Transcript
Speaker: I think what we need to do is explain how our principles of free speech, free inquiry, will help serve the cause of justice. The First Amendment, the constitutional freedom of speech and freedom of conscience that is the bulwark of our democracy.
Speaker: There was a passion in what was being said, affirming this, what people consider a sacred constitutional right, freedom of speech and freedom of association.
Speaker: From the UC National Center for Free Speech and Civic Engagement, this is Speech Matters, a podcast about expression, engagement, and democratic learning in higher education. I'm Michelle Deutschman, the center's executive director. Typically, I'm your host, but today I'm handing the mic over to the center's assistant director, Beth Niehaus.
Speaker: Beth joined the center this summer after spending 13 years on the faculty of the University of Nebraska-Lincoln's Department of Educational Administration, serving as assistant professor, associate professor, and beginning 2025 as full professor.
Speaker: Her research examined teaching and learning in higher education and the institutional conditions that shape it, with particular attention to service learning, study abroad, campus climate, free speech, academic freedom, and most recently, artificial intelligence.
Speaker: While new to the role of assistant director, Beth is no stranger to the Center. Rather, Beth's involvement with the Center has spanned more than six years, including as a fellow, senior fellow, and curriculum consultant.
Speaker: We're thrilled to have her officially on the Center team, and I'm going to turn the rest of this episode over to her. Take it away, Beth. Thanks, Michelle. It's great to be part of the Center, and I'm excited for the conversation today about AI and Tobacco Seed.
Speaker: There is no question that generative AI has presented a number of challenges for higher education, from academic integrity and assessing student learning to workforce readiness and the value of a college degree.
Speaker: Today, though, we're going to discuss a different set of challenges. What AI means for higher education's larger democratic mission. What do students, faculty, and staff need to know about how AI is reshaping governance, democracy, and civic engagement?
Speaker: To help us dig into this very challenging topic, we will be joined by Nathan Sanders and Bruce Schneier, co-authors of the book, Rewiring Democracy, how AI will transform our politics, government, and citizenship.
Speaker: We will discuss how AI may help restore college students' faith in democracy, the biggest risks that AI poses to our political system, and of course, what this all means for higher ed. But before we dive in let's turn to class notes, a look at what's making headlines.
Speaker: The public comment period for the Department of Education's proposed overhaul of the higher education accreditation system closed this past Monday. The stakes are high given that the administration's proposed changes would give the federal government greater influence over what accreditors evaluate and how they operate.
Speaker: Critics fear this could transform accreditation from a quality assurance mechanism into a vehicle for political priorities and and that expanding the federal government's role in accreditation would undermine academic freedom and institutional autonomy.
Speaker: The Department of Education will issue a response and publish the final accreditation rules by November 1st. In news from the courts, a federal district judge of Massachusetts issued a preliminary injunction blocking the U.S. government from implementing a new rule that would limit stays for international students to no more than four years.
Speaker: In his ruling, Judge F. Dennis Saylor wrote that the policy would introduce significant uncertainty into the higher education system and rejected the government's claim that the rule was needed to protect national security.
Speaker: Saylor also expressed concern that government officials could deny extensions to students for politically motivated reasons. The ruling temporarily starts with the implementation of the international student visa restrictions, but is not a final ruling on the policy itself.
Speaker: In another loss for the Trump administration, the Supreme Court blocked the implementation of new restrictions on vote-by-mail ahead of the midterm elections. The ruling leaves current state election procedures in place and prevents changes that could have significantly affected voters who vote-by-mail, including many college students.
Speaker: The Trump administration argued that the restrictions were necessary to prevent voter fraud based on unsubstantiated claims of fraud in the 2020 presidential election. The Supreme Court's order applies to the upcoming midterm elections, while the broader legal challenge over the administration's authority to impose the restrictions is still to be resolved.
Speaker: If you are interested in hearing more about vote by mail and other issues facing student voters in the upcoming midterm elections, check out the bonus episode of Speech Matters posted last week, Ask Me Anything About Student Voting.
Speaker: Now back to today's guests, Nathan Sanders and Bruce Schneier, co-authors of the book, Rewiring Democracy, how AI will transform our politics, government, and citizenship.
Speaker: Nathan is a data scientist, physical scientist, civic technologist, and organizer in science communications, and currently serves as an affiliate researcher with the Berkman Klein Center for Internet and Society at Harvard University.
Speaker: Bruce is an internationally renowned security technologist, a fellow at the Berkman Klein Center, and a lecturer in public policy at the Harvard Kennedy School. Welcome, Nathan Sanders and Bruce Schneier.
Speaker: Thanks for having us. Thank you so much. Great to be with you. Well, as is tradition on the Speech Matters podcast, I wanted to start by asking you about your journeys. How did each of you become interested in studying issues around ai and democracy?
Speaker: Nathan, why don't we start with you? Happy to. My journey really starts when I was a student. I trained in the physical sciences and physics and astronomy, but I recognized that I had a really deep interest in democracy and how government serves people.
Speaker: And so I took the opportunity as a graduate student to do a science policy fellowship through the Rappaport Institute here in Boston at the Massachusetts State House. And I got to work with just amazing, inspiring legislators, ah folks like State Senator Pat Jalen and Representative Denise Provo.
Speaker: And it was an eye-opening experience. I got to see all the ways that an elected official who's really deeply committed to serving their constituents can make a big impact. And that was inspiring. And I also, as you can imagine, got to see examples of things that didn't work as well as they should in our state government.
Speaker: And so ever since then, I've you know really been devoted as a career, as a passion to figuring out how to use the types of skills that I have as a scientist, as a technologist to try and make government work better for people. And I come this through security, through cybersecurity, really thinking in terms of systems and how they can be abused, how they can be broken.
Speaker: And few years ago, I started thinking about economic, political, and social systems, which of course leads to both democracy and capitalism. And now that sort of AI is running headlong into those systems, I think a lot about how to make them work and how to make them work well versus how to make them work poorly.
Speaker: Mm-hmm. Yeah. And as Nathan commented on, a lot of times they do work poorly, unfortunately. you know, it's a human system that's been around for a long time.
Speaker: So before we get into the the nuts and bolts, for our listeners, i I wanted to start by defining our terms, because both AI and democracy can mean a lot of different things.
Speaker: So when you two are talking about AI and democracy, What do you mean by each of those terms? And Bruce, maybe we could start with you with AI. yeah So we are deliberately sloppy because we're not writing a textbook and our goal isn't to be precise. Our goal is to be able to discuss the issues.
Speaker: So where what we mean is a broad collection of technologies that mimic human cognition in some way. Now, there's a lot of edge cases. Is this in? Is this out? Yeah.
Speaker: We kind of don't care that much. It's changing. But it seems clear to us that there are these technologies that are engaging in thought in some way, that are mirroring things that used to be the exclusive purview of humans.
Speaker: And that's really what we're talking about. So, Bruce, you talk about AI as this big umbrella, and I think a lot of our listeners probably are thinking about chatbots, generative AI, things that, you know, students could use to write a paper for them.
Speaker: Can you give a few examples maybe of what AI means beyond a chatbot? It's interesting because chatbots is 2023 that, you know, the discussion today is not about chatbots. It's about agents.
Speaker: It's about AIs that autonomously do things that affect the world. And it's interesting. I mean, I think it is definitely interesting to watch people who don't have a lot of experience get their ideas of AI from blogs from the news or from videos they watch and are not really paying attention to things that are changing.
Speaker: So that's generative AI, AI that generates either speech or videos or images that were chatbots and are now more agents. But there's also a lot of predictive AI.
Speaker: it The AI that's giving you turn-by-turn directions when you pull up Google Maps is not chatting. The AI that's reading chest x-rays. The AI that is is driving a car.
Speaker: These are are our different sorts of systems, but they're all under the same umbrella. But it is important, I think, to watch and pay attention to what the technologies are doing today.
Speaker: mean as A few weeks ago, they're proving mathematical theorems at the level of of graduate students and mathematics professors. That wasn't a thing three years ago.
Speaker: And now it's kind of old news already because we've passed it by a few weeks. So these these things change all the time. Moving on to our other complicated topic. Nathan, can you tell us a little bit about what you mean by democracy?
Speaker: Yeah, I would acknowledge. I think how we define democracy is at least as complicated as how we define ai I think we analyze it in the book primarily in terms of how we share power across society.
Speaker: Who gets the resources? Who gets to make decisions that impact other people? Of course, part of that is elections, and it's the question of what individuals we put in positions of elected power. But of course, that's not all there is in democracy.
Speaker: In the book, we also talk about how decisions are made in operating government and administration, how we decide who is eligible for an important benefit under a program like a health care benefit. We talk about how judicial decisions are made, both in the public court systems and also in arbitration, how we sell disputes among each other.
Speaker: um We talk about how laws are made, how policy is made. We really try and look step by step in all the aspects of democracy across the three branches of government and citizenship as well.
Speaker: We try and think about how technology is already starting to impact each one of them, AI in particular. So young people in particular, seem to not have a lot of faith in democracy these days.
Speaker: but As a system for managing power, for getting things done, recent research from the Democracy Project and researchers at Tufts actually found that only about a quarter of their Gen Z respondents strongly agreed that democracy is the best system of government.
Speaker: And only about a third agreed or strongly agreed that U.S. democracy in particular can address address the issues we're facing. So Nathan, as a follow up, I'm wondering if you can talk about what you see as the most promising possibilities for AI that could improve how young people in particular experience democracy in practice.
Speaker: I love that question. If you don't mind, maybe I'll give two examples that we've written about. One is on how politics is done, how voters and constituents engage with the political process, how they engage with elections.
Speaker: I think there's a huge opportunity to leverage AI to make that kind of political process more accessible to people who are not insiders and to young people in particular. Bruce and I have been writing about and looking for examples of that across the United States, especially as we get into the midterm elections. And to be honest, I haven't seen all that many that seem all that inspiring to me.
Speaker: But we also see examples in other countries where I think people are using technology in a way that is more innovative in the sense of really making politics look different than it traditionally has. We've written a lot about the new Japanese political party called Team Mirai, which is full of primarily young people, primarily appealing to young voters, in part by leveraging technology as a way to get more constituent input and making decisions about their party platform and in using it to mobilize voters.
Speaker: They've built, for example, AI interviewer applications that allow an individual constituent to have a conversation with effectively a chatbot, but in a way that's not just ephemeral, but actually leads to a demonstrable recommendation to the party.
Speaker: You know, the voter had a conversation about this issue, and that results in a written recommendation to the party saying that they should change their platform and listen to that voter, and the party has the opportunity to respond. And we've written about some cases where the party has said, yes, we are going to listen to this individual.
Speaker: We're going to change our platform to adopt their recommendation, and the voter gets to see that. I think it's a powerful example of how mobilization and input and in party politics can be more responsive using technology.
Speaker: We also write about applications of AI and government administration. And here's where I hope young people will not feel resigned that the way AI is being rolled out today Under our current administration here in the U.S., that it has to be that way, but rather they should look at ai as an axis over which to demand things change.
Speaker: So the concrete example that I'm thinking of is the Department of Health and Human Services, under their previous administration, have put in place guardrails for how private insurers could or could not use AI in decision-making to say that someone is eligible to submit a health ah expense, and that would be covered or reimbursed.
Speaker: And the new administration, the current Trump administration, chose to reverse that guardrail and actually change how AI is used in the marketplace by incentivizing insurers to deny coverage on the basis of automated reviews of those claims.
Speaker: So the incentive that's being put in place now is for insurers to save costs by saying no more. That's not a feature of the technology. There's nothing about AI that says it has to be used to limit people's access to healthcare. It could be used for the opposite. It could be used to speed up approvals, to say that when someone really is eligible, that decision should happen ASAP, so they're not left waiting for care.
Speaker: That's the kind of demand I would like to see the public in general, but especially young people galvanize around it to make that change happen and not be resigned that AI has to be implemented in government the way it is today.
Speaker: So it sounds like there's a lot of potential for AI to be used in ways that make government more responsive to people's needs, that at the very least enable government government officials, political parties to gather and understand the preferences of voters.
Speaker: Beyond the the voter preferences piece of this, though, Are there ways that you're seeing AI being used or or ways that you can imagine AI being used, even if they're not happening yet, that would help people engage more and participate more directly in the political process?
Speaker: We see a lot of examples of ai being used to facilitate conversation, to facilitate negotiation, consensus building, you know, ground up, deciding what to do, you know, the democracy at at the most local level.
Speaker: And that's something that AI can facilitate. Now, lot of the AI uses we talk about are not unique to AI. Like we can imagine human facilitators doing all that work, but we don't have enough human facilitators. And AI can do things at a scale that is just not possible with with the humans we have.
Speaker: And there's a lot of of space between our individual preferences, our voices, and and policy. And largely, there's not a lot of conversation because it's hard.
Speaker: It's hard to call you a congressperson. It's even harder for them to listen because they're getting so many calls. And AI can facilitate a lot of those processes. Now, the people have to want to do it.
Speaker: but AI is not going to magically make your congressperson interested in listening to you. And I think that is a lot of dissatisfaction we're seeing and young people about democracy.
Speaker: Because democracy today, as it's practiced in many places around the world, doesn't really work for people. And we could blame all sorts of things for that. And AI can make it better. it doesn't It's not necessarily going to, it doesn't have to.
Speaker: But if we, if countries, if governments choose to make it better, AI can help. I was going to add that we've written about some of the examples springing up all around the world of how civic technologists are trying to reimagine how representative democracy works using AI.
Speaker: And I think there are really interesting examples of ah new forms of citizens' assemblies that can involve many, many people in a policymaking conversation, not just their elected representatives, but individual constituents to help drive policymaking at a way that's much more scalable than a traditional civic assembly, which might happen just with, you know, 50 people sitting in a room together.
Speaker: One of the examples of this that Bruce and I have written about is the platform being developed by the organization Crown Shy based in Scotland. It's been commissioned by the Scottish government. The Scottish government is looking for tools to help them hear more from Scottish constituents and to really prioritize responsiveness to the demands of the Scottish public public in parliamentary action.
Speaker: And so they've been working with a local Scottish company to develop tools that are also reusable, that are open source, that other countries, that other countries, other communities can use for decision-making as well.
Speaker: That's just one example. There are lots of efforts like that going on around the world. We've had AI-assisted or facilitated civic assemblies here in New England, in Maine and Connecticut over just the last couple months. And I'm really interested to see what kinds of impacts those have over time.
Speaker: I want to circle back It's related to the the citizen assemblies, but you know Bruce, a minute ago, you were talking about using ai to facilitate dialogue or conversation among people.
Speaker: And one of the hesitations that I hear from people about AI being used in that way is that there's this fear that you lose the humanness of dialogue, of that exchange of ideas when it is facilitated or mediated by a computer.
Speaker: And so I'm wondering if you have thoughts on how to keep the humanity present and and central in these types of dialogues, in these types of citizen assemblies?
Speaker: How do we use the AI for things that the AI is good for, but keep the things that human that humans really should be doing? Well, that's what the research is all about. And there are lots of examples of AI being used in ways that facilitate human connection rather than diminish it.
Speaker: know Our podcast right now is being facilitated by a computer. Everything that's happening is being facilitated by a computer. We are all in separate locations. And yet, maybe this is a human conversation.
Speaker: And possibly an AI will be used to summarize this. We can imagine AI being used to link people together. A group out of MIT, Cortico, is a project that they use AI as a listener and a summarizer and a way to move ideas between groups of people.
Speaker: And it it is things that skilled facilitators do. mean, sure, we we we can build an AI system that silences voices, that turns your opinion into a yes, no, that dehumanizes everything, but we don't have to.
Speaker: And the people doing the research are really looking into ways to use the technologies that make the human connection better. That you come out of the conversation but with a consensus, with the people believing they've been heard, they've listened to other people, they've learned things, they've come to conclusions.
Speaker: and And it's very powerful. But sure, you can use it in a dehumanizing way. In the end, the technology does what the humans want. And if the humans want more humanization, we can build a technology to do that.
Speaker: So we talked a little bit about voting towards the beginning of this conversation. and And with the midterm elections coming, if I want to circle back to that, Nathan, are there ways that you see AI being used or, again, you could imagine it being used to improve voter turnout and engagement?
Speaker: Well, let me talk about what I think is kind of the 800-pound grill in the room that hasn't come up in this conversation yet, which is ai is a political issue. And what I see as the biggest problem AI is creating in the world today, which is around the concentration of power and wealth and a very small number of companies and people at the scale of trillions of dollars, meaning trillionaires and trillion-dollar companies.
Speaker: I'm really actually encouraged to see how much attention in the political debate that issue is starting to achieve. It's been you know very, very much top of mind for Bruce and I. It's the primary issue that we wrote about when we talk about solutions in our book, Rewiring Democracy.
Speaker: And I really hope it will be front and center as, you know, recently former President Obama called for it to be in the political conversation the after the midterms. I think in terms of voter engagement, making sure that people are seeing a clear policy platform from each of our major political parties and our political candidates about what they will do to solve that issue is critically important.
Speaker: And then making people, making sure that people show up to vote so their voice is heard on that critical issue of concentration of power is important. To me, that is absolutely front and center. And Beth, to your question, I think there's also a lot that AI can do, not as a policy issue, but as a technology for voter engagement as well, although I think that's very much secondary to that primary issue.
Speaker: We talked earlier about that example of the political party in Japan, Team Uri, that's been, I think, really demonstrating what can look like to do politics in a fundamentally different way with AI. One of the other examples of what they've done is using AI to develop tools that ah organizers, political organizers can use to mobilize their constituencies.
Speaker: They've built a action platform, which is a a website online that helps helps organizers come together and engage volunteers in doing things like putting up posters around their neighborhood or showing up to speak at a local event on behalf of the party or political issue.
Speaker: And they've succeeded in this kind of gamified approach of giving people recognition and visibility into those actions that they're community members are taking, they've succeeded in mobilizing millions of those individual ground level political actions.
Speaker: That's not something that's really, it's not something that requires AI or that wouldn't have been possible to do before the technology existed. But what I think is exciting is that ah a very small political party with a small number of people can use the technology of AI to create digital technologies that have that kind of impact on organizing.
Speaker: And they can see the kind of on on the ground potential of organizing in a different way. I think AI as an assist assistive tool to help political movements and organizers make their work of working together with people physically on the ground in communities better is something that has a lot of potential.
Speaker: In the U.S., we know of several companies that are using AI in the political process, either to make get-out-the-vote campaigns easier. There's one company that is using AI to establish walking routes for door knockers.
Speaker: to make that process more efficient. There's a company that that uses AI to link candidates with donors. so that candidates can get better funded. There's an AI that helps candidates are run for office, file the paperwork and get signatures.
Speaker: And here, don't think Congress, don't think Senate, think city council. where There's no money, no but no time, no expertise. ah AI tools can make it easier for more people to run for office.
Speaker: ah These tools are all in development. i mean, the the varying stages of how well they work, but they are all ways to make democracy better. It sounds like there's a lot of examples in there that could be used by students or by student affairs professionals who are working around political engagement, voter engagement, lots of ways that that can be adapted by higher ed institutions to better engage students. in elections and in politics more broadly.
Speaker: so Nathan, I appreciated the first part of your answer there that you think the biggest issue around AI and elections is actually AI as a an issue.
Speaker: um And so I'm wondering if you can talk a little bit more about that, particularly about the relationship between how AI is currently being governed and its potential for influencing our larger democratic processes.
Speaker: Oh, such a huge topic area. Maybe I'll call out just two aspects of that, recognizing there's so much more we could say. First of all, in terms of how ai is being governed, obviously in the last few days, you know, we're speaking on September 17th.
Speaker: There has been so much discussion about AI safety and Dario Amadai's proposal and others' proposals about how regulation and international cooperation can help us as a society get more control over the far future development of AI.
Speaker: which maybe isn't all that far in the future in terms of its potential to pose significant or even existential risks. And I think that conversation is important. I hope that governments will put in place regulations. I mean, I think the simplest thing that Amadai had proposed is just the idea of having embedded regulators within companies working side by side with visibility to what the staff are doing.
Speaker: I think approaches like that that have proven, if not completely successful, at least helpful in other industries generally make sense. But I think they're also insufficient. I think in order to really enable the public and society to have a real tangible impact on AI development, we have to think about the economics of how AI is developed.
Speaker: So that's why I think proposals like some version of a token tax that would fundamentally change the economics of AI by putting downward pressure on usage and not allowing the sort of venture capital funded acceleration of AI development and usage that we're seeing today to better regulate that economic development are critical and should be discussed more.
Speaker: The other aspect I want to talk about is how government can shape and set baselines and expectations about how how AI is developed. Regulation is one way to do that, but it's not the only way. Another way is by setting a baseline for the ethical development and operation and even cost of AI by providing a public option for AI so that the technology is not only being developed in the largest corporations in the world that are clearly developed. working for their own financial interests, but there's also an option that's developed and explicitly aligned to the purpose of public benefit.
Speaker: Today in the US, we don't have that kind of a public option AI, but other countries are developing that. The example that Bruce and I know best and have written about the most is from Europe, from Switzerland, where they've developed a public AI model called Apertuse that is developed by Swiss institutions led by a collaboration of Swiss universities.
Speaker: Some of the ways that they've differentiated from the commercial AI developers is that they run entirely on Swiss public computing infrastructure, pre-existing supercomputers in Switzerland that are built for research and government use that are run on hydropower.
Speaker: They've also gone above and beyond, i believe, any other AI developer to make sure that they're not training on stolen data, but rather only on content that does not exclude use for AI training publishers who have chosen to make their content available.
Speaker: um I think that should be attractive in the marketplace. i I hope that a lot of companies want to use AI that's developed responsibly and sustainably in that way. And having a a public option developed for the public interest, I think is a important complement to regulation in shaping the behavior of commercial actors.
Speaker: So, Nathan, thinking about those recommendations you have for improving AI governance, what would it take to make that happen?
Speaker: ah The answer to that I think is both easy and very, very hard to achieve, which is political will. ah That's why I said earlier, I'm really glad that AI, which I think, you know, by virtue of its economic and societal impacts is clearly so important to our future, is now coming to center stage in the political debate and in the midterm elections.
Speaker: So many people have called for, and I would agree with them, a clear platform, a clear policy position from our major parties and from the candidates for the most contested races so that we that the public can see really the alternative visions of the future.
Speaker: And from my perspective, I hope that both parties would look to take significant action to resist the concentration of wealth. that AI is producing today. But regardless of the choices the individual parties make, I think voters need to be able to express that political will. I think that's the only way that we're going to see the kind of radical change that I think is necessary to make sure that AI is not a disruptive negative force on our future.
Speaker: And that speaks to the difficulty. and It's very hard in the United States for the people to express political will. it's It's hard for us to pass policies that the people want and the money doesn't want.
Speaker: So a lot of the the problems with ai are they're not unique to AI. we say this in our book, that AI exacerbates existing problems with democracy. doesn't really create new ones.
Speaker: But those problems are going to be the problems that we have to solve to regulate AI, which feels can can feel impossible. And now we're back to, you know, why young people are souring on democracy. It's kind of not working in so many cases.
Speaker: The public... option for a i makes me think about the issue of trust. Because in theory, at least, a publicly developed ai should be something that we trust. But this connects again back to the idea that we don't have a lot of trust in our institutions and even in our government institutions right now.
Speaker: And you will have talked a lot about how is In order for AI to be a positive for our democracy, we really have to think about trust and can we trust the a i And so I'm wondering what you think it would take to build an ai or to build AI into our democratic systems in ways that we can trust and is actually not just that we do trust, but is worthy of our trust.
Speaker: And it's important that you separate those two things. Because it's different whether a system is trusted but their system is trustworthy. Lots of systems that are trustworthy are not being trusted in today's political climate and and vice versa.
Speaker: When we look at trustworthy systems, especially complex ones, we think about transparency, oversight, and accountability.
Speaker: This is not unique to AI. This is anything. And if we have a system that is controlling our life, We want to know how it works. we want transparency. We want oversight. Someone is paying attention to how it works and making sure it works properly. And then accountability when it fails us.
Speaker: And, you I say that like it's easy, but it can be incredibly hard. AI is an opaque system. It's often hard to understand how it's working or what it's doing.
Speaker: You can watch the outputs, But, you know, that gives you only limited information. In our market economy, there's not a lot of oversight. know, we have exempted computer systems from normal product liability laws and and other laws. And we did that in the ninety s the 80s, because of the rapidly evolving technology. we didn't want to hamstring.
Speaker: Well, things are different now. These the technologies affect life and property. And you don't get accountability if the powerful can avoid accountability through the legal process, the regulatory process.
Speaker: So I'm giving a ah ah quick answer of how to do it. But when you get into the details, it turns to be really hard. I want to add one thing to that, which is I think part of how we build trust is by having clear incentives.
Speaker: And I think that's where the difference between a public option for AI and a commercial or even just a private option for AI can really differentiate. Obviously, we've seen lots of big tech companies that have been built under the explicit set of incentives of financial motivation, you know, of shareholder value, returning shareholder value.
Speaker: Google, Meta, Microsoft, the list goes on. And I think there are lots of good reasons why people don't trust the products coming from those companies. you know They've implemented strategies like surveillance capitalism, gathering data on people and exploiting that data through advertisement to generate profit. And I think that's a reason to question their trustworthiness.
Speaker: We've also tried a private nonprofit strategy for AI development. i mean, famously, OpenAI was founded as a nonprofit that was focused on responsible and safe AI development. And just a few years later, they've gone commercial and they have a trillion dollar IPO planned from their perspective, hopefully in the next few weeks.
Speaker: Anthropic was then spun out from OpenAI researchers who didn't like the direction OpenAI was going in. And now I think many people see um OpenAI Anthropic as following really parallel paths. And I think that's because in the private sphere, it's very difficult to resist that kind of a incentive structure of generating value, shareholder value and profit, um even if it starts out formulated as a nonprofit exercise.
Speaker: Government is very much not perfect. There are lots of very good reasons for people not to trust their governments. And I hope people will take democratic action, political action to try and fix those ah issues of trustworthiness where they occur.
Speaker: But think the thing that is fundamentally different for a publicly run project is its incentive. A publicly run project can be oriented towards producing public benefit. It does not need to be oriented towards producing profit and shareholder value the way that those private enterprises are.
Speaker: and what We know in a democracy is the best thing we can do is mutual, is is multiple power centers watching each other. Different branches of government, the press, citizen organizations, that that really is the way we get oversight and accountability.
Speaker: In thinking about oversight and accountability and transparency, i think one of the big challenges, and Bruce, you referenced this briefly, is that we don't really understand how AIs are working.
Speaker: Even the engineers who are working with them don't always understand why they're doing what they're doing, why we're getting the output we're doing, why they're going off in directions they're not supposed to go off in sometimes.
Speaker: And so is it possible to get a level of transparency and understanding with AI that we would need in order to have that accountability?
Speaker: but We have a lot of experience with opaque black boxes that we have no idea what they're thinking or how they reach the decisions. We call those people. And all the time. i mean, i i don't i don't even know how I reach a decision. i can give you reasons, but like we know they're largely justifications.
Speaker: And I have no idea what's going on in your head. And if we probably and knew, I think it was weird because we go off in all directions. So this isn't new. And largely in people, we judge them by their output.
Speaker: mean, I trust you because maybe we know each other. Maybe we're friends. We've been colleagues for years. We've been introduced by somebody we both trust. mean, all of these mechanisms that humans use, and none of them involve opening up your brain and seeing how it works.
Speaker: And brain scientists also disagree on how it works. So I think all those things get brought to bear. with these new intelligences, with ease with these AIs.
Speaker: So I'm not too worried about that. Do you think that the that we will come to a point where we're just interacting with AIs like we do other very confusing, opaque humans?
Speaker: Yeah, unfortunately, we're there sooner than we like. I was in a hotel last week and I called the concierge and it was an AI. And it didn't identify itself as an AI. I think that is something I want to see as a rule that it should say I'm an AI.
Speaker: But it was pretty clear was an AI. And honestly, it answered my question. So it did okay. I wasn't upset that it was an AI. I use term-out-of-directions on my phone all the time.
Speaker: And that's an ai Now, these are already being integrated into our lives, probably faster than we want, probably faster than is prudent.
Speaker: But the profit motive is to use these technologies where they can plausibly replace humans because humans are expensive and annoying and go on lunch breaks and and AIs don't.
Speaker: So without rules, I think we're going to see a lot of interaction with AI thrust upon us. Not necessarily consensual, but but they're there.
Speaker: And as these become more capable, and we can discuss when that happens and if it happens and and maybe it doesn't happen, they're going to be used in more and more places. No company owner actually wants to hire humans.
Speaker: They are necessary evil in order to get profit. If a CEO could fire all the humans replace them with AIs, they would do that tomorrow. And the fact that they can't is why we also have jobs.
Speaker: Will that change? Probably. you know, is it in the next five years, 10 years, 50 years? We have no idea. So given everything we've been talking about here, the profit motivation, the sort of way that the economics and governance of AI are happening, the speed at which this is being implemented,
Speaker: What would you each say is our our reasonable best and worst case scenario when it comes to yeah ai and democracy? what What are you afraid might happen, but what do you hope will happen?
Speaker: I'm happy to chime in with a hope. All right. my My hope is that we'll broaden the debate about AI beyond just the technology itself to think about how it interacts with those pre-existing humans and political systems that Bruce was talking about.
Speaker: And we'll recognize that so many of the problems that AI is creating in the world today are really just people exploiting problems in our existing system and that it's a system that needs urgent renovation. Yeah.
Speaker: You know, talking to you here from the US, I think about, for example, campaign finance, which is a system I think so many of us have felt is badly broken for decades. And today we see AI companies exploiting that by dumping tens, maybe hundreds of millions of dollars into our coming elections to make sure that they get what they want in terms of policies that will benefit their companies and their valuations.
Speaker: That's not a problem that AI created. It's not really a technology problem, but it's absolutely shaping how AI is going to influence our society. And so I think campaign finance reform is an example of a democratic system that I hope we'll recognize requires urgent renovation in order to steer AI's impacts on society in a positive direction.
Speaker: And we won't limit the conversation just to aspects of the technology itself. When I look at the positive and negative aspects, I think of it in terms of power. AI is fundamentally a power-enhancing technology. It makes people who use it more powerful.
Speaker: And they can be powerful for good and bad reasons. i They could want good or bad things. More democracy, less democracy. My hope is that AI is democratizing in power.
Speaker: That it empowers all of us more than it empowers the already powerful. My fear is it goes the other way, that AI will make the already powerful even more powerful and further increase the power differential between them and everybody else.
Speaker: So given that our listeners work largely in higher education, where do you see higher ed having a role here? How can colleges and universities help us realize the benefits do the good, democratize power while guarding against the those worst outcomes that you might be imagining.
Speaker: So I teach at both Harvard and the University of Toronto. I teach in the public policy school. So I'm teaching graduate students who are going into to policy, into politics, to work in an NGO, to work on you know these questions.
Speaker: And what I try to teach is how to think about the issues. and This is changing incredibly fast. Things that were political issues two years ago are less issues today and new ones are appearing and and in another year they'll they'll be different again.
Speaker: So agility is extraordinarily important today. so I try to teach that. I try to teach a facileness with the technologies.
Speaker: think we have a lot of bad outcomes when people making the rules don't understand the technology they're making the rules about. But really, I see the universities as a counterpoint to the corporations.
Speaker: The corporations are at leading the research. Most of the the papers about these technologies come from people who working at the corporations. And they have an agenda.
Speaker: And universities are the place where we're going to see more, a wider wider range of research. Research on things that might not be immediately profitable.
Speaker: research on things that might be unprofitable but important. So i universities, I want to be a backstop to a corporate takeover of the technology and of the future of this technology.
Speaker: You mentioned how fast everything is moving with AI, and that's what I see as being one of the biggest challenges that higher ed is facing in being that counterpoint to the corporations.
Speaker: our Our institutions don't move quickly. Our faculty... don't move quickly, and that's by design. um you work in in higher ed I don't know that you have the answer here, but how can our institutions actually move quickly enough to keep up with what's happening with AI, whether it be you know, how we're researching a i as that counterpoint, but even how we're teaching students and and how our faculty can
Speaker: keep up with what students know and are doing with AI? what How do we we take our overworked faculty and staff and our cash strapped institutions and actually do what we need to do?
Speaker: You know, if I could solve that, think I could solve a lot of things. And it's not just academia. How does a congressional office deal with AI? How does a policymaker deal with this?
Speaker: but They're moving incredibly fast. Remember a couple of years ago that prompt engineer was a job description? That lasted six months. And universities were putting together classes in teaching that.
Speaker: And then it disappeared. it This is going to require a lot of rethinking, not just how we teach, but how we pass laws. mean, do we have a hope of passing laws, of regulating at the speed of the tech, not with our current system?
Speaker: and Because lawmaking also takes a long time and is slow and has old, out-of-touch people in charge. I think we need to rethink a lot of our institutions for this age.
Speaker: Not sure if we can. but the moment requires it. So we've talked a lot about how hard all of this is, but we like to end speech matters by offering our listeners listenersors something tangible to reflect or hopefully act upon.
Speaker: So for students, faculty, or campus leaders who care about democracy, what is one thing they can do to push for positive uses of AI in politics, governance, elections?
Speaker: What can we actually do here? I would say make sure your voice is heard. If you're listening to this issue or think you're at AR more broadly, and if you have a passionate point of view on that, make sure that your elected officials, the policymakers in your community, know that's your perspective, that they're hearing from you, and don't take no for an answer to that.
Speaker: a lot of the work that I do is with a civic technology project called Maple, the Massachusetts Platform for Legislative Engagement. Our whole project is about making it easier for people to testify to our state legislature to have their voice heard in our state policymaking process.
Speaker: By the way, we see lots of valuable assistive uses for AI in building that tool. We use it, for example, to summarize legislation so that you don't have to read a hundred page legal document, a bill,
Speaker: to know what policy is being proposed, but you can have a much more accessible summary that's easier for you to understand and weigh in on. There are lots of ways that AI can help in that process, but most importantly, if you have a point of view on an issue, make sure that's heard. Participate in the democratic process. Talk to your elected official. Show up at a campaign event.
Speaker: I hope that everyone listening will feel free and responsible for doing that. Yeah, I agree. The way we change politics is get involved in politics. And if we don't get involved, then others decide for us.
Speaker: These are important issues. They're both national and local. right national I guess they're global, national and local. at at the At the large end, We think about the companies and their power and these models.
Speaker: At the local end, it's it's data centers and how our local power and what our resources are being used and what compensation we get. And if we want to change things, we need to get involved.
Speaker: there's no There's no other option. Well, I think that is the perfect message for listeners of podcasts that's fundamentally about free speech and civic engagement to to end on. So Thank you so much to Nathan Sanders and Bruce Schneier for joining us today.
Speaker: Thanks for having us. Thank you. That's a wrap. Thank you so much to Nathan Sanders and Bruce Schneier for joining us. Please tune in next month for a discussion about viewpoint diversity in higher education with Center Fellow Milad Mohabali and Heterodox Academy President John Tomasi.
Speaker: If you enjoy listening to Speech Matters, please subscribe on Apple Podcasts or Spotify, leave us a review, and forward this episode to a colleague who wants to learn more about AI democracy and higher education.
Speaker: Talk to you next time.
