This is a proposal for a fellowship award to Principal Investigator (PI), Prof Subramanian Ramamoorthy, for the establishment of a new research centre focussed on the development of a radical new concept for human-centred Artificial Intelligence. Driven by the mission of creating assistive autonomous systems that are person-centred and teachable at the individual user level, we will bring together a range of innovations including large scale interactive data gathering, executable cognitive modelling of human behaviour, introspective and compositional frameworks for model building, and computationally rational decision making frameworks.
In addition to our core scientific goals, we aim to be innovative in how we organise the centre. We draw on the concept of a focussed research organisation, structuring the centre as a foundations team working with three domain specific laboratories - addressing assisted living, surgical assistance and autonomous vehicles respectively. This organisational structure will enable closely collaborative work with a range of external partners, with a view to field deployments and spinouts in each of these application areas.
Alongside the UKRI funding requested, this proposal benefits from £1 million in direct university contribution in the form of funding for additional personnel, as well as a £100K equipment donation and in-kind contributions from our external partners valued at £1.21 million.
The key objectives of the centre are to:
O1: Devise a modelling framework for human behaviour that yields 'pre-trained' executable models capable of achieving broad domain performance
O2: Devise novel tools and technologies for iterative and introspective model learning, as well as interactive decision making for assistive autonomy
O3: Demonstrate transformative benefits in terms of user acceptability and quality of assistance in the selected application domains (assisted living, surgical assistance, autonomous vehicles)
O4: Develop a growing international community and evidence base around deployed systems based on the proposed methodology
Through the centre's activities, we will seek to create benefits for a wide range of stakeholders, including:
AI community, by addressing fundamental gaps associated with broad domain human behaviour modelling and their use in computationally efficient interactive decision making frameworks,
Industry in the relevant sectors mentioned above, including our external partners already listed in this proposal, and a much more widely extended network of stakeholders to be engaged during the course of the fellowship,
Early Career Researchers, who will be trained within the centre, and those who will participate in our various collaborative projects and events,
Policy makers and various publics, who will benefit from our wide-ranging engagement activities, including those involved in exercises such as patient and public involvement consultations,
The broader public, through the potential economic growth via products and spinouts emerging from the proposed work.
The proposed work builds on the PI's strong track record of robot learning in human-centred environments, including sustained contributions to topics such as integrated prediction and planning in interactive environments, interactive task learning through embodied conversation, and methods for safety evaluation and risk-sensitive design in open environments. These ideas have been developed through major projects including, e.g., ISCF funded ORCA HUB, DARPA XAI programme and UKRI TAS, in which the PI had leadership roles. The PI has demonstrated experience of translating this world-leading science to industrial practice, having played a key role in the establishment and growth of Five, a UK-based technology company in the autonomous vehicles domain which was acquired by Bosch in 2022.
|