Prime Minister Rishi Sunak says he will take advice from the new £100m foundation model taskforce before deciding whether to launch a sovereign UK artificial intelligence model to compete with the likes of OpenAI and Google. Calls have been made by UK industry figures for such a model to be created, but speaking to Tech Monitor Sunak said the taskforce would be given the freedom to make recommendations about the future of AI policy.
Foundation AI models are trained on massive data sets and power tools like the ChatGPT chatbot from OpenAI, as well as image generators and other generative AI applications, including in the health and material sciences field. The largest models take significant and expensive compute time to train and are owned or controlled by big tech companies such as Google, Meta and Microsoft through its partnership with OpenAI.
Hyperscalers such as AWS and Google are starting to open access to alternative models from smaller startups like AI21 Labs and Anthropic but organisations like the Alan Turing Institute say it’s important to have a sovereign, government-backed model available as an alternative, but for some, this doesn’t go far enough and still creates a “lock in” for public data.
University of Oxford Professor and Alan Turing Institute director of foundational AI research Michael Wooldridge previously told Tech Monitor public sector organisations and researchers often have to sign non-disclosure agreements to access these models through hyperscaler cloud platforms. “The most important technology of the 21st century should not be the exclusive preserve of big tech.”
Finding the right approach
As part of the recent budget announcement, Chancellor Jeremy Hunt unveiled a £900m fund to improve compute power in the UK, particularly for smaller businesses, public sector organisations and academic institutions. This includes dedicated AI resources and an exascale supercomputer. The new foundation model AI taskforce, which is modelled on the Covid-19 taskforce established in 2020 will directly use that compute power and direct research into artificial intelligence models.
Sunak wouldn’t commit to having them direct some of the resource towards creating a sovereign model, dubbed BritGPT, but said it would be in the debate. “We’re trying to figure out what the right thing for us to do is, and that will be a mixture of things including looking at building foundation models, access to talent, and research and development,” the prime minister told Tech Monitor.
He said the taskforce would have the power to make those decisions, and the £100m in startup capital it is being given to capitalise on the benefits foundational model AI tools can bring to the economy. It is estimated billions could be added to the UK GDP by 2030 if it embraces “safe and well-deployed” AI across the economy.
“This is the first step and I think what we’re doing here in the UK is exciting,” Sunak said. “If you talk to someone like [former Google CEO] Eric Schmidt, who has a global perspective, he’ll say what’s happening in the UK is some of the most exciting stuff in the world. This is a fast-evolving area and we can’t pretend that we can stop it from happening or wish it away. The right thing to do is engage with [AI] intelligently so we can get the balance right between supporting innovation and the need to protect ourselves.”
BritGPT? Sovereign AI model could secure data
Speaking at a select committee hearing in February, BT chief data and AI officer Adrian Joseph said the country was in an AI arms race with the rest of the world and warned without the right investment and government direction the UK would be left behind. This included the need to build and train a BritGPT independent large language model. “There is a risk that we in the UK will lose out to the large tech companies.” without building such a system.
A number of UK AI companies have been sold to US and European rivals in recent years including DeepMind, the neural network research company founded out of University College London in 2010. It was acquired by Google in 2014 and has one of the largest large language models developed to date, called Gopher.
Dame Wendy Hall, regus professor of computer science at the University of Southampton also appeared before the panel of MPs last year and seconded the urgent need for better sovereignty over large language models and AI technology, particularly when used on NHS data. “We are at the beginning of the beginning with AI, even after all of the years it has been around, we need to keep our foot on the accelerator or risk falling behind,” Dame Hall said.
Woodridge said it is “utterly unacceptable for government organisations who want to use AI to feed sensitive UK data to the algorithms of Big Tech companies. The single biggest hurdle to overcome in developing a sovereign AI capability is our national lack of compute resources.”
One of the areas the taskforce has been told to focus on is the use of AI within healthcare, including using NHS-owned data to improve patient diagnosis, treatment and drug discovery but doing so in such a way that end-users, doctors and companies trust both the system being used and the output. This is an area where a sovereign model could be used.