The Eric and Wendy Schmidt AI in Science Postdoctoral Fellowship, a programme of Schmidt Futures, aims to accelerate the next scientific revolution by supporting talented postdoctoral researchers to apply AI techniques across the natural sciences, engineering and mathematical sciences. The new initiative adds to Schmidt Futures’ existing philanthropic efforts to support the development and application of AI in innovative ways.
As an initiative spanning the full breadth of Mathematical, Physical and Life Sciences (MPLS) Division, the programme is not ‘owned’ by a single department and is entirely cross-disciplinary, and in this way has already begun to bring together different parts of the AI landscape in novel ways, sparking interactions from departments and fields less traditionally involved in AI research (for example, the Department of Biology). A key element of the AI in Science programme is the training aspect; the programme will develop a suite of training that provides the Fellows with all they need as non-AI specialists to use AI in their respective research fields. In future the aim would be to make this resource more widely available to all researchers at Oxford.
All of Oxford’s Eric and Wendy Schmidt AI in Science Postdoctoral Fellows have the opportunity of becoming Associate Research Fellows at the University’s newest College: Reuben College. Founded in 2019, Reuben College aspires to create a community of scholars embracing opportunities for interdisciplinary collaboration, and developing initiatives to generate wider impacts and positive benefits from research, entrepreneurship and public engagement. Reuben College’s students and Fellows are loosely clustered within four themes of the College ‘Cellular Life’; ‘AI & Machine Learning’, ‘Environmental Change, and ‘Ethics & Values’ or within the strategic priorities of ‘Public Engagement with Research’, and ‘Innovation & Entrepreneurship’. The College is looking forward to the contribution of the Eric and Wendy Schmidt Associate Research Fellows to its interdisciplinary, ‘start-up’ culture.
The first cohort of ten Schmidt AI in Science Fellows have now been recruited and were formally welcomed to the University at a joint event with Reuben College at the Museum of Natural History on 4 May. Each of them had the opportunity there to outline the focus of their proposed research; the ten projects are summarised below.
Professor Sam Howison, Head of Division for the Mathematical, Physical and Life Sciences, said: “It’s an exciting time to be part of the AI revolution and wonderful to see the ten Schmidt AI in Science Fellows bringing together different parts of AI in creative ways and across the whole MPLS Division.
Shuxiang Cao
“Efficient and automated calibration of superconducting processors using AI techniques”
The project aims to develop an AI to automatically calibrate the changes in device parameters of a quantum processor, to maintain its stable performance. This will significantly benefit both academia and industry for controlling large-scale superconducting quantum processors and has the potential to be adapted to other types of quantum processors.
Qi Hu
“Embedded AI for next generation biomedical optical imaging systems”
Optical microscopes suffer from distortions introduced by imperfect hardware and innate sample structures that can detrimentally affect image quality. Adaptive optics (AO) uses reconfigurable devices to correct aberrations, but such methods have limitations. The work will embed neural network algorithms into the feedback control of the AO system and lead to improvements in performance.
Holly Pacey
“Maximising LHC discovery potential with graph neural networks (GNNs)”
The LHC's ATLAS experiment collides protons at near-light speeds, with the goal of discovering evidence for beyond Standard Model (BSM) particles. However, BSM searches to date use only properties of each individual collision to discriminate between BSM- and SM-like collisions. The project will use graph neural networks (GNN) analysis to revolutionise ‘anomaly detection’.
Carlos Outeiral Rubiera
“Protein expression optimization using large language models”
Engineered proteins, a crucial ingredient in vaccines, medicines, and diagnostic tests, pose production challenges due to the complex DNA 'dialects' used by protein-producing microbes. Even small dialectal mismatches in the engineered DNA can impair the metabolic machinery and greatly reduce yield. Our project will use AI to 'translate' between DNA dialects to enhance protein production, aiming to show the potential of artificial intelligence in synthetic biology.
Heloise Stevance
“A Virtual Research Assistant for the next generation of sky surveys”
Starting in 2024, the Rubin Observatory will record the most exhaustive sky-survey humanity has ever undertaken: it will detect the explosions of stars hundreds of millions of light-years away, seeing 10 million new events each night. The project will create a model that is intelligent enough to serve as a virtual research assistant and allow human scientists to fully exploit this revolutionary source of astrophysical data.
Tianning (Tim) Tang
“Intelligent wave breaking characterisation with machine learning”
The project will use machine learning (ML) to characterise the initiation of wave breaking and seek a novel equation-based description of the wave breaking discovered by ML. The ultimate aim is to discover the formulation describing the breaking process after the initiation. This would have benefits in ocean engineering but also contribute to the ultimate question: can we discover physics with ML?
Jake Taylor
“Modelling Exoplanet Atmospheres with Machine Learning”
Modelling the atmospheres of exoplanets is a computationally demanding task and becoming more demanding in the James Webb Space Telescope era. The project will use machine learning to emulate the calculation of molecular opacity. Combining this new tool with GPU programming will significantly improve our ability to interpret the atmospheres of these alien worlds.
Elnaz Azizi
“Translating grid substations data into actionable information via unsupervised load monitoring”
Extracting useful knowledge from the aggregated load at substations level plays an important role in the planning and operation of the distribution grid – enabling smart solutions for demand-side energy management as well as fault detection and recovery. The research aims to develop a learning-based method to extract information from existing grid measurements (active and reactive power, voltage and current).
Rachel Parkinson
“Employing AI to identify the complex interactions of environmental stressors on pollinator health”
There is a critical need to investigate how environmental change affects pollinator behaviour so that steps can be taken to mitigate economic and ecological risk. Sound is a typically overlooked component of behaviour, and the project will result in a powerful tool for integrating computer vision and sound to automatically track the behaviour of insects. The technology will have application in the risk assessment of environmental stressors.
Richard Creswell
“Applications of AI/ML to epidemiological time series”
Infectious diseases such as COVID-19 represent an important target for mathematical models which describe how a disease progresses through a population. A model’s parameter values may be informed by prior knowledge of the disease, but often they must be inferred from data such as the daily number of deaths or weekly number of positive tests. This raises numerous challenges which will benefit from the application of advanced techniques from AI and ML. The lessons learned are expected to apply directly to other problems.