The use of artificial intelligence (AI) by government agencies is increasing, TNO reports after extensive research. Compared to 2021 (165), more than 1.5 times more AI applications were found within the government, exploratory research shows. Municipalities in particular are large users: of the total 266 AI applications found, 105 (39%) were used here. Furthermore, transparency about the use of AI is also increasing, making it easier to find out exactly what techniques the government is using and for what purpose.
Why this is important:
AI has the potential to improve society. It also benefits the government.
Type of AI application and government organization
A large proportion of applications (105) were found in municipalities. Many applications are also developed in partnerships (46). This is followed by agencies (23) such as, for example, Rijkswaterstaat, DUP, and RVO.
For the first time, more examples of implementation (102) than of applications in the experimental phase (95) were found. This shows that the application of AI has become more mature.
Knowledge processing, archiving, and anonymization
AI is most commonly used for knowledge processing, archiving and anonymization (60) followed by inspection and enforcement (54). The increase in anonymization software can be explained by the influence of the Open Government Act (Woo).
Several examples of successful applications in the public domain are cited. Such as an anonymization tool for shielding confidential data, an algorithm that uses camera images to detect ship movements in the harbor, one for fisheries and for road inspections, and a virtual policy assistant for answering parliamentary questions faster and better.
The most commonly used techniques are image recognition (80) with speech and text recognition second (70). Interestingly, only 8 applications used generative AI and LLMs (large language models). This is partly because generative AI can be seen as an online service, mostly used by individuals within an organization.
Transparency
There is more transparency on the application of AI in public service by 2024. For example, it is possible to find out from more examples which AI technique was applied. This year, only 38 examples (14%) are unknown what type of AI technique was applied, compared to 72 (44%) in 2021.
Related Posts
