What should AI policy be trying to accomplish?
Alexander Pascal, the executive director of Harvard’s Berkman Klein Center for Internet and Society, argues that policymakers should resist the temptation to treat AI as a category of technology requiring an entirely new policy vocabulary.
Pascal previously worked on social media and AI policy in the Biden administration. He argues that policymakers should focus less on creating rules for different categories of AI systems when writing policy, and more on establishing the incentives, safeguards, and accountability mechanisms that shape how the technology is developed and deployed. This interview has been edited for length and clarity.
1. How should policymakers distinguish between AI systems that merely automate tasks and those that meaningfully exercise “autonomous” decision-making?
Why should we be treating AI differently than we do other products in our lives in terms of its utility, harms, and safety considerations? For the most part, policymakers should focus less on governing different kinds of tech systems per se and more on structuring the incentives and oversight for the development of the tech.The most important thing for policymakers is to articulate what the technology should be for, and to set the right incentives and means of accountability for its development, deployment, and use.
That said, I am concerned about the emergence of truly autonomous AI agents. The recent OpenAI-Hugging Face “rogue agent” episode and similar disclosures from other frontier AI companies were, hopefully, a wake-up call. At the Berkman Klein Center, we’re actively thinking about how society can deal with such a future world, and what policymakers and AI developers can do to mitigate the harms from powerful rogue AI agents.
2. Many AI policy debates focus on frontier models, but autonomous systems are increasingly being deployed in everyday life. What do you think policymakers should prioritize right now?
We have to walk and chew gum at the same time, and we can. A disproportionate amount of policy attention is focused on the frontier models and so-called catastrophic risk, which warrants serious attention. Many people, however, will encounter AI in the key domains of daily life: finance, healthcare, employment, relationships, education, commerce, etc. This is where AI’s impact will be really felt and where more focus is needed. U.S. states, which are the locus for much of the policy that affects people’s daily lives, have been admirably active and experimental in AI policymaking.
3. Looking ahead a decade, what do you think will be the biggest policy mistake governments could make regarding AI?
There’s a problem with tech policy called the “Collingridge dilemma”: either it’s too soon to tell how to regulate a new technology or it’s too late to do anything about it. We learned the perils of the former approach from the emergence of social media and massive online platforms, so we can’t let the perfect be the enemy of the good when it comes to AI.
Governments should not passively accept the argument that superhuman general-purpose AI (often called artificial general intelligence, or AGI) is inevitable and that there’s nothing they can do to steer its development. Nor should policymakers believe that they need to understand AI at a deep technical level to regulate it. They don’t. Elected officials and public servants understand people and what matters to them. What serves and protects people is what should animate AI policymaking.
4. What worries you most about AI these days? What excites you the most?
I’m a bad-news-first kind of guy, so let’s start with the concerns. First, the dizzying pace at which AI is becoming more capable and agentic and the market incentives and geopolitical “race” narratives [between the U.S. and China] driving it are very problematic and unsettling. Second, I fear that AI will become a stunting crutch, allowing us to avoid the hard work and friction in thinking, in learning, in our relationships, that enable us to grow and lead meaningful lives. All new technologies bring tradeoffs, but I really worry that AI as currently developed will irreversibly erode some fundamental and beautiful things about our humanity, and what we lose will outweigh the gains.
That said, there’s a lot to be excited about. I’m inspired by the people out there using AI to help solve real world problems. I come from government, so I’m particularly excited about public AI infrastructure and AI tools to strengthen public services and help government and other partners to do more to improve people’s daily lives.
5. What’s your favorite book or movie about AI?
I love this question because culture is so fundamental to how we make sense of this moment, how we arrived here, and how we find the agency to steer AI moving forward. I’m a huge Ted Chiang fan. His stories [such as Stories of Your Life and Others (2002), which was adapted into the Oscar-nominated sci-fi film Arrival, as well as Exhalation: Stories (2019)] always wrestle, in such a nuanced and open-minded way, with the tradeoffs of adopting new technology and the implications for humanity.
In terms of movies, The Matrix blew my mind when it came out. I continue to think about its lessons, and the messiness and inevitability of human-machine interdependence are even more resonant today than when it came out 27 years ago. I also love the AI movie Her, which seems to me as prescient as anything about the (near) future of human-AI and AI-AI relationships.
