UKRI Centre for Doctoral Training in Foundational Artificial Intelligence

About the course

Our Foundational AI CDT addresses the national need for AI workers by training researchers capable of advancing core AI algorithms. These graduates will help shape the social, scientific and economic landscape through scientific breakthroughs and the creation of companies on the basis of novel AI technology.

Current AI machines are largely “dumb” - they don’t understand their physical environment, nor have enough understanding of human culture to communicate in natural ways. Our vision is that AI is in its infancy and that AI breakthroughs are key to controlling and shaping the future technological landscape. However, creating effective AI is challenging given our limited understanding of how intelligence works.

In forging AI creators, we will therefore encourage students to look beyond Computer Science and be open to interaction with other scientists researching intelligence.

The most successful existing AI technology (Deep Learning) is based on the way brains process information. Future developments may be inspired by neuroscience and we need to be alert to the insights offered by this and other fields.

A key research objective of CDT students is to make new algorithms for next-generation AI technologies by incorporating more knowledge about the real-world and human culture into the AI agents themselves. Such machines will be able to offer unparalleled insights into our data-rich world and provide us with transparent and interpretable explanations.

A unique aspect of the CDT is to give students the deep technical skills they require to be leading researchers in AI and also the skills to be a deep tech entrepreneur.

Funding and Recruitment Update:
The CDT in Foundational Artificial Intelligence was originally established with funding from the Engineering and Physical Sciences Research Council (EPSRC). As this initial funding period comes to an end, the CDT will continue its mission and activities with the support of our industry partners. This new phase of funding will ensure the long‑term sustainability of the programme and enable us to build on the foundations laid by the original EPSRC investment, continuing to train world‑leading researchers and advance cutting‑edge work in foundational AI.

Funded PhD Opportunities:
Explore current opportunities for fully funded PhD studentships within the CDT in Foundational AI. This section provides details on available funding, eligibility, and how to apply. 
At present, the CDT is not recruiting new PhD students. New opportunities will be added as they arise, so please check back regularly.

People

Centre Management

• Prof. David Barber (CDT Director)
• Prof. Lourdes de Agapito Vicente (CDT Co-Director)
• Prof. Danail Stoyanov (CDT Co-Director)
• Prof. Niloy Mitra (CDT Co-Director)
• Prof. Gabriel Brostow (CDT Co-Director)
• Claire Hudson (CDT Centre Manager)

Impact Report

This report introduces the students and highlights the key achievements since the centre opened in 2019. We aim to provide a clear picture of our impact and the foundation of the CDT we’ve built for continued success in the years ahead.

Selected publications

• R. Bendikas, V. Modugno, D. Kanoulas, F. Vasconcelos and D. Stoyanov, “Learning Needle Pick-and-Place Without Expert Demonstrations,” in IEEE Robotics and Automation Letters, vol. 8, no. 6, pp. 3326-3333, June 2023, doi: 10.1109/LRA.2023.3266720.
• Bharti, Ayush & Naslidnyk, Masha & Key, Oscar & Kaski, Samuel & Briol, François-Xavier. (2023). Optimally-Weighted Estimators of the Maximum Mean Discrepancy for Likelihood-Free Inference. 10.48550/arXiv.2301.11674.
• Lewis-Smith, Andrew and Jaš Šemrl. “Implication Algebras and Implication Semigroups of Binary Relations”  In proceedings of RAMiCS 2023.
• Jipsen, Peter and Jaš Šemrl. “Representable and diagonally representable weakening relation algebras”  In proceedings of RAMiCS 2023.
• L Schaefer, O Slumbers*, S McAleer, Y Du, SV Albrecht, DH Mguni - Ensemble Value Functions for Efficient Exploration in Multi-Agent Reinforcement Learning - arXiv:2302.03439
• Kirk, R, Zhang, A, Grefenstette, E. Rocktaeschel, T “A Survey of Zero-Short Generalisation in Deep Reinforcement Learning”, Journal of Artificial Intelligence Research (JAIR), Vol 76 2023
• V Zantedeschi, L Franceschi, J Kaddour, M Kusner, V Niculae - DAG learning on the permutahedron -ICLR 2023
• A Lynch, GJS Dovonon, J Kaddour, R Silva - Spawrious: A benchmark for fine control of spurious correlation biases - arXiv preprint arXiv:2303.05470
• Y Yin, J Kaddour, X Zhang, Y Nie, Z Liu, L Kong, Q Liu - TTIDA: Controllable Generative Data Augmentation via Text-to-Text and Text-to-Image Models - arXiv preprint arXiv:2304.08821
• J Kaddour - The MiniPile Challenge for Data-Efficient Language Models - arXiv preprint arXiv:2304.08442

Partnerships

It is through the generosity and support of our partners that we are able to train the next generation of leaders in foundational AI. By working closely with our partners, we are able to understand their requirements and formulate student projects which match those needs. See below for further information on our current partners. There are many ways for partner organisations can get involved, from project creation and sponsoring studentships, to training and in-kind contributions.