of Cambridge, which created the Alchemite™ algorithm. This can quickly generate machine learning models using training data that is ‘sparse’ (i.e., contains gaps), a scenario where other ML methods fail.
The next step is to make machine learning a productivity tool – the use of ML methods in day-to-day R&D still faces constraints. They are usually applied by researchers who become experts in their use – coding around them and constructing their own models tailored to the data and applications they are targeting. The next step is to create routine ML productivity tools that empower data analytics and experimental design by scientists who are not ML specialists, so that we no longer remark on the use of AI or ML in a project any more than, for example, we would highlight the use of Excel. This is our vision at Intellegens.
So, congratulations and thanks to our Nobel pioneers. We’ll continue to do our piece at Intellegens to build on their foundations.