Artificial intelligence will face a range of opportunities and challenges, and policymakers need to consider more research into some areas of the technology, according to a forum convened by GAO.
The Forum on Artificial Intelligence was convened by the Comptroller General and included participants from industry, government, academia and nonprofit organizations. It considered the future impact of AI in four main areas: cybersecurity, automated vehicles, criminal justice and financial services.
Forum participants agreed that investment in AI technologies could lead to improvements in productivity and economic outcomes, but that data used by AI could be vulnerable to hackers. Within cybersecurity, AI could be used as a tool for protecting and defending against attacks but policymakers need to consider how autonomous systems can be made secure, and whether machine-learning algorithms adhere to ethical norms.
The forum agreed that automated vehicles hold promise for enhanced mobility but policymakers should consider the appropriate regulatory framework for automated vehicle safety assurance.
The use of AI within criminal justice could help with the allocation of law enforcement resources but also raises questions about civil rights and privacy violations.
Within financial services, AI could improve client services and enhance surveillance monitoring but policymakers need to question the necessary mechanisms to address ethical considerations, trade-offs and protections.
Forum participants also raised some other challenges in relation to AI: for example, the potential for data used to be biased, along with a lack of access to computing resources or sufficient human capital. The forum also found that the widespread adoption of AI also raises questions about current laws, regulations and ethical frameworks.
The particular policy issues that forum participants believed required further attention were data-sharing and providing mechanisms for sharing sensitive information while protecting public privacy, improving safety and security and updating the regulatory approach that will affect AI.
Participants also highlighted several areas of AI that they believed warranted further research which include establishing regulatory sandboxes, developing high-quality labeled data, understanding the implications of AI on jobs for the future, and exploring computational ethics and explainable AI.