Five internationally-recognised researchers have been appointed as the first Turing Artificial Intelligence (AI) World-Leading Researcher Fellows.
The new fellows, who will conduct ground-breaking work on AI’s biggest challenges, are:
- Professor Zoubin Ghahramani, University of Cambridge
- Professor Samuel Kaski, The University of Manchester
- Professor Mirella Lapata, The University of Edinburgh
- Professor Philip Torr, University of Oxford
- Professor Michael Wooldridge, University of Oxford.
The fellowships, named after AI pioneer Alan Turing, are part of the UK’s commitment to further strengthen its position as a global leader in the field.
Retaining and attracting some of the best international research talent in a highly competitive international environment will increase the UK’s competitive advantage and capability in AI.
Transformative impact
The fellows’ research will have a transformative effect on the international AI research and innovation landscape by tackling some of the fundamental challenges in the field.
It could also deliver major societal impact in areas including:
- decision-making in personalised medicine
- synthetic biology and drug design
- financial modelling
- autonomous vehicles.
About the fellows
Professor Zoubin Ghahramani, University of Cambridge
Professor Ghahramani is:
- Senior Director and distinguished researcher at Google
- Former Chief Scientist at Uber
- a Fellow of the Royal Society.
He will hold his fellowship jointly while continuing to work at Google.
He aims to develop the new algorithms and applications needed to address limitations faced by the AI systems that underpin technologies such as speech recognition and autonomous vehicles.
This includes ensuring they can better adapt to new data, and apply data-driven machine learning approaches to simulators to understand complex systems.
Professor Samuel Kaski, The University of Manchester
Professor Kaski holds a joint position as:
- Research Director of the Christabel Pankhurst Institute for Health Technology Research
- Director of the Finnish Centre for Artificial Intelligence at Aalto University.
Through his fellowship, Professor Kaski aims to overcome a fundamental limitation of current AI systems, that they require a detailed specification of the goal before they can help.
In difficult design and decision tasks such as drug design we often cannot give that, because the desired outcomes uncertain and evolving.
The new tools will be applied to help in drug design and to improve diagnoses and treatment decision making in personalized medicine.
Professor Mirella Lapata, The University of Edinburgh
Professor Lapata:
- is a Fellow of the Royal Society of Edinburgh
- is Director of the UK Research and Innovation (UKRI) Centre for Doctoral Training in Natural Language Processing
- holds the Royal Society Wolfson Merit Award
- was the first recipient of the Karen Spärck Jones Award.
Professor Lapata aims to develop an AI system, inspired by the human brain, that is capable of advanced reasoning and able to draw conclusions from large and varied sets of data.
This would address the limitations of current AI systems which cannot match the sophistication of the human brain in integrating large amounts of information from different sources.
Professor Philip Torr, University of Oxford
Professor Torr is:
- a Fellow of the Royal Society and of the Royal Academy of Engineering
- Founder of spin-out Oxsight and AIstetic.
He aims to make deep neural networks more robust.
Deep networks are widely used in applications from fraud-detection to self-driving cars but are have surprising vulnerabilities.
Through his fellowship Professor Torr will create a new centre of excellence. The centre will make deep learning reliable, robust and deployable as well as capable of efficiently handling the enormous quantities of data they will be fed with.
Professor Michael Wooldridge, University of Oxford
Professor Wooldridge is:
- Head of Computer Science at Oxford
- Co-programme Director for AI at The Alan Turing Institute
- a recipient of the British Computer Society’s Lovelace Medal
- a Fellow of four AI and computing societies and associations.
Professor Wooldridge will be working with industrial partners:
- Accenture Global Solutions
- JP Morgan
- Oxford Asset Management
- Schlumberger
- Vodafone.
He aims to improve the agent-based AI models that are increasingly used in sectors such as financial modelling and logistics.
Professor Wooldridge’s team have previously used agent-based models to understand the causes of catastrophic collapses in global markets, so called “flash crashes”, and they will continue this work in the project.
Globally competitive advantage
Business Secretary Kwasi Kwarteng said:
As home to Alan Turing, the father of artificial intelligence, the UK boasts a globally competitive advantage in AI, and the Turing AI World-Leading Researcher Fellowships will ensure that we continue to attract and retain the world’s most talented AI innovators.
Backed to the tune of £18 million, these five outstanding researchers will use AI to tackle the great societal challenges of our time that could improve how we live and work, from personalised medicine to autonomous vehicles, all while cementing the UK’s status as a global science superpower.
Government Chief Scientific Adviser, Sir Patrick Vallance, said:
These five internationally-recognised researchers appointed as the first Turing AI World-Leading Researcher Fellows will help enable us to attract top talent from across the globe and ensure that the UK stays at the forefront of AI research and innovation.
This expertise will increase the UK’s capabilities in AI and equip us to face greater and more complex challenges.
The fellows are supported with an £18 million investment by UKRI.
Industry collaboration
In addition to this, 39 different collaborators including IBM, AstraZeneca and Facebook are making contributions worth £15.7 million to the fellows’ research programmes.
The fellowships are being delivered by UKRI’s Engineering and Physical Sciences Research Council (EPSRC).
EPSRC Executive Chair, Professor Dame Lynn Gladden, said:
The Turing AI World-Leading Researcher Fellowships recognise internationally-leading researchers in AI, and provide the support needed to tackle some of the biggest challenges and opportunities in AI research.
These fellowships enable the UK to attract top international talent to the UK as well as retaining our own world-leaders. Attracting and retaining top talent is essential to keep the UK at the leading edge of AI research and innovation.
Innovation strategy
It follows the publication of the government’s innovation strategy last week.
The strategy identified seven strategic technologies for the UK to prioritise and build on existing R&D strengths, including AI, where the UK has globally competitive advantage.
AI is a significant global opportunity to increase economic wealth and transform society, with many other countries investing heavily in research.
Global opportunity
Today’s announcement is part of a £46 million investment in AI research leaders through Turing AI fellowships.
The fellowships, alongside wider investment, are designed to build on the UK’s leading role in AI and boost its reputation as a great place to study, invest or work in AI.
The investment includes:
- five fellows awarded by The Alan Turing Institute in 2019
- Turing AI acceleration fellowships awarded by UKRI, announced last year with £20 million of funding.
Fellowship investment
The Turing AI fellowships investment is be delivered in partnership by:
- UKRI
- Office for AI
- The Alan Turing Institute, the national institute for data science and AI.
Professor Sir Adrian Smith, Director of The Alan Turing Institute, said:
Inspired by the legacy of Alan Turing, the UK already attracts many of the most internationally-renowned talents in AI, and these fellowships will further enable some of our brightest minds, across a range of disciplines to ensure AI research continues to have a positive and transformative effect.
Further information
The Turing AI World-Leading Research Fellows
Project Zoubin Ghahramani, University of Cambridge: Advancing modern data-driven robust AI
UKRI support: £2.6 million
Modern AI is dominated by methods that learn from large amounts of data and underpin technologies such as:
- speech recognition
- production recommendation
- autonomous vehicles.
They are also the basis of recent breakthroughs in AI like the game-playing systems that can beat humans at chess, Go and poker.
They underlie practical advances in science, engineering and medicine such as automated tools for analysing genomic data and medical images.
Number of limitations
Despite significant advances, these systems face a number of limitations.
These include:
- poor handling of large amounts of additional, useless information
- uncertainty and changing circumstances
- gaps in their ability to combine symbolic and statistical reasoning
- the lack of automation of many of the stages of learning.
Professor Ghahramani aims to develop new algorithms and applications to address these limitations.
Adapting and reporting
For example, by ensuring that AI systems can better adapt to new data and report when they are unable to find a correct conclusion instead of continuing to an incorrect one.
He aims to develop better tools to automate the process of building and maintaining machine learning systems and apply approaches from data-driven machine learning to simulators. Simulators are widely used to model and understand complex systems in science and engineering.
He aims to use the tools developed through his fellowship to address problems in modelling and optimising complex systems with many interdependent components, such as electrical grid efficiency and transportation systems.
Project partners:
- Graphcore
- Invenia Labs
- Tractable Ltd
- Wayve Technologies Ltd.
Total project partner contribution: £1.45 million
Professor Samuel Kaski, The University of Manchester: Human-AI research teams, steering AI in experimental design and decision-making
UKRI support: £4.4 million
Machine learning, where solutions to problems are automatically learnt from data, is a form of AI with great promise for addressing a number of challenges.
This includes healthcare, where AI can detect patterns associated with diseases and health conditions by studying healthcare records and other data.
However, machine learning is still limited by the fact that we need to set appropriate objectives and rewards to tell AI systems which outcomes are desired.
Desired outcomes
This is difficult when we only partially know the goal, as is the case at the beginning of scientific research.
Professor Kaski aims to develop new ways for machine learning systems to help humans in the problem-solving process of:
- designing experiments
- interpreting what results mean
- deciding what to measure next
- to finally reach trustworthy solutions to problems.
This new approach will be applied to three challenges:
- diagnosis and treatment decision-making in personalised medicine
- the guidance of scientific experiments in synthetic biology and drug design
- the design and use of digital twins to design physical systems and processes.
New approaches and tools
In drug design, for instance, the most advanced current tools are able to generate candidate molecules if we can specify a precise objective function for them.
However, that is difficult for humans to do, and if our specifications are not perfect the intelligent system will very cleverly produce results we do not want.
This is where the proposed new approaches and tools will help.
An AI centre of excellence will be established at The University of Manchester, in collaboration with The Alan Turing Institute and a number of partners from the industry and healthcare sector. The centre will have strong connections to the networks of best national and international AI researchers.
Project partners:
- Aalto University
- Apis Assay Technologies Ltd
- AstraZeneca
- Delft University of Technology
- Etsimo Healthcare Oy
- Gendius Limited
- Greater London Authority
- Health Innovation Manchester
- IBM
- Kyoto University
- Spectra Analytics
- University of Birmingham
- University of Cambridge
- University of Toronto
- Zero Carbon Farms Ltd
- Greater Manchester Combined Authority
- NHS through the Pankhurst Institute.
Total project partner contribution: £6.4 million
Professor Mirella Lapata, The University of Edinburgh: TEAMER, Teaching machines to reason like humans
UKRI support: £3.9 million
Progress in deep learning, which aims to mimic the human brain to process information and make decisions, has led to advances in:
- speech recognition
- natural language processing
- robotics.
However AI systems utilising deep learning still suffer from drawbacks in reasoning, that is taking pieces of information, combining them and using them to draw logical conclusions or devise new information.
Visual and textual information
Computer systems can draw conclusions from visual and textual information.
However, the human brain is far more sophisticated at correlating and integrating information from different sources and re-using previous experience to transfer it to radically different challenges.
Current AI systems fail when exposed to data outside the information they were trained on.
Misleading associations
They adhere only to superficial and potentially misleading associations instead of learning true causal relationships.
They are also unable to reason on an abstract level and provide us with full understanding of how they came to a specific conclusion.
Advanced reasoning
Professor Lapata aims to develop neural networks, inspired by the human brain, that are capable of advanced reasoning.
The specialised components forming these networks would have differing strengths which would ensure they have a more well-rounded ability to reason, being able to:
- draw conclusions from large and varied sets of data
- deal with change
- be creative
- explain their predictions and decisions.
Project partners:
- ARM Ltd
- BBC
- British Library
- Google DeepMind UK
- Dyson Limited
- Huawei
- IBM
- Naver Labs Europe
- RAS Technologies GMBH
- Scottish and Southern Energy SSE plc
- Brainnwave Ltd
- Wallscope
- Amazon Research Cambridge.
Total project partner contribution: £4.2 million
Professor Philip Torr, University of Oxford: Robust, efficient and trustworthy learning
UKRI support: £3 million
Deep neural networks imitate human intelligence and are capable of learning from huge amounts of data.
They have a wide range of applications, from fraud detection and visual recognition to self-driving cars, but their limitations are becoming increasingly evident.
These include a vulnerability to adversarial examples, where data is presented with the intent of causing AI models to make mistakes.
Classifying information
In safety-critical applications such as autonomous vehicles or medical diagnosis this presents issues around AI models incorrectly classifying information.
For example, it has recently been demonstrated that the neural networks underlying autonomous vehicle autopilots can be fooled by markers on the ground into swerving into the opposite lane.
Through his fellowship Professor Torr will create a new centre of excellence. The centre will make deep learning reliable, robust and deployable as well as capable of efficiently handling the enormous quantities of data they will be fed with.
Project partners:
- Five AI
- Facebook (International)
- Remark Holdings
- Horizon Robotics.
Total partner project contribution: £1.9 million
Professor Michael Wooldridge, University of Oxford: The large agent collider, robust agent-based modelling at scale
UKRI support: £3.5 million
Agent-based models are increasingly used in areas such as financial modelling, logistics, and supply chain management.
Built from autonomous decision-making AI ‘agents’, they provide a novel way of modelling many systems.
However current approaches to agent-based modelling have many limitations, including a lack of suitable modelling languages, and an inability to verify their predictions.
Overcoming challenges
Professor Wooldridge aims to carry out the fundamental science needed to overcome these challenges.
He will use state-of-the-art AI and machine learning techniques to develop, populate, calibrate, and validate agent-based models at scale.
Working with major industrial partners, he will test and refine techniques on a range of real-world case studies. He will transform agent-based modelling from an ad hoc, trial-and-error process into a robust and trusted discipline.
Project partners:
- JP Morgan Chase
- Accenture Global Solutions Limited.
Total project partner contribution: £60,000
Top image: Credit: MF3d/GettyImages