Image
Loading...
In late August, the MPA cohort returned to Durham for their second in-person residency and their first on Sanford’s “home court.” After beginning their journey in Washington, D.C. this summer, the cohort arrived at Duke for five packed days of coursework, professional development, and community building. The week included a civics immersion in Raleigh, networking with faculty and fellow graduate students, and sessions designed to prepare students for their fall courses. But it was the culminating keynote dinner, a Saturday evening panel on artificial intelligence at The Commons in Duke’s Brodhead Center, that left the deepest impression.
A Week Rooted in Community and Practice
The residency began with opportunities to connect across the broader Duke community. Students toured campus, joined lunches with peers in the MPP and MIDP programs, and sat for optional headshots on Sanford’s lawn, small but memorable rituals that mark the transition into graduate study.
Thursday’s offsite day brought the cohort to Raleigh for an inside look at North Carolina state government. They met with senior executive branch officials, legislators, and staff, hearing firsthand how state-level dynamics mirror and diverge from federal politics. Friday and Saturday featured a rhythm of classes, workshops, and career-focused sessions, from an introduction to Sanford’s career services to a workshop on leading inclusive organizations.
All of it built toward Saturday evening, when the cohort dressed up and gathered at The Commons for the residency’s capstone event. Against the backdrop of Brodhead’s gothic arches and modern glass, the group sat down to dinner and an animated conversation on the future of artificial intelligence and its implications for policy and governance.
It is unbelievably difficult to compete in AI. You need massive amounts of data, compute, and talent. The question is whether regulation makes it even harder, or whether we can create a regime where smaller companies also thrive.
Matt Perault (MPP'08), Head of AI Policy at Andreessen Horowitz
A Panel of Practitioners and Scholars
The keynote brought together voices at the center of the AI policy debate. Sanford’s Robyn Caplan, an assistant professor and expert on platform governance, moderated the conversation. The panelists included:
- Ronnie Chatterji, Duke professor and Chief Economist at OpenAI, who has served in both the Obama and Biden administrations.
- Matt Perault, Head of AI Policy at Andreessen Horowitz and a former Facebook policy leader, now a Senior Fellow at NYU’s Center on Technology Policy.
- Anne Washington, Sanford’s new Rothermere/Harmsworth Duke Associate Professor of Technology Policy, a scholar and public interest technologist with experience at Apple, CRS, and NYU.
The discussion unfolded less like a lecture and more like a spirited roundtable, with Caplan posing questions and the panelists trading perspectives, disagreements, and personal reflections.
Image
AI’s Transformative Potential and Its Limits
Chatterji opened by reflecting on what sets AI apart from technologies that promised transformation but fell short. “We reached 100 million users in two months,” he noted of ChatGPT’s launch. “That was a product that didn’t have any marketing. It wasn’t even supposed to be a consumer product. That scale of adoption is something we haven’t seen with technology before.”
He highlighted three factors that make him bullish on AI’s trajectory: rapidly increasing capabilities, dramatic cost reductions, and the sheer convenience of use. “Capabilities are actually increasing really dramatically… researchers have this look in their eyes because they are witnessing a revolution in their field,” he said.
Yet Chatterji also emphasized that AI is not a plug-and-play solution, especially in large organizations. “You can’t just give someone an AI and say, ‘go type in ChatGPT and find efficiency.’ Change management is required, and if you don’t consult your workers, it’s not going to be successful at all.”
That tension between promise and practice surfaced repeatedly throughout the evening. AI, the panelists agreed, may reshape economies and professions, but only if institutions adapt with intention.
Jobs, Knowledge, and Misleading Narratives
Caplan pressed the panel on a recurring anxiety: job loss. She cited a Microsoft study that listed “historian” as one of the most likely professions to be replaced by AI, a conclusion she found absurd. Chatterji agreed, calling such task-based studies misleading.
“If you tell me, ‘Ronnie, your PowerPoint slides, your speech notes, your lecture class, AI can do all of that,’ you might check those tasks off,” he explained. “But my job is not just those tasks. Jobs are indigenous to the environment, and they evolve.”
Rather than building policy around flawed predictions, Chatterji argued, governments should focus on equipping people with skills for the next wave of opportunities. “I don’t think we need a Historic Protection Act,” he quipped. “What we need is to give people the skills to pursue new opportunities.”
Caplan’s concern extended beyond jobs to knowledge production itself. “These systems are very good at summarizing. They do not produce new knowledge,” she warned, raising questions about what happens if AI undermines industries like academia and journalism that generate original information.
Governance, Competition, and the “Little Tech” Agenda
Where Chatterji framed AI in terms of economics and adoption, Perault focused on governance. From his vantage point at Andreessen Horowitz, he sees how policy decisions shape the ability of startups to survive in a market dominated by massive firms.
“It is unbelievably difficult to compete in AI,” he said. “You need massive amounts of data, compute, and talent. The question is whether regulation makes it even harder, or whether we can create a regime where smaller companies also thrive.”
Perault rejected the caricature that venture-backed firms want no regulation at all. “Our funds run on a ten-year cycle. If AI crashes in three years, that’s bad for us. We need a healthy, thriving market.” For him, the key is focusing governance on harmful uses of AI, rather than restricting the science of its development. “If harm happens, is someone held accountable? That should be the question.”
He also touched on the politics of AI policy, noting that his work engages both Democrats and Republicans. “We are truly bipartisan in our policy work,” he said. “My hope is for a bipartisan consensus around smart policy.”
Technology for the Public Interest
Washington brought a complementary perspective, reminding the audience that technology must serve everyone, not just industry or affluent consumers. “Governments have to worry about the person without broadband, or the one still using AOL dial-up,” she said. “We have to decide when to take the stairs and when to take the elevator. If we only build elevators, we risk losing the skills we need to keep moving forward.”
She raised practical challenges of deploying AI in government: scaling successful pilot projects, maintaining systems across political transitions, and investing in institutional capacity. “It’s a perennial problem: nobody wants to pay for maintenance,” she observed. “But if we don’t, every new administration just starts over, and we lose continuity.”
Washington also highlighted AI’s environmental and resource costs, noting the tension between scrappy innovation and resource-intensive corporate models. For her, governments in particular must be discerning about what kinds of AI systems they adopt and how they sustain them.
I'm immensely grateful for the quality time with other members of the cohort this week. We are trusting one another more and becoming the ‘3 AM’ (ok, maybe 9 AM?) peer support group.
MPA Student Fiona Bell
Image
Lessons for Policy Professionals
For the students gathered in The Commons, the evening was both a window into cutting-edge debates and a mirror reflecting their own career paths. Panelists offered direct advice as well:
- Chatterji urged students to think strategically about becoming generalists or specialists, and to maximize Duke’s network.
- Perault reminded them not to burn bridges. “The person across the table today might be your collaborator tomorrow.”
- Washington underscored the value of the cohort itself: “You’re building your 3 a.m. call friends, the people you’ll lean on when something goes sideways.”
Fittingly, the warm lights and festive songs of Duke’s semester celebrations could be seen and heard (billowing out from Duke’s plaza) as Sanford’s first MPA cohort finished out the evening.
The residency week showcased the hybrid spirit of the MPA program: rigorous classes, exposure to real-world policy environments, and conversations with practitioners shaping the frontiers of public policy. The keynote panel on AI crystallized that blend, part seminar, part strategy session, and part inspiration for the work ahead.
Most importantly, this week offered what Sanford does best: community-building.
MPA student Fiona Bell wrapped it up perfectly: “I'm immensely grateful for the quality time with other members of the cohort this week. We are trusting one another more and becoming the ‘3 AM’ (ok, maybe 9 AM?) peer support group.”