When the Tory government was still in power, the UK's unelected peers called for the creation of a made-in-Britain "sovereign" large language model for the public sector.
Now the plan to build a bespoke LLM has been criticised as "pointless" by a distinguished machine learning pioneer appointed to the House of Lords just six months ago.
Lionel Tarassenko, President of the University of Oxford's Reuben College and leader of its AI & Machine Learning research cluster, spoke at a Lords debate last week to discuss the contribution of science and technology to the UK economy.
He foreshadowed his criticism by expressing surprise that the debate did not mention the recent awarding of the 2024 Nobel Prize for Physics to Professor Geoff Hinton and for Chemistry to Sir Demis Hassabis, before going on to consider what it would take for Britain to build a deep-tech startup to become the UK’s first $1 trillion company.
The Lord warned that the increase of undergraduate fees to well north of £9,000 is causing a "loss" of PhD students in STEM subjects, citing figures that indicate there has been a 39% decrease in the number of UK-based computer science graduates between 2019 and 2022.
"Many students from disadvantaged backgrounds will no longer consider PhD study as a financially viable option," he warned. "This trend must be reversed."
READ MORE: AI Scientist LLM goes rogue: Creators warn of "significant risks" and "safety concerns"
After calling for a "spinout system" to scale companies founded around universities and the creation of a "coherent industrial strategy", Lord Tarassenko turned his attention to a "sovereign LLM capability" for Whitehall, which was first recommended in a Lords report on large language models and generative AI published in February 2024.
"A sovereign UK LLM capability could deliver substantial value if challenges around reliability, ethics, security and interpretability can be resolved," the report advised. "LLMs could in future benefit central departments and public services for example, though it remains too early to consider using LLMs in high-stakes applications such as critical national infrastructure or the legal system."
"We recommend that the Government explores the options for and feasibility of acquiring a sovereign LLM capability," it continued. "No option is risk-free, though commissioning external developers might work best. Any public sector capability would need to be designed to the highest ethical and security standards, in line with the recommendations made in this report."
Discussing the proposed LLM (which, we argue, should be called ChatGB if it ever goes into production), Lord Tarassenko pointed out that big tech has already seized the advantage and is racing ahead of the old world in Britain and Europe.
"The House of Lords Select Committee report on large language models called for a 'sovereign LLM capability', but it would be pointless, because of the prohibitive costs, to compete with US big tech companies in training hyperscale LLMs such as GPT-4 or Gemini," he said.
"Instead... we should be backing UK companies developing trustworthy AI using medium-scale LLMs and proprietary datasets, giving them privileged access to our sovereign data assets."
Of the world's most popular LLMs, at least five were made in the US: OpenAI's ChatGPT, Anthropic's Claude, Google's Gemini, Meta's LLaMA 2, and X's Grok.
The French company Mistral has made LLMs, as has China's Baidu and Germany's Aleph Alpha.
Whilst a number of British firms are working in the LLM space, none of them (to the best of our knowledge) have become household names.
Meanwhile, in the US, the erstwhile Liberal Democrat leader Nick Clegg announced earlier this week that Meta's open-source models are now being used by government agencies, including the State Department.
Without mentioning the country that gave him a career, Clegg wrote: "In a world where national security is inextricably linked with economic output, innovation, and job growth, widespread adoption of American open source AI models serves both economic and security interests. Other nations - including China and other competitors of the United States - understand this as well, and are racing to develop their own open source models, investing heavily to leap ahead of the U.S."
Get in touch with jasper@thestack.technology if you've got a great idea about how the UK can regain its lost ground in the large LLM market.
