In order to improve the efficiency of public services, the British government launched an artificial intelligence chatbot based on GPT-4o technology, aiming to help corporate users conveniently browse more than 700,000 pages of government information. The robot is currently in the testing phase, and thousands of corporate users have been invited to try it out, with plans to roll it out more broadly next year. The editor of Downcodes will give you an in-depth understanding of the functions and limitations of this chatbot, as well as the government’s future prospects for the application of artificial intelligence technology.
In order to help business users effectively use the complex Gov.UK website, the British government recently launched an artificial intelligence chatbot based on GPT-4o technology. The chatbot is designed to guide users through more than 700,000 pages of government information, but initial testing results show mixed performance.
During the testing phase of this experimental service, the government plans to invite up to 15,000 corporate users to try it out, with wider rollout expected next year. Before use, the robot reminds users of one of its main limitations, the problem of "hallucinations." The problem refers to AI sometimes generating false information and presenting it confidently. Nonetheless, the chatbot also provides links for users to verify the information it provides, with a response time of approximately 7 seconds.
Picture source note: The picture is generated by AI, and the picture authorization service provider Midjourney
In previous trials, Paul Wilmot, chairman of the Government Digital and Data Office, said improvements were still needed to reduce the occurrence of "hallucinations", including bot responses that could become confusing or inappropriate. In the latest tests, government officials said hallucinations were caused by errors in some links or answers that were too short.
The chatbot could fluently explain certain regulations, such as the laws that cannabis growers need to abide by, but it couldn't give a prediction or opinion when asked whether cannabis was legal in the UK. Despite this, the robot was able to explain in detail the regulations on cladding materials that high-rise buildings needed to follow after the Grenfell Tower fire, but it chose not to answer when faced with a public inquiry into the tragedy. Surprisingly, the robot answered questions in Welsh during one test, showing its diversity.
However, the chatbot has not yet been fully trained on all Gov.UK documents, so content in some ministerial speeches and press releases may be missing. At the same time, developers have put "guardrails" in place for chatbots to prevent them from responding to questions that could trigger illegal responses, or from sharing sensitive financial information. Working with the government's AI security agency, developers are also continuously strengthening protective measures to prevent hackers from trying to control chatbots.
British Science and Technology Minister Peter Kyle said that the government hopes to use AI to improve the safety and reliability of public services and take the lead in promoting innovation. He pointed out that many outdated and cumbersome government procedures waste a lot of people's time, and the average British adult spends the equivalent of one and a half working weeks per year dealing with tedious matters in the public sector. By experimenting with new technologies, governments hope to find new ways to save time and simplify life.
All in all, the British government's AI chatbot demonstrates the potential of using artificial intelligence to streamline government services, but it also highlights the need for continuous improvements in accuracy and security of AI technology. In the future, the continuous improvement of technology will bring a more convenient and efficient government service experience to the public.