ai@cam: Cambridge University's Flagship AI Mission
ai@cam is the University of Cambridge’s flagship mission on AI. Over the last year it has supported 5 AI-deas projects, supported 46 research projects across 20 Departments, and engaged in policy and public dialogue to better steer national AI priorities.
In The Cultural Conundrum we reflected on the challenges of adapting existing culture to the disruptions of AI. The institutional approach at the University of Cambridge is our flagship mission on AI, ai@cam.
Cambridge has a tradition of being at the cutting edge of machine learning with cutting edge work in speech, language and computer vision. Today we enjoy the legacy of those achievements and a host of leading researchers, so why would we need a new initiative on AI?
It’s a good question, what would be the purpose of a new initiative at the University of Cambridge.1 Let’s assume that the initiative is needed. Then another question arises. Why hasn’t this initiative arisen naturally? Why do we need to intervene?
Given these questions, instead of launching a new institute, we chose to explore what the aspirations of the University’s AI researchers were and how we could suppot them. Across December 2021 to March 2022 we surveyed Cambridge’s AI community and ran a series of workshops with staff from across the University, both academics and professional services. We reported with a landscaping document in April 2022. It captured Cambridge’s aspirations:
- be a global leader in AI research, driving progress in technical domains, applying AI to tackle major scientific and social challenges, and aligning the development of AI technologies with societal interests; — lead the field in AI education, training the next generation of AI leaders and making AI accessible to all in society; and — become an innovation hub, and a go-to source of advice, that connects research, business, civil society, and policy leaders working to deploy AI for wider social and economic benefit, shaping the development and deployment of AI for the benefit of all in society.
We identified obstacles and enablers for those ambitions and placed them at the heart of the landscape document. We convened a steering committee selected from across the University.
From a projects perspective, we’ve launched 5 AI-deas projects. We see these as “academic unicorns”. Interdisciplinary projects that are focussed on societal impact and will leverage state of the art AI solutions to deliver that. They are focused on local decision making, climate & nature, women health, fertility and early infant neurodevelopment, mental health and language inclusion.
Alongside this we supported 46 research projects from 20 Departments access the new Dawn high-performance computing system. These included departments that traditionally exploit high performance computing (e.g. Materials Science), but also less traditional users in the humanities such as English and History.
Our policy lab has engaged in public dialogue in collaboration with the Kavli Centre, convened UK policy makers around questions of current interest,and worked with the Bennett Institute and the Minderoo Centre to provide policy reports. We maintain an agile group, currently 3 staff, and invest our funding into projects that cut across interdisciplinary boundaries and address the barriers to translation of solutions into practice.
Our philosophy is captured in our branding, which uses Turing Patterns (reaction-diffusion equations) to highlight how localised activity, if initiated well can cause organic global structure to emerge.
All these ideas build on previous experience from Data Science Africa, Accelerate Science and the Data Trusts Initiative.
Next steps are to focus more on the core machine learning community, support for innovation and support for translation. Watch this space!
-
One can imagine that it might be useful to raise the profile of the work, or to attract more funding, and many initiatives are launched with these objectives, but in my experience such initiatives tend to lower thinking and increase politics. ↩
Click to see what the machine says about the initiative and the book
Machine Commentary
NL: Can you provide further machine commentary on how ai@cam is fitting with the approach described in this post?
Claude: The ai@cam initiative demonstrates strong alignment with the five Ps framework:
Purpose: The initiative began by explicitly defining its purpose through community consultation, resulting in clear aspirations around research leadership, education, and societal benefit. Rather than imposing a structure, it let purpose emerge from stakeholder needs.
People: The approach centered on bringing people together, starting with extensive surveys and workshops across the Cambridge AI community. The steering committee’s cross-University composition reflects this people-centric approach.
Projects: The five AI-deas projects represent concrete, actionable outcomes that operationalize the broader purpose. Supporting 46 research projects across 20 departments shows commitment to turning purpose into practical action.
Principles: The initiative’s principles are reflected in its organic, bottom-up approach, symbolized by the Turing Patterns branding. Key principles include interdisciplinarity, barrier removal, and societal impact.
Process: The initiative maintains an intentionally agile structure with just 3 staff, avoiding over-bureaucratization while still having clear processes for project selection and support. This reflects the framework’s guidance about balancing process with purpose in young institutions.
The initiative particularly exemplifies the framework’s emphasis on starting with purpose and people before moving to process. By first understanding community needs and aspirations rather than imposing structure, ai@cam has developed in an organic way that serves its stakeholders while maintaining agility.