UK in Shock AI Revelation

UK Houses of Parliament. Image by Diliff, CC BY-SA 2.5, via Wikimedia Commons

Anyone would think that the UK House of Commons was going to be replaced with a bunch of AI robots (hang on… is that an idea?).

The reaction to an article in the Times newspaper (paywall) about the Keir Starmer government’s proposed use of AI was startling. On Bluesky, a Guardian journalist thought it sounded like a joke. Some critics appeared to think that AI wasn’t intelligent enough to be trusted with the complexity of government and others that it was too clever by half, and it was irresponsible to let it have the reins of power (spoiler alert: it won’t).

All of these reactions display a lack of familiarity with AI which should not come at any great surprise, I suppose, given the hype that surrounds it. The term AI has travelled far from its origins as a discipline in Computer Science and has arrived as a vague notion in marketing.

There is no proper definition of intelligence so the argument about whether machines can or cannot be regarded as intelligent is not very meaningful. All we can really say is that a software system that can take the place of a human in a particular role and perform that role in a reasonably competent way could be regarded as an instance of AI.

Those who still argue that an AI translator, say, is not intelligent so cannot be regarded as AI might consider that Elon Musk’s network of communications satellites called Starlink doesn’t actually link any stars, Cillit Bang doesn’t explode when used and a Ford Mustang is not a horse. AI is clever software that is marketed as intelligent— maybe it is, but, then again, maybe it isn’t; it’s not important.

What is without doubt is that while AI is not about to take over the world, it is useful.

Most of the UK government proposals are entirely modest. Use text analysis to summarise large documents, making searching easier, and introduce an AI-based customer service system.

To AI developers these sorts of applications will seem run-of-the-mill. To people in large companies who have already invested in and are successfully using this technology, it will not sound ambitious.

But it’s a start.

The UK government like any other generates many documents, reports and memos, one is Hansard. This is a verbatim record of the debates in the House of Commons, the UK’s main legislative chamber. It records everything that it said and by whom in every Commons debate. It is the official document of record.

Currently, you can search an online version of Hansard for particular debates, parliamentary sessions or references to members. But it is tedious wading through such an inevitably large document and it is easy to miss something.

One of the tasks that the government expects AI to help with is to look at MPs’ contributions to debates to judge their views on proposed legislation and glean information about how that legislation might be adapted to be more acceptable or effective.

Getting an AI system to do this analysis is always going to be much quicker than employing a team of special advisors to do it by hand. The chances are that it will be more accurate, too. Yes, AI makes mistakes, it sometimes invents things — so-called hallucinations — but when it is given the information that it has to use and all the answers are somewhere in that information, then the chances of these errors are much reduced.

And, we shouldn’t forget, people make mistakes, too.

There are several projects mentioned on the Incubator for Artificial Intelligence (i.AI) website that the government is developing.

Redbox is intended to empower civil servants saving them time and increasing their efficiency. Redbox can summarize lengthy documents in minutes, a task that previously took hours.

Caddy is essentially an AI-powered customer service system. It will act as a ‘co-pilot’ for customer service advisors in the public sector by providing expert advice. The aim is to improve the accuracy and consistency of information given to the public.

Lex and Parlex are tools that simplify research by allowing users to ask questions in plain language. They process large amounts of information to provide summaries and analyses of that information to make it easier for users to find the information they need. Lex deals with the vast amount of legislative information the government holds, while Parlex deals with Parliamentary information such as Hansard.

Consult is designed to automate and improve the analysis of public feedback following government consultation. It is expected to save around £80 million a year

These are not rocket science. AI is good at these sorts of tasks where they are provided with a corpus of documents and, following natural language instructions can search, analyse and summerise those documents to produce a result. And they can achieve good results in a tiny fraction of the time it would take a human to do the same job. So, why wouldn’t you do it?

These projects directly impact government, but i.AI are working on other projects that are less clear in their aims. i.AI is working with the NHS to increase efficiency, they say, but there are no details about what it is they intend to do. The likes of Lex and the other applications I have mentioned are dealing, in the main, with information that is in the public domain. The NHS is obviously a different matter as it holds a great deal of private information which must be kept that way. The benefits of more rapid diagnosis and personalised treatments need to be balanced with the need for privacy. The scale of this type of project is very different, too, and big software projects, AI-based or not, are always tricky.

The government’s plans for AI in government are not scary. The technologies are already available, in use in business and industry and, used well, should lead to efficiency and cost savings.

It seems clear that the UK government’s AI initiatives represent a pragmatic step toward modernizing day-to-day government operations. While initial reactions may have been marked by scepticism and misconceptions, the proposed applications demonstrate practical uses of AI to enhance efficiency, accuracy, and service delivery.

The government needs to address those misconceptions and scepticism by being transparent about their aims and making a proper effort to inform the public and build trust that these new tools will serve the public good.

The potential cost savings, improved productivity, and better-informed policymaking surely make the effort worthwhile. The future of AI in government has the potential to reshape and improve how services are delivered and so we should applaud their efforts, not treat them as a joke.