Steps Suggested for Trump to Boost Government AI Integration:
The Office of Management and Budget (OMB) has unveiled two memos aiming to speed up artificial intelligence (AI) integration across federal agencies. These memos provide guidance on enhancing public services and efficient procurement of AI systems. However, for the government's AI initiative to progress, other aspects of the administration must align, preventing inconsistencies or inaction.
Effective AI adoption doesn't happen merely by directing agencies to adopt technology. To realize this vision, there must be clarity on the desired performance from AI systems, the technical features that can deliver these results, and the means to verify if the systems perform as intended. As of now, these elements lack a solid foundation. To rectify this, the White House should consider the following steps:
- Specify AI Performance Outcomes for Federal Agencies:
Defining specific AI performance outcomes is crucial for effective adoption. This entails understanding the type of AI performance relevant in each agency's domain. For instance, the Department of Justice might focus on ensuring fairness and minimizing bias, while the Department of Energy may prioritize secure, resilient systems for critical infrastructure. Clarity on performance outcomes will enable better procurement decisions and offer valuable insights to the private sector.
- Prioritize R&D Focusing on Features Connected to Outcomes:
Even if agencies identify the outcomes they desire, such as fairness, reliability, or security, knowing which technical features achieve these ends remains essential. This is similar to the cycling world, which developed this knowledge over time through concerted research and testing under real conditions. The White House should urge the Networking and Information Technology Research and Development (NITRD) subcommittee to update the National AI R&D Strategic Plan with a focus on studying how technical parameters relate to measurable AI performance outcomes.
- Support NIST's Ability to Develop Evaluation Protocols:
Understanding performance outcomes and the technical features that create them is essential but insufficient for effective adoption. Agencies also need to test whether AI systems perform as intended. The National Institute of Standards and Technology (NIST) has initiated efforts to develop evaluation methods for this purpose, though budget cuts pose a threat to its technical capacity. The preservation of NIST's capacity is essential for turning emerging design knowledge into the evaluation infrastructure required to scale AI adoption across the government.
Agencies have already identified more than 1,700 AI use cases, yet most are still experimenting without the technical foundation needed to scale AI effectively. The administration must address these gaps to realize its vision of accelerating AI adoption. Adopting these recommended steps would significantly help translate the administration's plans into tangible action.
Source: Anna Moneymaker / Getty Images
- The initial memo from the OMB should include a clear policy on specific AI performance outcomes for federal agencies, considering unique AI performance needs across various domains, like the Department of Justice focusing on fairness and minimizing bias, or the Department of Energy prioritizing secure and resilient systems.
- The National AI R&D Strategic Plan, overseen by the NITRD subcommittee, needs to prioritize research and development that focuses on understanding how technical features translate to measurable AI performance outcomes.
- To ensure industry-wide applicability, the White House should consider defining and sharing these performance outcomes, which would offer valuable insights to the private sector.
- As AI systems must be evaluated to verify their performance, the government should support the National Institute of Standards and Technology (NIST) to develop comprehensive evaluation protocols and safeguard its technical capacity, despite potential budget cuts.
- A more coherent approach to innovation, including in policy-and-legislation and politics, is crucial to address the identified gaps in AI research, development, and implementation.
- Sharing findings and best practices from the government's AI research, including case studies and evaluation results, would contribute to the general news discourse on AI technology and encourage further innovation.
