Anna Scaife, University of Manchester, £1,458,618 AI4Astro: AI for Discovery in Data Intensive Astrophysics The data rates from modern scientific facilities are increasing and it is no longer possible for scientists to extract scientifically valuable information by hand. This is particularly important for the Square Kilometre Array (SKA) telescopes. Consequently, in this era of big data astrophysics using machine learning to extract scientific information is essential to realise a timely scientific return from facilities such as the SKA. This research will consider how existing techniques can be adapted to achieve the key scientific goals of the SKA telescope. It will target the development of new machine learning approaches which address key aspects of SKA scientific processing. Maria Liakata, University of Warwick, £1,227,974 Creating time sensitive sensors from language & heterogeneous user generated content Widespread use of digital technology has made it possible to obtain language data (social media, SMS) as well as heterogeneous data (mobile phone use, sensors). Such data can provide useful behavioural cues on an individual level and for the wider population, enabling the creation of longitudinal digital phenotypes. Current methods in natural language processing (NLP) are not well suited to time sensitive, sparse and missing data or personalised models of language use. The proposed research will address specific challenges within NLP, will have an application to AI and mental health, and outputs will include novel tools for personalised monitoring behaviour through language use and user generated content over time. Neil Lawrence, University of Cambridge, £2,380,212 Innovation to Deployment: Machine Learning Systems Design The AI systems this project is developing and deploying are based on interconnected machine learning components. This research focuses on AI-assisted design and monitoring of these systems to ensure they perform robustly, safely and accurately in their deployed environment. This project addresses the entire pipeline of AI system development, from data acquisition to decision making, and proposes an ecosystem that includes system monitoring for performance, interpretability and fairness. This project places these ideas in a wider context that also considers the availability, quality and ethics of data. Timothy Dodwell, University of Exeter, £1,336,283 Intelligent Virtual Test Pyramids for High Value Manufacturing There is a paradox in aerospace manufacturing. The aim is to design an aircraft that has a very small probability of failing. Yet to remain commercially viable, a manufacturer can afford only a few tests of the fully assembled plane. How can engineers confidently predict the failure of a low-probability event? This research will develop novel, unified AI methods that intelligently fuse models and data enabling industry to slash conservatism in engineering design, leading to faster, lighter, more sustainable aircraft. Yarin Gal, University of Oxford, £1,379,408 Democratizing Safe and Robust AI through Public Challenges in Bayesian Deep Learning Probabilistic approaches to deep learning AI, such as Bayesian Deep Learning, are in use in industry and academia. In medical applications they are used to solve problems of AI safety and robustness. But major obstacles stand in the way of widespread adoption. This project proposes building new AI challenges to assess safety and robustness, derived from applications of AI in industry. The challenges identified will set the course for a community-driven effort leading to a self-sustained ecosystem and will bridge practitioners and AI researchers. This will offer new research opportunities for the AI community, helping to develop new safe and robust AI tools. Democratising safe and robust AI aligns with the UK's strategic plan set by Hall and Presenti and will help put the UK at the forefront of AI globally.

Technologies defining Artificial Intelligence (AI), and therefore, research in these technologies of AI, will be an all-pervading underpinning of future development in many sectors. The last decade has seen a dramatic rise in computer power, an explosion in data, and scientific break-throughs. Together, these advances have led to the emergence of data science and the resurgence of AI - 'machines that think', as imagined in Alan Turing's landmark research paper published in 1950. The role of the Alan Turing Institute (the Turing) is to foster these sciences and grow them in order to change the world for the better. The Turing was established by EPSRC in response to a letter to the Prime Minister from the Council for Science and Technology entitled 'The Age of Algorithms' which recommended that 'The Government, working with the universities and industry, should create a National Centre to promote advanced research and translational work in algorithms and the application of data science'. The AI for Science and Government programme proposes an integrated programme in AI and data science to transform engineering through digital twins; data-driven health through delivering personalised medicine with early disease detection and machine-learning based diagnosis; physical and life sciences, by applying AI to experiment outputs; and government through data-linking for policy development, particularly through the criminal justice system. The research programme is aligned with the Industrial Strategy, in particular the AI Grand Challenge to put the UK at the forefront of the AI and data revolution, the first mission of which is to make use of data, AI and innovation to transform the prevention, early diagnosis and treatment of diseases by 2030.

Cells are a fundamental unit of biology and cellular heterogeneity is a hallmark of multicellular life. Within an organism - plant, animal or human - cells can display enormous functional diversity, fulfilling distinct roles that underpin the overall function of the organism. The diversity of cell function reflects distinct "molecular cell identities", emerging from variation in the genomes of each cell, how they are regulated, and the genes they express. Incorporating a cellular perspective into functional genomics experiments is therefore key to unravel how variation in sequence and regulation of the genome within an organism can contribute to its overall function and phenotype. The Cellular Genomics programme (CELLGEN) builds on EI expertise in data science, bioinformatics and single-cell analysis to investigate the impact of genomic and transcriptomic heterogeneity in healthy plants and animals. Recent developments in molecular and computational science have ushered in an era of single-cell genomics, in which the analysis of the genomes, epigenomes and transcriptomes of individual cells can readily be analysed. In several model systems, these approaches have highlighted the extent to which genome function - but also genome sequence - can vary between cells of the same organism during the healthy lifespan. Here, we seek to leverage and develop these approaches to explore the origins and consequences of genomic and transcriptomic heterogeneity within model and non-model organisms. These analyses generate high-value, highly-dense datasets, and our programme takes a data-science led approach. Our programme of work leads with WP1 "Data Science for Cellular Genomics" which develops approaches to curate and harmonise datasets, to enable reproducibility, reusability, comparison, and most importantly integration across studies. WP2 "Consequences of somatic mutations on traits" investigates the diversity and consequences of intraorganismal genomic variation. We will develop approaches to measure emerging genomic heterogeneity (tandem repeats and polyploidisation) and their immediate and long term effects on gene and isoform expression in cells - using model cell lines or primary polyploid cells from plants (trichomes) and mouse (megakaryocytes). We will test novel hypotheses about the relationship between cellular genomic variation and the wider phenotype of the organism, for example measuring the impact of genetic variation between individual cells on gene regulation. This can then be directly related to how genes are expressed at the tissue and whole organism levels. WP3 "Cellular heterogeneity and expression regulation" characterises cellular transcriptomic heterogeneity and its impact on cell and organism responses to the environment. Here we will explore the regulation of gene expression in plant and animal models of cell lineage commitment (haematopoiesis) and responses to environmental challenges (fish, plant) and dietary intervention. This programme of work will enable researchers working on diverse biological systems to work synergistically to explore common themes across the tree of life. It will position EI as a world leader in single-cell developments for non-model organisms, plants and animals, going beyond cell-type classification and delivering novel approaches for scalable cellular functional genomics underpinned by advanced data science approaches. By exploring consequences of genomic and transcriptomic variation during a healthy lifespan and into how cellular diversity underpins organismal adaptation to environment, CELLGEN will generate new insights into the rules of life.

Since the start of the industrial revolution the CO2 concentration in the atmosphere has steadily risen. Scientists have confirmed that the recent loss of Arctic sea ice in summer directly follows this rise in human-induced CO2 emissions, reducing from about 7 million km2 of Arctic sea ice in the late 1970s to around 3.5 million km2 in the 2010s. While climate models suggest Antarctic sea ice extent should also reduce in response to rising CO2, satellite observations reveal that during 1979-2015 the opposite was in fact true. The trend in Antarctic sea ice extent has been a small increase of approximately 1.5% per decade. In 2016, however, this increase was abruptly interrupted by a dramatic reduction in sea ice extent that was far outside the previously observed range. Since the extreme event in 2016, Antarctic sea ice extent has almost returned to its pre-2016 values, highlighting the significant variability in Antarctic sea ice conditions that can occur from one year to the next. These variations in sea ice are important to the whole Earth's climate, because they affect the melting of the glacial Antarctic Ice Sheet, and the capture of atmospheric heat and CO2 by the Southern Ocean. The recent extreme swings in Antarctic sea ice extent, and the challenge of accurately predicting, understanding and modelling them, emphasise the need to: (i) increase our knowledge of the processes that drive Antarctic sea ice variations, including extreme events, and (ii) understand the drivers and climate implications of Antarctic sea ice loss over different time-scales, from weeks to decades. To address this knowledge gap requires a significant research programme, one that takes year-round observations, including throughout the harsh Antarctic winter, and is effective in improving the underlying processes in the latest computer climate models. Our project, known as DEFIANT (Drivers and Effects of Fluctuations in sea Ice in the ANTarctic), will embark on one of the most ambitious observational campaigns aimed at understanding Antarctic sea ice variability. Scientific measurements from the German research ship Polarstern, the UK's new polar research ship Sir David Attenborough, the British Antarctic Survey's Rothera research station, aircraft overflights and satellites will work seamlessly together with cutting-edge robotic technologies (including the underwater vehicle Boaty McBoatface and a suite of on-ice buoys) to provide us with comprehensive, year-round measurements of atmosphere, sea ice and ocean. The knowledge gained from these observations will enable our team to develop new ocean and climate models in order to more accurately represent Antarctic sea ice processes. The analysis of these improved models will allow us to better understand the underlying drivers of the sudden decrease in Antarctic sea ice, determine the impact of these extreme events on the global ocean circulation, and forecast the implications for the movements of heat and CO2 through the climate system. By developing new observations, new satellite records, and new models, DEFIANT will deliver a major advance in our understanding of the Antarctic sea ice system and its wider impacts on global climate.

Abstracts are not currently available in GtR for all funded research. This is normally because the abstract was not required at the time of proposal submission, but may be because it included sensitive information such as personal details.