Leading UK researchers will develop cutting-edge artificial intelligence (AI) technologies through prestigious fellowships announced today, 27 November.
The novel AI techniques they will develop could have wide-ranging impact, for example:
- combatting cancer
- developing digital twins that can aid us in modelling and understanding air pollution
- improving cybersecurity through developing more robust and transparent AI algorithms.
Fifteen researchers have been awarded Turing AI Acceleration fellowships, named after AI pioneer Alan Turing.
The fellowships are supported by a £20 million government investment, being delivered by UK Research and Innovation (UKRI), to lead innovative and creative AI research with transformative impact.
Cutting-edge AI
For example, Professor Aldo Faisal at Imperial College London will develop new AI to help clinicians to choose the best treatment for hospital patients post-diagnosis, through introducing or changing drug treatments.
Professor Yulan He, at the University of Warwick, will explore how we can develop AI-driven systems for use in chatbots, for example.
The systems are able to understand and respond correctly to everyday human language usage, bringing us closer to human-like AI.
And Dr Sebastian Stein, at the University of Southampton, aims to develop trusted AI systems which put citizens at their heart and involve them in decision-making, rather than viewing them as passive providers of data.
These could be used in a wide range of applications, from using crowdsourced information to track the spread of infectious diseases and issue guidance to helping people to manage their energy needs.
The birthplace of AI
Science Minister, Amanda Solloway said:
The UK is the birthplace of artificial intelligence and we have a duty to equip the next generation of Alan Turings with the tools that will keep the UK at the forefront of this remarkable technological innovation.
The inspirational fellows we are backing today will use AI to tackle some of our greatest challenges head on, transforming how people live, work and communicate, cementing the UK’s status as a world leader in AI and data.
Digital Minister, Caroline Dinenage said:
The UK is a nation of innovators and this government investment will help our talented academics use cutting-edge technology to improve people’s daily lives, from delivering better disease diagnosis to managing our energy needs.
The Turing AI Acceleration fellowships will accelerate and support the careers of a diverse cadre of the best and brightest AI researchers enabling them to become world-leading researchers in the five years of the award.
This will sustain and strengthen the UK’s leading international position in AI.
Academic and industry collaboration
These fellowships will increase collaboration between academia and industry, with each fellow bringing together a wide range of partners on their projects to accelerate the impact of their transformative AI technologies.
Partners have already committed to cash and in-kind contributions in excess of £10 million.
The Turing AI Acceleration fellowships are part of the £46 million investment that was announced in the 2018 budget following the government’s review of the UK AI industry.
A major government investment
Five fellowships have previously been awarded (The Alan Turing Institute) and the Turing AI World-Leading Researcher fellowships call is in progress.
These fellowships are part of a major government investment in AI skills and research that also includes 16 UKRI centres for doctoral training in AI announced by Prime Minister Boris Johnson (The National Archives).
The Turing AI fellowships are delivered through UKRI’s Engineering and Physical Sciences Research Council (EPSRC), in partnership with:
- Department for Business Energy and Industrial Strategy
- Office for AI
- Alan Turing Institute.
Supporting leading researchers
EPSRC Executive Chair Professor Dame Lynn Gladden said:
The Turing AI Acceleration fellowships will support some of our leading researchers to progress their careers and develop ground-breaking AI technologies with societal impact.
By enhancing collaboration between academia and industry and accelerating these transformative technologies these fellowships will help to maintain and build on the UK’s position as a world leader in AI.
Further information
For media enquiries contact James Giles-Franklin, UKRI External Communications, at james.giles-franklin@ukri.org.
The quality of the applications received was high.
Due to clear focus in some applications on fundamental mathematical sciences, an additional £2.3 million has been contributed from the Additional Funding Programme for Mathematical Sciences.
A total of 15 fellows will be supported.
The Turing AI Acceleration fellows
Professor Damien Coyle, University of Ulster: AI for Intelligent Neurotechnology and Human-Machine Symbiosis
Professor Coyle aims to develop AI technology that will play a crucial role in new forms of wearable neurotechnologies, devices which measure signals from the brain and enable their wearer to interact with technology without movement.
Enabling movement-independent communication through this brain-computer interface could help those who are unable to communicate following a serious injury or illness.
Professor Coyle is leading a national trial in partnership with 17 hospitals to evaluate AI-enabled neurotechnology for consciousness assessment in prolonged disorders of consciousness following severe brain injury.
Wearable neurotechnologies and brain-computer interfaces have other applications in ambulatory brain monitoring, medical and neuroscience research, recreation, gaming (neurogaming) and sport, among others.
Dr Jeff Dalton, University of Glasgow: Neural Conversational Information Seeking Assistant
Dr Dalton aims to improve the capability and performance of voice-based personal assistants, similar to Alexa and Siri. Current systems are only capable of simple tasks and limited conversations.
Dr Dalton aims to research information agents that can collaboratively interact with users to accomplish long and complex information tasks, for example researching the causes of climate change.
The research will use deep-learning methods for machine reading of text and learning from user interaction to enable agents that are developed more quickly and easily without specialised experts.
Dr Theo Damoulas, University of Warwick: Machine Learning Foundations of Digital Twins
Dr Damoulas aims to establish the machine learning foundations for AI-enabled digital twins: digital representations of assets and processes that are tied to their physical ‘twins’ through streaming data, information flows and interventions.
Such distributed, constantly-improving, digital twin systems will enable robust simulation and evaluation of ‘what if’ scenarios and pave the way for continuously learning and disentangling complex intertwined phenomena in the real world.
The resulting principles and advances will be demonstrated in environmental and urban digital twins that will allow us to better understand and predict air pollution over cities while optimising our policies and mitigation strategies, for example.
Professor Aldo Faisal, Imperial College: Reinforcement Learning for Healthcare
Professor Faisal aims to develop a fundamentally different approach to decision-support systems than conventional AI methods previously used, though the use of reinforcement learning.
This can learn and distil the best plan of action to treat a patient, by harnessing existing hospital data and the expert knowledge of clinicians.
The AI system will not only learn to recommend optimised medical interventions such as prescribing drugs and changing dosages as needed by a patient, but also learn to make recommendations and their causes in a manner that is meaningful to decision-makers.
This helps them make the best final decision on a course of action, like an AI clinical colleague.
The machine learning approaches developed by Professor Faisal could also be used in other regulated sectors such as aerospace or energy, where there is a need for decision-making support but a scarcity of highly skilled human experts.
Professor Yulan He, University of Warwick: Event-Centric Framework for Natural Language Understanding
Natural Language Understanding (NLU) is a branch of AI that aims to allow computers to understand text automatically.
NLU may seem easy to humans, but it is extremely difficult for computers because of the variety, ambiguity, subtlety, and expressiveness of human languages.
Professor He will explore how we can develop AI-driven systems with reasoning capabilities which are able to read and comprehend text and formulate an answer automatically when presented with a query, bringing us closer to human-like AI.
Such systems will find a wide range of applications including intelligent virtual assistants, automated customer services, smart home, and question-answering in the finance and legal domains.
Dr José Miguel Hernández Lobato, University of Cambridge: Machine Learning for Molecular Design
Many existing challenges, from personalised health care to energy production and storage, require the design and manufacture of new molecules.
However, identifying new molecules with desired properties is difficult and time-consuming.
Dr Hernández Lobato aims to accelerate this process by creating new deep generative models of molecules that operate by placing atoms in 3D space, generate synthesizable molecules and are robust and data efficient.
These models can then be used to design more effective flow batteries and solar cell components, and accelerate the drug discovery process to create economic and effective drugs that can significantly improve the health and lifestyle of millions.
Dr Antonio Hurtado, University of Strathclyde: PHOTONics for ultrafast Artificial Intelligence (PHOTON-AI)
The popularity of smart devices and online services, and increasing demand for AI within sectors such as energy, healthcare and finance means the ability to process fast and efficiently large volumes of data is becoming ever more important.
Dr Hurtado aims to develop AI systems inspired by the powerful capabilities of networks of neurons in the brain and utilising photonic devices that create, manipulate or detect light.
These light-enabled AI systems will be able to operate at very high speeds while retaining low energy consumption. Their potential capability to perform complex computational tasks at ultrafast speed could see them used to:
- improve meteorology forecasting
- process images at very fast rates for medical diagnostics and autonomous vehicles technologies
- analyse wind patterns in offshore wind energy farms to optimise their performance.
Dr Per Kristian Lehre, University of Birmingham: Rigorous Time-Complexity Analysis of Co-evolutionary Algorithms
Optimisation, the problem of identifying the best solution among a vast set of candidates, is fundamental in AI and essential to UK competitiveness.
New optimisation algorithms are needed that deliver high-quality solutions, even in unforeseen scenarios.
For example, next-generation controllers for charging fleets of electrical vehicles must cope with fluctuations in renewable energy sources and in charging demand.
The fellowship gives Dr Lehre resources to develop the theory required to reliably exploit co-evolutionary algorithms for optimisation. Such algorithms simulate competitions between predators and prey.
The work will facilitate the adoption of novel and competitive AI techniques to inform decision making across UK government and businesses.
Professor Giovanni Montana, University of Warwick: Advancing Multi-Agent Deep Reinforcement Learning for Sequential Decision Making in Real-World Applications
Professor Montana aims to make advances in Deep Reinforcement Learning (DRL), an area of machine learning which teaches artificial decision makers such as robots and software agents how to interact with the world in order to achieve a desired goal.
By allowing autonomous systems to learn a wide range of skills without human intervention, DRL will allow them to be effective in various applications, such as:
- industrial assembly lines and warehouse management systems
- driverless cars
- decision making for the most appropriate form of medical treatment for patients.
Dr Christopher Nemeth, Lancaster University: Probabilistic Algorithms for Scalable and Computable Approaches to Learning (PASCAL)
A key challenge in AI research is to extract meaningful value from unstructured disparate data sources to make decisions that can be trusted and understood to improve society.
Dr Nemeth aims to develop an end-to-end framework, from data to decisions, that naturally accounts for data, modelling and decision uncertainties to provide transparent and interpretable decision-making tools.
The algorithms developed throughout this research project are based on Bayesian machine learning and will be applicable to a wide range of domains, with specific applications in the areas of autonomous vehicles and cyber security.
Dr Raul Santos-Rodriguez, University of Bristol: Interactive Annotations in AI
Dr Santos-Rodriguez’s fellowship will focus on human-centric machine learning.
He will explore new extended forms of supervision and interaction between AI systems and humans to build trust and accountability while making the learning process more efficient.
In particular, focusing on developing methods for humans to provide informative and actionable feedback in order to shape the behaviour of AI systems, allowing humans in return to fully understand and measure the effect of their contribution.
The fellowship will cater for data and AI practitioners, domain experts and end-users across multiple fields, including health, IT, engineering and social media.
For example, novel tools for adapting decision support systems through rich annotations that accurately reflect their needs and constraints will be co-created with clinicians in intensive care units across the UK, changing the mindset from the current practice of clinicians being the ones that adjust to the way that AI systems work.
Dr Sebastian Stein, University of Southampton: Citizen-Centric AI Systems
Dr Sebastian Stein aims to develop trusted AI systems which put citizens at their heart and involve them in decision-making, rather than viewing them as passive providers of data.
These citizen-centric AI systems could be used in a wide range of applications, from using crowdsourced information to track the spread of infectious diseases and issue personalised guidance, to helping people manage their energy and transportation needs in a more sustainable manner.
Professor Ivan Tyukin, University of Leicester: Adaptive, Robust and Resilient AI Systems for the FuturE
To ensure that the transformational impact of AI is fully realised, new data-driven AI systems will need to be trusted and understood.
These systems are based on empirical data and are impacted by a complex range of factors.
These factors include uncertainties which are inherent within any empirical data and uncertified AI design practices leading to biases and a lack of transparency.
Altogether, they create major societal risks such as wrong medical diagnosis, incorrect financial advice, and crashes of autonomous cars at scale.
This project aims to go beyond current AI theories to create new-generation of data-driven AI that are adaptive, resilient, robust, explainable, trustworthy, and certifiable.
The theory will also allow creation of technologies to maintain existing AI assets and enable a formal understanding of the fundamental limits of data-driven AI, independent of application and learning algorithm.
This will enable AI practitioners, through understanding AI limitations, to influence policy and prevent incidents before they occur.
Dr Adrian Weller, University of Cambridge – Trustworthy Machine Learning
Machine Learning presents tremendous opportunities for society but also introduces risks, such as embedding unfair biases or creating new vulnerabilities.
It is therefore crucial that we can understand and deliver what is needed for such systems to be trustworthy.
Dr Weller will build solid technical underpinnings via new theory and practical algorithms for trustworthy deployment focusing on three key measures:
- fairness
- interpretability
- robustness.
To make progress and enhance impact, these measures will be grounded in specific settings, while engaging with real world practitioners and stakeholders in order to face the inevitable interactions and trade offs that occur in context.
The work will examine applications in criminal justice and healthcare: two domains with high stakes decisions and clear outcome goals.
Machine learning presents great hope to improve the consistency and efficiency of sound judicial decisions and beneficial patient outcomes, but also raises important ethical concerns which this work will address.
Professor Christopher Yau, The University of Manchester: clinAIcan – Developing Clinical Applications of Artificial Intelligence for Cancer
Professor Yau aims to develop novel AI-driven predictive models that will allow us to describe how cancers evolve at the molecular level.
He aims to exploit the fact that cancers, whilst never exactly identical, often share similar development trajectories which we can learn about by collating information from across deep high-resolution molecular profiles of many cancers.
By embedding our extensive biological knowledge of cancer within AI models, he will develop intelligent systems that will produce predictions that are more realistic, interpretable and better explain the progression of cancers.
This will help to improve the efficiency of drug development, decisions on treatment and to provide patients with more information about their illness.
Top image: Artificial intelligence (credit: ipopba/GettyImages)