Do We Need A “Civilian ARPA” for AI?

In 1999, I read a great book by Bruce Sterling called “Distraction.” Billed as a fictional view into the status of U.S. public service in the year 2044, the technology and cultural ideas packed into the book still resonate more than 15 years later. The book, in part, motivated me to strive to make a difference in public service, if only to avoid some of the more dystopian views in the book. Apparently I’m not the only one who found the book packed with ideas, Cory Doctorow also wrote a great review of it in 2008.

The book’s central premise: that all of us could suffer “distraction” from what really matters, especially in a world with 300+ cable channels, a 24/7 news cycle, and always-on social media, is an idea I’d like to explore more fully given our rapidly changing world.

Technology is amoral; it is how humans choose to use it that determines good vs. bad outcomes. I’d submit that soon, perhaps very soon—if we can avoid being distracted from what really matters—we will need to have some substantial discussions and decisions regarding how we use technology to perform the business of running a nation and co-existing in the world.

Of particular note, the rise of machine learning and artificial intelligence (AI). How will this impact us as individuals and as a nation? And how will it affect public service?

Where Will Artificial Intelligence Improve Public Service the Most?

In the United States, public service should be about what “We The People” choose to do together that we can’t do by ourselves.If we are to be, as President Lincoln aptly described, a nation “of the people, by the people, for the people,” the question is then where do both machine learning and artificial intelligence fit into that vision?

In 2015, as an Eisenhower Fellow in Australia and Taiwan, I had several discussions with CEOs and government representatives about the rise of the Internet of Everything and how the devices, sensors, and the data streams associated with the Internet of Everything would begin to interrelate with machine learning. Since that time, we’ve continued to see impressive gains in what artificial intelligence can do as well as the rise of several open source AI projects that allow individuals to begin to experiment with AI at scale.

Expertise is only gained by doing experiments—something startups know quite well. Experiment and expertise both have the same root “ex peria” meaning “out of danger.” Experiments are risky, they don’t always work—however you are always guaranteed to learn something as a result of the experiment, even if it’s what to try differently for the next round.

We need a place, or places, in public service where we can collaborate with citizens and private sector partners to develop new ways to conduct the business of public service.

Given this need, I suggest we consider developing a civilian “ARPA for AI”—referring to what used to be the Advanced Research Projects Agency created in 1958 by President Eisenhower. ARPA invested in the early research and development (R&D) that gave birth to ARPAnet, now known as the internet. While ARPA now exists as DARPA with the addition of “Defense,” where is its civilian counterpart? Especially on the topic of AI?

There are several small, civilian organizations conducting scattered R&D, though none with the same success or focus as DARPA. Additionally, AI has moved beyond R&D. An “ARPA for AI” should perform experiments in the context of the organizations and processes it seeks to improve, as well as take time to listen and collect data on the processes to be improved both before and after implementing an approach. This would help reveal what works and provide tangible returns on investment to decision-makers about the benefits of such improvements.

I submit the potential benefits of AI to our nation and world are primarily in the civilian domain. The U.S. should help pioneer and show the world how AI can be used to make people more free, prosperous, and secure in keeping with the our Constitution. AI by itself is amoral, how we choose to use it determines whether it is good or bad.

The questions worth asking now: How can AI make us more free, less distracted, and more effective as individuals, as a nation, and—for those who choose to do as such—as participants in public service? By public service, I mean members of the public, private sector partners, and government professionals working together.

The Need for Deeper, Focused Substance vs. Quick Shiny

It would be disadvantageous to pursue the benefits of AI in an uncoordinated fashion across individual agencies. Agencies and departments are already fragmented (in part by design), and this would miss some of the deeper, substantial gains possible with rethinking how public service is delivered.

2016 marks 240 years since the events of 1776. The U.S. could use the next 10 years to perform a systematic “upgrade” in how it conducts the business of the people, of the nation, and of public service—just in time for the nation’s 250th birthday in 2026.

This would include rethinking beyond how we use technology to also rethink how we improve the stakeholder experience. Most importantly, we need focus on the meaningful elements of public service that could improve the U.S. and the world, and also focus on either automating or ending the less meaningful and often rote elements of public service that may no longer be necessary.

While startups often create a place where experiments and new methodologies can be tested, several elements of public service cannot fail—which means we will need to identify a systematic, substantive approach that identifies:

1. Which elements of public service must run-on-time and not fail, such as crucial parts of defense, the economy, etc.
2. Which elements of public service are most likely to produce significant “returns on investment” if new, better ways of performing them are found at the local, state, or national levels, and thus might be best for in-situ experiments.
3. Which elements of public service are rote or less meaningful in today’s rapidly changing world, and thus might be best decreased, automated, or halted?

These three questions will require us to dive into the substance of public service and take the time to understand the local, state, and national interactions involved in delivering such stakeholders services. It will require more than a quick, shiny gleam of ‘Let’s do artificial intelligence.’

Taking the time to listen, learn, and understand context will be essential.If we did decide to use the next 10 years to perform a systematic upgrade to public service, we will need to resist the temptation to pursue quick, shiny solutions and focus on tackling the hard, thorny, substantial issues that have been accumulating for decades.

Joy in Tackling the Really Hard Challenges

Three years ago, I parachuted into the Federal Communications Commission (FCC) as its new chief information officer (CIO). The FCC had previously employed nine CIOs over the course of eight years. One of the disconcerting trends I quickly noticed was a reluctance to tackle hard problems. These thorny issues were risky and complicated, and wouldn’t be solved overnight.

There was a huge incentive for folks to avoid them and instead pursue only quick, “shiny” activities that did not systematically improve how the organization delivered results to its stakeholders. Similarly, there was a huge disincentive to take risks and try something new.

Yet by the time I came on board, we had to tackle the hard, thorny, substantial problems because they had continued to grow—to include an ever increasing amount of our IT budget being spent just to maintain 207 different legacy systems and a general slowdown in how fast we could deliver new prototype solutions. I took professional joy knowing it was going to be hard, risky, and require a focus on substantial, long-term investments with measurable improvements along the way. In particular, this meant making a wholesale move to public cloud and commercial services, which back in 2013 was a pioneering effort.

A year ago, the FCC moved any remaining on-premise IT to public cloud and commercial service providers over Labor Day weekend. It took a team of hundreds of positive #ChangeAgents working day and night. Once we were done, we achieved:

1. Substantial improvements in how fast the FCC could prototype new solutions, from six months to 48 hours.
2. A sizable decrease from greater than 85 percent to 50 percent in terms of how much the FCC was spending to maintain legacy systems.
3. Unity among the team of #ChangeAgents to be creative problem solvers who continue to proactively search for new ways to deliver results.

That final and probably most important achievement—creating a team of proactive #ChangeAgents—is what we need now in public service.

Networks of Creative Problem Solvers Across Organizations

We need creative problem solvers who proactively search for new ways to deliver better results, especially when it comes to exploring how machine learning and AI can improve public service for us all.

AI, combined with the Internet of Everything and advances in the distribution of storage, processing, and services—to include interoperability across software-as-a-service and platforms-as-a-service cloud solutions and improved peer-to-peer, blockchain-like distributed services—holds the possibility of truly enabling us to be a nation “of the people, by the people, for the people.” We’ll need to move all agencies to cloud and commercial services, and we’ll need to make them more agile so they can then go further and explore machine learning and AI at an inter-organizational scale to improve public service.

No longer does the “business of government” have to be done just in D.C.; it can be distributed and shared in a way that’s open, visible, and participatory for all who want to be involved.

The FCC did a mini-example of this in late 2013, when we provided an open-source, downloadable app that allowed you to test your connection speed and, if you wanted, anonymously share that statistic by provider with the FCC to better inform decision-makers at the agency. This crowd-sourcing approach is just the tip of the iceberg as to what’s possible if people want to share data to improve the results and decisions of public service in aggregate.

In addition, lots of the hiring and procurement functions of public service are paper-, time-, and human-intensive. This is partly a result of the need for federal agencies to follow an equitable process. Yet an open-source AI could perform the same functions, and probably faster. Humans can still be in the loop for the deeper, more creative functions—and they could also help “teach the machine” in the months and years ahead as well, benefiting both in the process.

The result would yield more time for humans to perform more challenging tasks and better benefit stakeholdersas less paper, time, and work would be involved per hiring or procurement function. By making these tasks open-source, the public could verify the process is both equitable and not unduly biased.

Closing Thoughts

I’m passionate about making organizations more effective and adaptive in turbulent environments. The decade ahead will require us to continue to adapt at increasing speed. This is particularly true of the social institutions that help us be a nation “of the people, by the people, for the people.”

On the idea of a “civilian ARPA,” I’m not alone in such a proposal; the Partnership for Public Service recently released a report (see page 31) echoing the theme of encouraging and sustaining innovation. In addition, on the idea of using AI to improve innovation and policymaking, I have been invited, along with my friend and colleague Michael Krigsman of CxOTalk,to co-chair an Institute of Electrical and Electronics Engineers (IEEE) Subcommittee on these topics.

Positive #ChangeAgents are leaders who “illuminate the way” and manage the friction of stepping outside the status quo for benevolent ends. Anyone can be a change agent; you do not have to wait to receive formal authority to do so. If you have thoughts on how to make us more free, less distracted, and more effective as individuals, as a nation, and in public service—share your thoughts, as it will take all of us to transform public service for the challenges and opportunities ahead.

Onward and upward as positive #ChangeAgents.

About the author:

Dr. Bray currently serves as senior executive and CIO for the FCC. Through the efforts of a team of positive “change agents,” he led the transformation of the FCC’s legacy IT to award-winning tech in less than two years. He rolled out new, cloud-based solutions achieving results in half the time at 1/6 the cost. He received the Armed Forces Communications and Electronics Association’s Outstanding Achievement Award for Civilian Government. In 2015, he was chosen to be an Eisenhower Fellow to Taiwan and Australia and was named “Most Social CIO” globally by both Forbes and Huffington-Post, tweeting as @fcc_cio. In 2016, he was named a Young Global Leader by the World Economic Forum and co-chair for an IEEE Committee focused on artificial intelligence and innovative policies for the future. Business Insider also named him one of the top “24 Americans Who Are Changing the World.”

Photo Credit: NASA