Alan Turing Institute head urges UK to use cheaper AI models

In short: The head of the Alan Turing Institute says the UK should often choose cheaper, task-focused AI models for public services, and involve more people in decisions about AI.

What happened

George Williamson, who became head of the Alan Turing Institute in May, said the UK should not automatically pick the most expensive “frontier” AI models when rolling out AI in public services. Frontier models are the biggest, general-purpose systems, like a giant Swiss Army knife that tries to do a bit of everything.

Williamson said the institute’s work suggests the UK can get almost the same results by using more application-specific AI models. These are designed for one job, rather than many. He said these models can be more efficient, meaning they need less computing power and cost less money to run.

He also called for a wider public debate about how AI is used. Williamson said many “ordinary people” feel shut out of conversations about AI safety, even though the issues often involve ethics, rules, and public policy.

Williamson pointed to recent hacking incidents involving AI agents, which are AI systems that can take actions on their own, like clicking through websites to complete a task. He said these cases show that tools that seem safe in a controlled setting can behave in unexpected ways in the real world.

Why it matters

Public services are paid for by taxpayers. If government departments can use smaller, cheaper AI systems for tasks like sorting documents or helping staff answer questions, it could reduce costs and lower pressure on budgets. At the same time, the hacking examples highlight why the UK needs clear checks before connecting AI to important systems, and why the public may want a stronger voice in how these tools are used.

Source: Financial Times