AI experts argue that the UK should develop its own AI language model, akin to BritGPT, to mitigate national security risks and preserve competitiveness.

Adrian Joseph, BT’s Chief Data and Artificial Intelligence Officer, has emphasized the importance of the UK government investing in “large language models” like ChatGPT, Bing Chat, and Google’s Bard. Speaking before the Commons Science and Technology Committee, he cautioned that without such investment, the nation might experience a decline in competitiveness and face threats to national security. Joseph highlighted the risk of losing ground to major tech firms and China, especially in critical areas like cybersecurity and healthcare, amid an escalating global competition.

Dame Wendy Hall, co-chair of the UK government’s 2017 AI review, stressed the necessity of developing a BritGPT during a session with MPs. She emphasized that failure to create such technology could relegate the UK to a nation focused on service industries, while leveraging the technology could drive economic growth and generate job opportunities. Concerns were raised about the high cost of computing power for advanced AI work, putting UK researchers at a disadvantage compared to well-funded US companies. Nigel Shadbolt, Chair of the Open Data Institute, highlighted the risk of university researchers being left behind due to limited access to essential computing power, calling for a sustainable solution.

The Blair Institute’s report suggests an increased role for the UK government in directly developing large language models to establish independent capabilities. With leading private sector entities investing billions in these systems, the report underscores the need for the government to formulate policies promptly, providing domestic firms and the public sector an opportunity to bridge the gap.

The report recommends establishing a “sovereign general-purpose AI capability” through direct investments in supercomputing infrastructure, with a portion specifically allocated for training large AI models. Over time, this infrastructure should be considered a utility, similar to water or energy systems.

By admins