The question of whether or not the government should use AI to solve their problems is extremely broad. The answer is: it is complicated. Idealistically, the answer seems easy enough: if it is used safely and responsibly then yes, of course[2].
The phrase “The Government” is a likely stand-in for a wide range of concepts and organizational bodies that range from small village to the United Nations, and the application of machine learning at each level would involve unique use-case combinations. For example, some municipalities are contracting companies like Gravy Analytics’ subsidiary Venntel to use machine learning to optimize traffic patterns. This same company also helps ICE track movements near the border. They are also contracted by the US Department of Defense offering curated telemetry data for counterintelligence and threat detection. All three use cases stemming from this one company raise different questions and bring with them their unique set of complications and potential for abuse[1].
One determining factor on the appropriateness of AI for a particular use case is the level of autonomy and trust we allow these systems. I think we can all agree that better traffic, less fentanyl, and more soldiers coming home still breathing are all positive things. The expectation for optimization bringing increased efficiency and productivity is just as real as the potential for hallucination and governance-breaking bringing harm or even death.
The safety mechanism - that we already possess – needed to shepherd the transition and guard the gate is the human in the loop. Machine learning an agentic system bring insights and broaden the conceptual footprint, but authority has to stay within the purview of the human operator.
RReferences
- Venntel. Mobility data.↩
- OECD. (2024). Governing with artificial intelligence: Are governments ready? OECD Artificial Intelligence Papers, No. 20. OECD Publishing. doi:10.1787/26324bc2-en↩