By Clifton Roberts, Global Director, Cloud and Data Policy
Times are changing. In the past, the notion of government data being released to the public was not only antithetical to U.S. government policy, but was illegal. Today, we are witnessing a significant shift in the government’s view on the matter. Aligned with Intel’s position that legislative and regulatory initiatives should support the free flow of data, the White House is embracing federal data as an asset that “provides the public with knowledge of government, society, economy, and environment – past, present, and future.”
Consequently, in June of this year, the White House Office of Management and Budget (OMB) issued its Federal Data Strategy, describing a comprehensive vision to responsibly liberate data in order to “support the foundations of democracy, deliver on mission, serve the public, and steward resources.” And in a public forum convened by the OMB and The Data Coalition at the Department of Commerce in Washington D.C. on July 8, I had the pleasure of joining other business leaders, academics, and members of civil society to offer Intel’s feedback on this strategy, specifically around an aggressive set of actions outlining a “firm basis of tools, processes, and capacities to leverage data as a strategic asset.” These actions, collectively referred to as the 2019-2020 Federal Data Strategy Action Plan, provides all federal agencies with clear guidance on how to manage and use federal data.
The time provided for each speaker was brief, given over fifty experts were assembled for the forum. Following are summaries of Intel’s recommendations on three crucial and foundational Federal Data Strategy actions.
Develop a Data Ethics Framework
Develop a Repository of Federal Data Strategy Resources and Tools
For this action, the OMB sought comments on priorities for populating a repository of resources and tools to assist agencies in implementing the Federal Data Strategy. Such a repository, also owned by the GSA, would include best-practice descriptions, case studies that demonstrate best-known-methods (in action), and tool kits for employing these practices. Of course, such priorities should be aligned to the most significant challenges faced by agencies in executing to the government’s call to responsibly liberate data.
Because data is stored across numerous federal agencies resulting in data silos, the mandate to migrate siloed data to a single location may not only prove to be very challenging, but also impractical, demanding the consumption of vast amounts of resources. Thus, a playbook for use of Secured Federated Machine Learning (SFML) should be a priority consideration in populating such a repository. SFML brings processing mechanisms to the data source for training and inferencing versus requiring that agencies migrate data to one place. Furthermore, according to Dr. Nikhil Deshpande of Intel’s Data Center Group, SFML allows for faster deployment and testing of smarter models, lower latency, and less power consumption, all while employing a combination of privacy-by-design techniques to ensure data de-identification and insights security.
Identify Priority Datasets for Agency Open Data Plans
The plan directs all federal agencies to “identify an initial set of priority agency datasets key to mission success as an initial focus for testing and implementing improvements to agency comprehensive data inventories and catalogs.” As the ultimate goal of the Federal Data Strategy aligns with Intel’s position for the responsible liberation of data that allows for data use while protecting individual privacy, mitigating potential bias, and enhancing cybersecurity, the datasets that have collected and harvested intelligence that helps to (i) advance U.S. industrial competitive advantage, (ii) improve citizenry quality of life, and (iii) advance national leadership in AI should be triaged as top tier, priority datasets as agencies’ preliminary focus for deployment testing and implementation. Furthermore, as agencies identify these priority datasets, investing in the development of voluntary international standards (i.e., for AI-related algorithmic explainability) and promoting diversity in datasets will serve as key factors to mission-critical success.
Overall, the Federal Data Strategy action plan is well-aligned with Intel’s data policy positions, as we have long maintained that governments:
- liberate public sources of information in systematized and user-friendly databases
- vigorously support the conception of reliable datasets to unleash the power of AI, and;
- promote incentives for data sharing between industry, the private sector, and the public
As Intel’s data policy leader, it was my pleasure and honor to share Intel’s positions with the OMB, and with others also offering their views on data as a U.S. strategic asset, and look forward to reviewing, and again offering feedback for, the next iteration of this initial draft Federal Data Strategy Action plan.