According to AI experts, the UK must create its own version of AI language model like BritGPT, or else it could lead to national security risks and loss of competitiveness.
BT’s chief data and artificial intelligence officer, Adrian Joseph, has urged the UK government to invest in “large language models” such as ChatGPT, Bing Chat, and Google’s Bard, during a hearing with the Commons science and technology committee. He warned that without such investment, the country could face declining competitiveness and a threat to national security. Joseph added that the UK may risk losing out to large tech companies and China, particularly in areas such as cybersecurity and healthcare, which are part of a “massive arms race” and a growing international competition.
MPs were told that developing a BritGPT was necessary by Dame Wendy Hall, who co-chaired the UK government’s AI review in 2017. She expressed that if the UK failed to create such technology, it would become a country of service industries. In contrast, harnessing the technology could drive the economy and create job opportunities. It was also noted that the computing power required for advanced AI work is expensive, which puts the UK’s leading researchers at a disadvantage compared to well-funded US companies. Nigel Shadbolt, the chair of the Open Data Institute, added that university researchers were at risk of being left behind because they lacked systematic access to the necessary computing power, and a sustainable solution must be found.
The Blair Institute report recommends that the UK government should play a more active role in the direct development of large language models to ensure the nation has independent capabilities in this field. As leading private sector entities are spending billions of dollars in developing such systems, the report highlights that the government has only a few months to devise policies that will allow domestic firms and the public sector to catch up.
The report further suggests that a “sovereign general-purpose AI capability” should be established with the aid of direct investments in supercomputing infrastructure, some of which must be specifically designated for training those large AI models. In the long run, this infrastructure should be treated as a utility, akin to water or energy systems.
