UK's AI Leadership Goal 'Unrealistic,' Experts Warn

The British government's ambitions to turn the United Kingdom into a global leader in artificial intelligence are "unrealistic," experts warn, adding that legal hurdles and lack of economic incentive pose major challenges.

See Also: Securing the Data & AI Landscape with DSPM and DDR

The Sunak government is slated to hold its AI Safety Summit in November to address challenges and identify opportunities presented by machine learning and deep learning technologies. The guest list reportedly includes U.S. Vice President Kamala Harris and Google DeepMind CEO Demis Hassabis.

"The aspiration for the U.K. to become a global leader in the development of the foundation models that support generative AI products and services is unrealistic," researchers from the University of Cambridge said in a policy brief. A major impediment is the lack of computing power required to build generative machine learning models, wrote report authors Ann Glenster and Sam Gilbert.

"Training foundation models requires vast amounts of compute, and little compute capacity is available in the U.K.," the researchers said. Instead of focusing on developing its own foundation models, they said, the British government should prioritize application of the existing large language models in different sectors to boost domestic AI and economic growth.

A majority of foundational model developers, including ChatGPT maker OpenAI, rely on cloud providers such as Amazon Web Services and Google Cloud services for computing capabilities. That's not an option for British companies, the researchers said, since offshoring sensitive information such as health data is "unpalatable" and "not reconcilable with U.K. law." They said Westminster should lobby companies such as AWS to establish computing clusters in the United Kingdom.

Jeremy Silver, CEO of London-based Digital Catapult, an agency that works with startups in emerging tech, said during a Tuesday parliamentary committee hearing on large language models that lack of government investment has prompted British AI and other tech startups to migrate to the United States, which has tax and business-friendly policies.

"In our experience, industry is not thinking about innovation as its priority. Rather, it is concerned about budgeting. That's one of the place where the U.K. suffers," he said.

Muffy Calder, vice principal and head of science and engineering at Glasgow University, told the committee that limitations on access to data required to train AI models is another looming challenge.

"We have really precious resources in terms of health, geospatial, environmental data," Calder said at the hearing on Tuesday. "And at the moment, the government has no clear guidance on how to value this data and make use of it."

Silver criticized the U.K. government's AI strategy in a policy paper published in March, which Silver characterized as "fragmented."

In the strategy, the Sunak government did not envision new legislation and instead stated that the principles of safety, transparency, fairness, accountability and competition can be implemented through existing institutions.

While developing "sector-specific frameworks" for AI is a more "pragmatic" approach in comparison to the European Union's "centralized" proposed AI Act, this approach could lead to "duplicated, or regulatory, frameworks that could lead to contradictory rules," th Cambridge researchers warned.

To avoid possible policy fragmentation, the researchers said the government could consider establishing an AI office to oversee and coordinate AI regulation across regulators.

In August, the U.K. lawmakers from the Science, Innovation and Technology Committee called on the government to speed up efforts to articulate a comprehensive artificial intelligence policy. If the EU's AI Act took precedence globally, they said, a different approach to AI governance would make it "difficult to deviate" (see: UK Lawmakers Call for Swift Adoption of AI Policy).