Cambridge AI Research refers to the academic study of artificial intelligence and AI alignment at the University of Cambridge. The university hosts several interdisciplinary research groups focused on machine learning, AI safety, and the societal implications of AI. These groups include the Leverhulme Centre for the Future of Intelligence (LCFI), founded in 2016, and the Cambridge Centre for AI in Medicine (CCAIM), established in 2020. Cambridge researchers have contributed to foundational AI concepts, including the development of the Transformer (architecture) architecture, and have published influential papers on AI ethics and governance.
Early Contributions and the Transformer
In 2017, researchers at the University of Cambridge, including Jakob Uszkoreit and Llion Jones, collaborated with Google to develop the Transformer (architecture) architecture, introduced in the paper "Attention Is All You Need." This architecture became the basis for modern Large language models such as OpenAI's GPT series and Google DeepMind's Chinchilla. Cambridge's Machine learning group, led by professors such as Christopher Bishop, has been active since the 1990s, producing research on Neural networks and Deep learning. Bishop's textbook, "Pattern Recognition and Machine Learning," published in 2006, is widely used in graduate programs.
Leverhulme Centre for the Future of Intelligence
The Leverhulme Centre for the Future of Intelligence (LCFI) was founded in 2016 with a grant from the Leverhulme Trust. It is an interdisciplinary center that brings together philosophers, computer scientists, and social scientists to study the long-term impacts of AI. LCFI has published reports on AI ethics, such as "The Malicious Use of Artificial Intelligence" (2018), co-authored with researchers from OpenAI and the Center for Human-Compatible AI. The center is directed by stephen-cave and has hosted annual conferences on AI safety, including the 2019 "AI and the Future of Humanity" event.
Cambridge Centre for AI in Medicine
In 2020, the University of Cambridge established the Cambridge Centre for AI in Medicine (CCAIM) to apply Machine learning to healthcare. The center is co-directed by mihaela-van-der-schaar and Andrew Lloyd Brown. CCAIM has developed AI models for predicting patient outcomes, such as a 2021 study that used Deep learning to predict mortality in COVID-19 patients. The center collaborates with Nokia Bell Labs and Samsung Research on projects involving Neural networks for medical imaging.
AI Alignment and Ethics Research
Cambridge has been a hub for AI alignment research, focusing on ensuring that Artificial intelligence systems behave safely and ethically. In 2019, the university launched the Centre for the Study of Existential Risk (CSER), which investigates catastrophic risks from AI. CSER researchers, including huw-price and sean-o-heigeartaigh, have published papers on AI governance, such as "The Role of International Cooperation in AI Safety" (2020). Cambridge also hosts the AI Ethics and Society research group, which examines bias in Machine learning algorithms and the societal impact of Generative AI.
Notable Researchers and Collaborations
Several prominent AI researchers have been affiliated with Cambridge. Jakob Uszkoreit and Llion Jones are alumni who later worked at Google DeepMind. Christopher Bishop is a professor at Cambridge and also holds a position at microsoft-research. mihaela-van-der-schaar is a professor of machine learning and a fellow of the royal-society. Cambridge has collaborated with Anthropic on AI safety research, including a 2023 workshop on interpretability. The university is also a member of the OpenPanel consortium, which promotes open research in AI. In 2022, Cambridge partnered with Arm Holdings to establish the Cambridge Centre for AI and Robotics, focusing on Neural network hardware.
Impact and Future Directions
Cambridge AI Research has influenced both academic and industrial AI. The Transformer (architecture) architecture developed with Cambridge input has become the standard for Large language models. LCFI's work on AI ethics has informed policy discussions at the european-union and the united-nations. CCAIM's medical AI models are being tested in clinical settings, with a 2023 trial at addenbrookes-hospital. As of 2024, Cambridge continues to expand its AI programs, including a new master's degree in AI ethics and a research initiative on AI alignment with OpenAI and Google DeepMind. The university's interdisciplinary approach ensures that AI research at Cambridge addresses both technical and societal challenges.