While AI has been making impressive advances, the systems we can build today are still far from being able to perform complex reasoning tasks, act autonomously in real-world environments, and interact with humans naturally. Moreover, many of the most successful AI systems require excessive amounts of data and energy to be trained, struggle to adapt to new tasks, and involve the use of opaque models to deliver the required performance.

This creates safety, sustainability, and ethical risks that have sparked much criticism and controversy – but it also raises important scientific questions regarding what methods will look like that can help overcome these limitations in the future.

Aims and Objectives

In this workshop, we want to engage with the breadth of the UK AI community to develop a vision for the future of AI and identify new research directions around which we can bring people together to build new collaborations across different sub-communities.

Taking “integrated AI and incremental autonomy” as a starting point as one possible direction this vision could take, we will use the afternoon hear from distinguished experts in different fields of AI - Leslie Pack Kaelbling (MIT) and Mirella Lapata (The University of Edinburgh) - and engage in small-group and plenary discussions to develop ideas for grand challenges and approaches to addressing them. We plan to capture and summarise insights from this and future workshops in what will become a roadmap for future research initiatives.

We invite everybody in the AI UK community, in particular those from underrepresented fields and disciplines, geographies, and societal groups to join us for this exciting online event.

Agenda (all times UTC)

13:00-13:30 Welcome and vision - Michael Rovatsos

13:35-14:15 Mirella Lapata, Movie Summarization as a Testbed for Machine Reasoning

14:15-14:45 Breakout groups - “How do we make a step change in AI?”

14:45-15:15 Break

15:15-15:55 Leslie Pack Kaelbling, Doing for Our Robots what Nature Did for Us

16:00-16:40 Breakout groups - “Mapping the landscape”

16:40-17:00 Summary and planning next steps

Who we are

The workshop is organised by the Enabling Advanced Autonomy through Human-AI Collaboration project, a collaboration between the University of Edinburgh, the University of the West of England, and the University of Oxford as part of the Towards Turing 2.0 initiative funded by EPSRC, The Alan Turing Institute, and The University of Edinburgh. It is jointly run by the University of Edinburgh’s Bayes Centre and School of Informatics in collaboration with the Alan Turing Institute.

If you are interested in our future activities and receive updates, we would like to hear from you! Please consider completing our survey to help us understand what our community is working on and interested in.

Please be aware that the event will be recorded and may be shared online in whole or in parts, but you will be able to participate in discussions via the chat if you would not like recordings of your video and audio feed to be shared. We will ask for your explicit consent to share your details with other attendees when you register.

Registrations are closed

Towards a Vision for Next-Generation AI