Skip to content
News

Ada Lovelace Institute responds to UK Compute Roadmap

Matt Davies, Economic and Social Policy Lead at the Ada Lovelace Institute, has responded to the UK Compute Roadmap.

Reading time: 2 minutes

Responding to the publication of the final version of the UK Compute Roadmap, Matt Davies, Economic and Social Policy Lead at the Ada Lovelace Institute, said:

‘Public compute – state provision of computing resources – could be a powerful tool for reshaping the AI sector. By reducing costly dependencies and giving a wider set of actors access to the resources they need to participate in AI research, investment in public compute infrastructure can help to bring about a more diverse and pluralistic AI ecosystem that works for people and society.

‘The UK government’s Compute Roadmap is a promising step in this direction, outlining how government intends to dramatically increase the scale of compute resources available through public entities. We welcome the government’s commitment to a blended allocation model – supporting safety initiatives and a variety of AI projects, not only large scale “frontier” research – and its stated intention to draw on a wide range of suppliers, reducing the risk of vendor lock-in. Much of what is contained in the roadmap aligns with recommendations made in our Computing Commons paper, published earlier this year.

‘The roadmap makes numerous references to the energy demands of compute infrastructure, suggesting that procurement will be used strategically to promote the development of more energy-efficient hardware. This is a positive step, but we think there is scope to go further, in particular by requiring information from vendors on climate and environmental metrics. In lieu of regulation requiring standardised reporting from data centre owners and cloud service providers, procurement for public compute provides a lever for increasing transparency in a sector where ecological impacts are too often obfuscated.

‘We would also like to see greater clarity from government on how it will ensure that the value created by these significant investments will accrue to the public and the wider AI research ecosystem, not only or primarily to incumbent commercial players. Our global mapping of public compute policies found a role for mechanisms such as licensing conditions and public stakes in ensuring that projects receiving public support drive the creation of public value. In other areas of policy – most notably energy and climate change, through the creation of GB Energy as a public investment vehicle – government has proved willing to explore bold measures of this nature. Similar creative thinking would be welcome to ensure that investment in public compute delivers for people and society.’

Related content