Before ChatGPT arrived, governments were interested in using chatbots to automate services and advice.
Early chatbots were basic, with limited conversational abilities, according to Colin van Noordt, a researcher from the Netherlands focusing on AI in government.
The recent development of generative AI has brought renewed hopes for more efficient public services. These AI systems can simulate human-like responses and potentially handle inquiries about government services across various areas.
However, generative AI is known for occasional mistakes or nonsensical answers, termed as “hallucinations”.
In the UK, the Government Digital Service (GDS) tested a ChatGPT-based chatbot called GOV.UK Chat. Early findings showed about 70% found the responses useful, but there were concerns about some incorrect information being presented as factual.
Portugal also introduced an AI-driven chatbot, funded by the EU, to assist with basic legal questions. While it performed well on simpler queries, it struggled with more complex questions, highlighting ongoing challenges with reliability.
Experts like Colin van Noordt caution against replacing human civil servants with chatbots entirely.
They suggest using chatbots as supplementary tools for quick information retrieval rather than cost-saving replacements.
Estonia, known for its advanced digital services, is developing chatbots under Bürokratt, using older Natural Language Processing (NLP) technology.
This approach prioritizes accuracy and transparency over the conversational abilities of newer AI models like ChatGPT.
While LLM-based chatbots offer more conversational quality, they also pose challenges such as less control over responses and potential variations in answers to similar queries.
Overall, the debate continues on how best to integrate AI into government services, balancing efficiency with reliability and accountability.
Credit : BBC
https://www.bbc.com/news/articles/cw0y6wwn9yyo