This transcript was generated using AI automation and may contain minor formatting or transcription errors. Please refer to the original audio to verify specific quotes or context.
Disrupting Politics: A Front Row Seat to the Collision of Technology and Democracy
Book forum at the Cato Institute
Featuring:
- Katie Harbath, Founder and CEO, Anchor Change
- Adam Conner, Vice President, Technology Policy, Center for American Progress
- David Inserra, Fellow for Free Expression and Technology, Cato Institute
Moderator: Jennifer Huddleston, Senior Fellow, Cato Institute
Jennifer Huddleston: Hello and welcome to today’s book Forum on Disrupting Politics. A front Row seat to the Collision of Technology and Democracy. My name is Jennifer Huddleston, and I’m a senior fellow in technology policy here at the Cato Institute. And I’ll be the moderator for today’s discussion. As a reminder, for those of you tuning in live either here in Cato’s audience or online, we will be taking questions from both the online and in-person audience. If you’re part of our online audience, you may join the conversation and submit questions directly on the event web page, Facebook, YouTube and X. But to begin with, I’m excited to introduce today’s topic and today’s author.
Increasingly, we’ve seen conversations about the intersection of democracy and technology. New technologies have provided new opportunities for information access, for creative expression, and for ways to connect with each other. But in some cases, people have also expressed concerns about what this is doing to the political process, both here in the U.S. and around the world. Today’s author has had a first front row seat to this since the beginning of the social media age, and continues to talk about it as we see new ideas like AI continuing this discourse. Katie Harbath is the founder and CEO of Anchor Change, a consulting firm focused on the intersection of technology, politics and democracy, and the author of today’s book, Disrupting Politics A Front Row Seat to the Collision of Technology and Democracy.
She previously spent over a decade at Facebook, where she helped lead the company’s global election related work and helped navigate major controversies involving political advertising, election interference, misinformation and content moderation. Her tenure spanned the 2016 presidential election, the Cambridge Analytica scandal, and growing scrutiny of social media platforms influence over political discourse. Prior to joining Facebook, Harbath held digital strategy roles at the Republican National Committee, National Republican Senatorial Committee, and several political campaigns. Please join me in welcoming Katie Harbath. Excuse me.
Katie Harbath: Thank you, Jennifer. And thank you, Cato. And thank you, all of you, for joining here today to talk about this. I want to bring you back to the day after the 2016 election. I was operating off of two hours sleep, and it was a shock, right, to those of us in here in the United States and around the world that Donald Trump had won. And I was sitting on Constitution Avenue driving back to the Facebook offices, not knowing what future I was driving towards, because I had spent 15 years of my career helping politicians to use digital tools in which to better connect with voters. And for the longest time, that work was seen as a positive. And now we were grappling with and confronting head on the fact that our tools were being used by foreign adversaries, being the content rules were being pushed by the president.
President elect of the United States himself. And now the work that we were doing was in question. And that was one of the reasons that’s why I opened my book, Disrupting Politics, where I take you from the 2000 election to the tech CEOs at the inauguration in 2025, where I bring you inside the rooms of the early days of campaigning, where we were trying to there was we were introducing candidates to Facebook, to Twitter, to YouTube, to then those ten years at Facebook where, as Jennifer mentioned, we worked on global elections and then the five years after that. And one of the reasons I wrote this book is I really thought that we needed to look at this entire 25 year time span, because there’s a lot that happened in the early 2000 that you can connect to what is happening today.
The 2008 election, which was the first Facebook election where President Obama used social media. But that led to the Tea Party in 2010, which led to Obama in 2012 using data in digital in completely new ways. And then you see Donald Trump and others continuing to use these technologies. But let me first bring you back to 2004. Think back to that time frame. We don’t have iPhones. We’re using flip phones and blackberries. Again, social media is Myspace and Friendster. We don’t even have YouTube. But I learned there my first job was at the Republican National Committee, where I was a communications assistant, and there I was helping them to just have a website and send out emails.
But we also had the idea at the time the Democratic nominee was Senator John Kerry from Massachusetts, and we were trying to peg him as a flip flopper. And we had all sorts of research. We had op eds. We did all of those things, but it was having some interns dressed as dolphins, following Senator Kerry and his staff around, and me creating a website called Flipper Cam, where those dolphins would had they had a digital camcorder and they were filming themselves going around the country. They would ftp those videos to us overnight. That’s how long it would took. And then I would have to upload them to the web in high res, low res, and then Windows Media QuickTime in real time because there was no YouTube.
It was not as easy. However, those dolphins took off. It hit that, hit that message in a way that all the other op editors and stuff couldn’t, to the point where even people magazine did a feature on them. And that was my first introduction into how those types of memes could really take off online. But what I also didn’t realize at that time was the fact that we were chasing clicks. We wanted more. We wanted more people to see what we were putting out there, and we weren’t thinking about it. In hindsight, I could tell we were kind of the proverbial frog being boiled right where slowly but surely you kept moving to the extremes. And I saw this start to happen at Facebook in 2012.
We saw again Obama using the platform in very new ways. And after that election, we started to have a lot more world politicians around the world being like, how do we do what Obama did? And so we started building out the global team and started traveling the world and talking to candidates in Brazil and Colombia, in India about how they could use these tools. But then things started getting harder after that 2016 election. We were then facing having to build political ad transparency. And how do you define what a political ad is harder than what you might think? I was dealing with I was going in rooms where we were getting pressure from both the left and the right, oftentimes to do opposite things.
I know Cato’s in a lot of conferences and talks about that concept of jawboning, where the government puts pressure or other entities would put pressure on us to try to do either leave more content up, take more content down, and everything in between. And we were operating without a playbook. There was no playbook for how to build this stuff out back then. And I realized we were operating off of a lot of good intentions. But as I say in the book, good intentions are not safeguards. And scale is not a strategy. And I think about that a lot as we think about this AI era that we are going into. If you think about 2008, again, it was revolutionary for Obama to have a Facebook page.
And then again, this multiple times the 2012 election. But think about 2024 with AI not used a ton in the elections. But here in the midterms, we saw it not only as a policy issue, but also as the campaigns using it more and more for research, for communications, for building social media, and for making ads and for making content online. And I think in 2028, we are going to see campaigns use this technology in ways we can’t even imagine today. And so I want to thank Adam, Jennifer and David for joining me here on this panel today. I want to particularly thank Adam for hiring me at Facebook all those many years ago. And I want to leave you with my mantra, which is panic responsibly.
I came up with this mantra in 2023 when I was thinking about all the initial AI panic, and I’m from the state of Wisconsin, and we have a brand called Drink Wisconsin, which, if you know my state, we like our beer. But that’s where I got the idea. I was doing laundry, and I was thinking of this panic, and I was like, panic responsibly. That’s it. Because I wanted to acknowledge the panic, but also I was seeing people frozen because they were being told if they use AI, it’s cheating. But if they don’t learn how to use it, they’re going to lose their jobs. And that can be a paralyzing spot in which to be in. And so I want to us to acknowledge the panic, but move through it with intention and be willing to pivot when needed. If your decisions don’t have the outcomes that you meant to have. So I thank you all for joining us today and I look forward to the conversation. Thank you.
Jennifer Huddleston: Katie, I know we have a lot to discuss, and I’m excited to bring up and introduce our two discussants as well. First, I’m joined by my colleague David and Adam. David is a fellow for free expression here at the Cato Institute. His research focuses on a variety of free expression issues, both here in the US as well as around the world. This has included topics like youth online safety, the use of social media and AI, and elections, as well as how content moderation decisions are made. In an effort to help policymakers and the public understand how free expression can flourish online. We’re also joined today by Adam Conner. Adam is the vice president for technology policy at the Center for American Progress and CAP Action, where he leads the organization’s work on technology policy, including issues involving artificial intelligence, innovation and democracy.
As Katie alluded to in 2007, he established Facebook’s first Washington, D.C. office and subsequently spent seven years on the company’s privacy and public policy team, developing its government and political outreach efforts and directing its election related work. So Katie just gave us an excellent overview of her book and of some of the questions that she’s thinking about as things have even emerged in the just weeks since this came out. You know, a very interesting timing of the book coming out shortly before the election. And now we’ve had the midterm elections, and I’m sure there will be a lot to dive into. But Adam and David, you know, you both have had an opportunity to read the book. You also have kind of worked in this space and in your own capacities as well, starting with Adam. If you could kind of give me your response and your thoughts on Katie’s work and on the book.
Adam Conner: Yeah, absolutely. You know, one, I’m just immensely proud that Katie wrote this book. I know it’s something she’s been thinking about and talking about for a long time. And I think a real labor of love. And, you know, I think Katie tells three stories in here, and they’re really intermixed from her kind of perspective as somebody who was really there on the front lines for a lot of this in living history. You know, the first is just her own journey, right from Wisconsin to Washington, D.C., and then to Silicon Valley. And then the other is at the same time kind of coming up in this new media world, this internet world, this internet and politics world that we both kind of grew up in professionally and how it evolved, how she, you know, was at the RNC and the NRSC and other places really kind of being the early adopter, the evangelist, the kind of person who not only took something new and made it a proof of concept, but convinced important people, senators and others to kind of try these things and understand how they could be harnessed and the kind of process of politics and campaigns and democracy.
And then the third is really the kind of story of Facebook at key moments that she was there. And I think as a broader lesson from not only other platforms, but I think kind of the larger AI companies in particular that are going now, I think, you know, in reading that book, it is sometimes jarring to go from from piece to piece, right, to, to kind of remember that this is something that both she lived or she and I lived that had something that was both your job, but also kind of broader consequences. And I think that story is really important. One of the things I wrote about last night, when I was trying to kind of put my thoughts together was, you know, it’s really a book in the of a lot of other books that talk about what happens in the halls of power or in newsrooms or political campaigns, but what we really saw, I think, more than any other kind of technology from the private sector, was that these kind of platforms meant that you could really play a role in that from inside a corporation in a way that hadn’t necessarily been possible with broadcast news or or newspapers or other things like that.
And I think it really is, I think a story that Katie is one of the first to tell, and I think a really good documentarian of it, but also something that kind of we should expect more of. Right. Right now there are people, some of whom Katie trained, you know, inside companies like OpenAI or Anthropic or Google, who were kind of writing the next chapter as it comes to political engagement and politics and campaigns and other aspects of our democracy within these kind of frontier AI companies. And I think that this is, I think, an interesting roadmap. I hope some of them will read as they think about the challenges they’re going to be facing.
Jennifer Huddleston: David, your thoughts?
David Inserra: Yeah. So I have some notes because I want to make sure I’m quoting Katie’s excellent book accurately in some places, but the through line that I identified in the book is destruction of principles and pressures throughout the book. You mentioned jawboning, the idea that politicians, legislators are saying, we’re going to legislate this way unless you do what we want, we’re going to harm your business. Same thing with executive branch officials making public or private statements that are applying pressure with an explicit or implicit or else tied behind them, and that pressures that that can put on platforms and companies that want to allow speech according to their principles, but that is under pressure.
Similarly, you talk about external groups. There are lots of pressures from both the left and the right trying to advocate. They’re trying to play the refs. They’re trying to advocate for more speech content moderation that favors them and maybe harms their opponent. And maybe they think there are legitimate harms involved, but they’re often playing. The end of that doesn’t work. Then see point one about moving to their politicians and having their politicians try it. And so this brings up a quote that you say talk about key, which is that every decision was lose, lose, leave something up. You’re platforming hate. Take it down. You’re picking sides.
Change a policy while you’re overreaching. Hold the line. Your abuse. It was never enough. Mike Masnick, a tech writer, has this idea that he calls the impossibility theorem that it is impossible to do content moderation at scale. Well, because people will fundamentally always disagree with the way you’ve moderated more or less different. It just doesn’t matter. And so I think this brings us back to the idea of having principles again, you talk about after the 2024 election, how there seems to be a change within Facebook. They’re allowing more speech in certain categories. They’re rethinking how they do fact checking. I personally felt that some of those content moderation changes were a good thing.
I thought it was appropriate to rethink, for example, fact checking. Was it scalable? Was it did it have certain biases built into it? Was it was it accomplishing the goal that it was meant to accomplish, or was it actually being maybe counterproductive to building trust? Same thing with some policies that might have been short limiting speech in certain areas where we wanted society to have conversations. But the key thing is we don’t want it to be a political football. We don’t want it to be the situation where a new politician comes in and all of a sudden we just feel like we have to change our entire content, content moderation strategies.
That’s, that’s that’s a terrible idea. And you, you talk about how the pendulum keeps swinging because there’s no anchor. The platforms respond to whoever’s in office, whoever is loudest in the press or whoever seems safest. Its survival, but it’s not strategic, and it puts users, employees in society at the mercy of executive mood swings and political wins. And I think that’s exactly right, in the sense that we don’t want a society where sort of the speech rights, a platform, and also the speech that we as users are able to engage with and speak are sort of dependent on those whims and wins. And so that’s a key trend and point that you make throughout the book. And I really enjoyed that. Thank you.
Jennifer Huddleston: You know, I’d love to dive more into some of the stories in the book, but having these three experts with real world experience, I can’t resist, you know, having a broader conversation around some of what we’re continuing to see on this topic of technology and politics and their interactions, particularly in the free expression space. A book has to be finished before it’s published. It turns out, and, you know, I imagine one cannot just constantly submit chapter updates immediately after publication. But of course, within weeks of this, we have had the 2026 midterm elections. So to the panel, what are your thoughts on how we saw some of these same phenomena playing out in the 2026 midterms? What were some of the positive ways we saw technology being used as platforms for expression? What, if any, of the same concerns arose as we’ve seen in the past?
Katie Harbath: I will say having a newsletter and a Substack is a great place to keep things updated. So like I personally made the book like end with the inauguration because I was like, oh, should I add a chapter about 2025? There is so much that happened, but I’m like, I can always do an update, but newsletters where I get to do those updates and anchorchange.substack.com.
Adam Conner: Thank you, ma’am, my best hype man.
Katie Harbath: I will say so. I think on the positive side, as well as something we have to keep an eye on, is the fact that campaigns were able to use this technology more, you know, especially when you think about less well-funded campaigns that may not have the resources and the fact that you now have to create content for so many different platforms. AI can be a big help in being able to do that. Like I did a survey this summer to campaign professionals about how are you using AI? And it was kind of the boring stuff you’d have an intern to write. It was like research. It was take a speech like help us to clip it up, etc. and I think that can be a positive thing for people to be able to do on the more tougher side.
And I think this will become an even bigger political football as we go into the 2028 election, is when you go to an AI model and you put in a prompt about politics, how should it respond? There’s one hand, there’s the where, when and how to vote that they work directly with. There is a nonprofit called Democracy Works that I’m on the board of that helps to aggregate that information from election officials. They would work with the Associated Press for the projected results. But what gets harder is if somebody puts in a prompt, what does candidate X think about issue Y? And should I vote for them? Because when you do that right now, most of these models are going to go out to the web and try to find the most recent news and pull that back in.
But given that most mainstream websites are blocking access to their sites, understandably so. While we try to have those negotiations, that also leaves a big information vacuum at the moment where you have what are called these pink slime sites, where politically motivated groups on either side are developing what they look like news sites, but they are pushing their preferred narrative. Do you pull from a candidate’s website, etc.? Do you only answer about that one candidate, or do you put their opponent’s content there? To what if one candidate has a very detailed proposal about that issue, but the other candidates never talked about it? These are the types of even thornier questions that I think we’re going to be debating over the coming years.
David Inserra: Yeah, I would just note that I was going to mention some of those same positives that I think that in this past election, we’ve seen a lot of people turn to AI and use it in ways which I think we may think are fairly benign. I think those are often positive, though, because for especially for campaigns that don’t have the funds you mentioned, like, and I also think there’s an opportunity to sort of catch or try to catch that same sort of virality. You mentioned the flippers, the dolphins. It’s like you don’t have to like it, but you can solve what the candidate did out in the LA mayoral race. There were definitely some viral clips that were generated by a.
It wasn’t even by the campaign. It was by his supporters. And that’s powerful. That’s the ability of people who support their candidate, not even the campaigns themselves, to be able to contribute to politics in a way that previously would have required, like a film crew and a and someone to do mocking up all the essentially a cartoon. And now the AI can do it for you. And I think that’s an amazing step forward. But some people, we should also recognize that, yes, there are ways in which all the specifics you mentioned, there are some more challenging questions and I what I look forward to is platforms also handling those a little bit differently, the options for competition within the space being that like some platforms are going to they’re going to express their own preferences, their own values.
They’re going to interject what they think are good or bad, or whether that different candidates and there will be other platforms that take, I think, a more measured approach that try to just present the facts, ma’am, and that’s it. And they’re going to try to stay away from the thorny questions. And I think that’s maybe for the best, because there will mean people who want one thing will be able to find it, and people who want a different thing will be able to find that too. And so I sort of hope that marketplace continues to exist, and I think it will continue to exist, hopefully, as we see increasing usage, usage of AI like products in future elections.
Adam Conner: Yeah, I think there’s a couple of things converging in reading Katie’s book makes some of that kind of come back and give us some perspective. You know, I think one, every election cycle there is a rush to define an election. Is that technology’s election, right? The Facebook election, Katie has a point in 2015, in the book the Meerkat election, which most people don’t remember, what was the start up that did live streaming? That obviously is, I think, more prevalent as a technology now, but maybe not. The company that started it, you know, was the Snapchat election. Was it the TikTok election? Is it the AI election? So that’s a trend we see often.
And I think maybe even more kind of a challenge with AI, right? As a general purpose technology is it’s being used in a lot of different ways. I was just with a bunch of campaigns, talking about how they use AI on the campaign trail. Right. And a lot of it is what you see kind of broadly as it diffuses into the broad workforce. Right. What are we kind of automating? What is it making it easier? What is large amounts of data able to do things like that? You know, I think the kind of 2 or 3 things I pull out from this election that I think are worth kind of examining in that kind of context of where these new technologies and kind of intersected with kind of one, you know, the, the for all the fears of deepfakes and some of the concerns about AI, you know, I think the public has adopted a little bit.
I do think what is maybe a really interesting free speech question and then a very different kind of how society handles the question is the kind of deep prevalence we saw and people willing to make deepfake ads as paid advertisements of their opponents saying things they hadn’t said before, which might end up being constitutionally protected speech, but feels grosser, I think, than almost any other aspect of the use of it as we talk about it. And I think one of the more interesting things is, is a practitioner of political campaigns is, you know, they are in fact, degrading and destroying the kind of the attack ad right. The efficacy of the attack ad is no longer going to be as useful for a political opponent if everybody thinks it could just be fake, right?
So it is a tragedy of the commons in this scenario too, that is kind of separate from whether or not it is an interesting free speech question. I think the second point is, you know, some of the things that we fought over the last ten years or so on what Donald Trump would do online. You know, his obviously a lot of ink has been spilled on platforms decisions to remove him after the incidents at January 6th. But what is I think, very interesting about the last few years is his choice to kind of contain himself into social and a bubble that, you know, obviously leaks out often through screenshots of the things, but is no longer, I think, the kind of 1 to 1 influence he had on that kind of traditional media.
Sorry that all my media world that used to exist, I’d say from 2015, you know, maybe to 2024 or sorry to 2020 and or whatever he got onto. And then the last thing I’ll say is, I think we kind of over indexed in media, in politics and in policy and tech on these things still being social networks. And I think really, you know, at this point, they are really just kind of places you get short form vertical video. You know, you go to Facebook, you go to Instagram, you go to TikTok, you go to YouTube, you’re watching shorts, you’re watching reels. They’re all 32nd short form video clips that we really perfected. Matt Yglesias has a good post where he says the internet is kind of a miracle.
We just kind of use it these days only to watch short from video that has, I think, attended speech challenges and kind of moderation intelligence. But I actually think the biggest conversation point here is that, at the beginning of social networking, at the beginning of Facebook, we had much more content creation to consumption ratios, right? More people were posting photos or status updates or tagging their friends and things. And that is really in the even as easy as it is to make vertical video today, that is really shrunk to a handful of creators and mostly consumption, right? We have actually, in some ways just retreated to kind of a much less of an attention span broadcast model that’s more algorithmically controlled than it might have been during the broadcaster cable era.
But I think we kind of don’t always feel that because the one small sphere of the internet that is still relatively lightweight and interactive is this kind of text based messaging that is Twitter. And, you know, that is not, I think, the norm for these broader networks, for the society at large right now. And I think we’re going to be grappling with a lot of those kind of, candidly, new lessons, but also old lessons from television for a long time.
Jennifer Huddleston: So I kind of want to build out on something. All three of you mentioned AI, and it’s not a tech policy event in Washington, D.C. in 2026. If someone doesn’t mention AI and the potential disruption of AI. Katie, you talked about this in your introduction as well. At the same time, a lot of these debates seemed similar to debates we’ve been having for a while, whether it’s over social media or whether it’s over election ads or any other number of things. We’ve seen movements in both a state and federal level at times around questions of the use of AI in campaigns or in elections or election speech, or the use of AI as it relates to public officials or political officials.
Some of the concerns about the deepfakes that were mentioned. We also saw this in the social media era with questions of what makes an election ad, the Honest Ads Act being debated, and things like that. What are your take on, you know, is this just a continuation of the same debates that we’ve been having with social media, with this new technology tool, or is there a difference in the debate around AI and political discourse that needs to be considered differently, both by those of us that are users and free speech advocates, as well as when we’re looking at the potential policy ramifications.
Katie Harbath: Well, I think history rhymes, right. It doesn’t repeat. And so I think there’s, you know, for instance, a lot of debate about the algorithms and what people get recommended. What gets, you know, is a share worth more than a comment versus a, like all of those types of things. And those are decisions that these companies make. And I do wish that we had more transparency into that decision making process, not just the algorithm itself. And now those decisions are like I was mentioning on the AI front. How do you respond to these prompts? What does that look like? A big debate to write of. It’s a better user experience to be able to actually get all of your answers in one spot.
Right? But then what does that mean for the news industry and people not clicking out to that? And what are those new? What are those new economic models need to need to look like? And so it’s not exactly the same, but there are a lot of lessons that can be learned from the pains that we went through in the social media era and having some of these debates, I also think to and one of the things I talk a little bit about in the book, but is coming up more and more right now, particularly around child safety, is the role that the courts play, right? And the role that the courts play that not only challenging the state laws, but then also we saw, was it last year or no, it was earlier this year.
Where it was it 51 AGS, a deal with meta, where it was kind of like they both weren’t sure if they were going to win in court. And so they decided to settle, and do that. But meta, as part of that, is trying to get other companies to have to also comply with some of these things that they were that they were talking about. So you’re seeing you’re seeing some of that new stuff of how to use courts up against one another. And the last thing I’ll say to you is, it’s interesting watching the AI pose, because I think they might have learned a little bit from what Facebook went through after 2016. You hear Zuckerberg talk a lot about how sometimes he thinks that the company took too much responsibility after 2016, and we became a bit of a political dumping ground where it was just easier to blame Facebook for, for everything when in the reality.
Right. It was much more nuanced about why Donald Trump won in 2016. But you see a lot of the AI CEOs now, everyone kind of being like, no, no, no, we want government regulation. It’s not just us, and it’s a little bit of a game of hot potato, because none of them want to be the one sole ones that like, get everything sort of thrown at them.
David Inserra: Yeah. So when I was thinking about how we think about AI, and I come from my research on misinformation, where you mentioned like every, it seems like every next election always has something, something new about it. And it seems like for the I don’t know how many past elections we’ve been told that, yeah, this is the next election is the one where deepfakes, where misinformation is just going to somehow overcome our sort of democratic institutions, and we’re going to throw it into chaos. And I remember thinking was a couple of years ago, the World Economic Forum had a report that said that AI powered misinformation and disinformation was the most severe threat facing our world today, not wars or climate change or economic issues.
It was misinformation and disinformation powered by AI. And I like to think that, especially as we’ve gone through another election cycle, I think that was that’s wrong. I don’t think that the threat is as severe as is made out to be. And in fact, I think in some cases, our at times to restrict stop misinformation may have even been counterproductive. There may have been ways in which some of the distrust we see in our society was a result of some of these efforts to restrict or to limit misinformation. There’s lots of bad things out there that we could restrict and was probably right to correct them. But there are other situations where maybe we went too far and we got some things wrong, and now there’s a lot of distrust within our society for these knowledge producing institutions.
So maybe we had a counterproductive effect. But more importantly, and this is a quote from your book, you mentioned how the threat to democracy, the threat of democracy weren’t primarily technological, but rather they were human. And I think that’s what we see with a lot of the dialog around misinformation, both in talking about old fashioned misinformation, but as well as AI, we’re talking about something that fundamentally is playing to human vices and foibles. Right? Like people who are biased, they tend to believe that things that they are already biased towards. And so they might accept misinformation just because it happens to agree with their existing beliefs about the world.
Did it really change that misinformation, change their viewpoint? Did it radically shift their their viewpoint? Not really. Sort of. It’s come formation bias occurring. And we see this play out because humans are the real problem and the technology may change. But I think some of the same issues present themselves. I think there might be, as you point out, it might change a little bit differently from depending on the technology. But to come back to it, I think that’s why we should maybe panic a little bit less about the specific technology. Maybe panic responsibly would be would be called for here, because I think we could use a lot more responsibility in our panicking over AI.
Adam Conner: Yeah, I think I’ll echo most of the people said, you know, I think, the pressure on the large big tech platforms of the last ten years has been because they have been kind of a sole source that has aggregated through their success, right, where lots of eyeballs end up. Right. And so they are kind of you go to the facebook.com or you go to the Instagram app. And I think the kind of broad nature of the general purpose technology of AI, we tend to focus on the kind of narrow aspects of it, right? Our chatbot, it’s the best way to kind of 1 to 1 understand this experience, and that will be a part of AI. I think it’s kind of harder on some of these broader concepts of speech or others to think it’ll be a 1 to 1 on that.
You know, I think within that, though, you know, I think we do need to be cognizant, particularly with these newer, more advanced frontier models, right, that we don’t necessarily they are a little bit of a black box. We don’t necessarily understand why they output certain things or or the things. And so I think that on a, on a much broader level, that is a kind of fundamental question we have about alignment and control and other pieces. But I think in this context of information, you know, I think that is a wild card. We do need to differentiate, you know, not that the algorithm is something that Facebook, for instance, a news feed, does everything about, but it has an inkling of the inputs in it and some of the shifts, even if it doesn’t understand them all.
And that is not as true, I think, in some of these frontier models. And I think that should be something that we think about. I think, you know, look, fundamentally the difference here is what I say is, like most of us didn’t anticipate, for good or bad, how these kind of social network technologies would change broader swaths of society. Right? Not just business, but, you know, kind of our society, our elections and things like that. You know, I think it was not the strategy. And I do think most of the AI CEOs are being truthful when they say, we think these technologies could have pretty significant harms. We are fearful of them. And so, you know, I think treating that rhetoric with at least some, you know, curiosity to, to intervention, I think is, is just a different world than I think, where we fought many of the kind of policy battles related to, to, to social networks. So I think it’s not a 1 to 1. But I think, as Katie said, right. We we rhyme history, not necessarily repeat
Jennifer Huddleston: Needless to say, there are a lot of people who think they know how content moderation decisions get made. You know, Katie, you give some great examples in your book about what it’s like to actually be in the room. And I think a lot of people think these would be really easy decisions. Like if I was just the one in the room, I would get it right. But you really have shown how these are oftentimes real struggles internally at a company of trying to find those balances. And while we have nonfiction works of stories like like yours and all three of our panelists who have at some point worked on these content moderation decisions, we also have fictionalized works.
Most recently, the movie the The Social Reckoning came out based on a true story. It’s like it’s a mix, right? Right. And so we often, though, can see in our discourse this idea that people know what happened. They think because they’ve seen this kind of fictionalized account. What are your thoughts or your takeaways on how the public discourse, whether it’s based on the various nonfiction memoirs that have come out about these decisions or on the fictionalized versions of it, are impacting the discussions that we’re having around these issues, or how they reflect each of your experiences.
Katie Harbath: One of the reasons I wanted to write the book and share some of these stories is that, when you hear these things talked about publicly, usually it’s around like one piece of content, right? And it’s kind of like Facebook, you idiots. Why why did you take that down? Right? Like I wouldn’t have done that for that. Or why are you leaving that up like that makes a ton of sense. And you oftentimes just hear about what the like a Zuckerberg or Musk or what decision they made. And you don’t necessarily realize because the companies don’t oftentimes want to talk about that. The robust debate that happens sometimes inside sometimes the robust debate did not happen.
But like many times it did. And a lot of research and a lot of talking to think tanks and civil society and all of that to try to have an understanding. But oftentimes we were trying to write policies, and David and Adam can talk about this a lot more in depth to not for just one piece of content and not just for how it works here in the United States, but how it works around the globe. And that starts to get a lot harder when you start to think about these things at scale and oftentimes to where it’s like and you’re grappling with, you know, for instance, we don’t necessarily want Facebook to have the power to take down the president of the United States in the speech that they have, but they also have that First Amendment right in which to do so.
And so it’s kind of I’m like, I’m not an elected person. I don’t know if I want that responsibility to do that, but we don’t have any other better mechanism in which to have those have those conversations. And so I think I would like us, as we continue to have these debates in the, in the public discourse and all that, of bringing some of those, those nuances and stuff and kind of having those, those debates so that we can understand. And the last thing I’ll say to you, and one of the reasons I selfishly wanted to write the book was, I don’t want anybody else telling my story. And I think, you know, we can all experience the same event in very different ways and we can have different perspectives.
And so I think it is important for as many people as possible to share their perspectives and their stories for us to get a much more well-rounded understanding of sort of why a company may have made the decision that they did.
David Inserra: So I want to approach this question a little bit roundabout way by saying that I think we tend to think of companies as sort of like singular, rational actors, sort of simple, simplistic model. But I think we all, even in our own lives, know that our families and, and other organizations were a part of their messy just where you want to go to dinner or where you want to where you want to go on vacation. There’s lots of different interests and preferences involved, and it can get sometimes quite heated about something that should be very simple. There’s an axiom in sort of when talking about bureaucracies, it’s called Miles’s law, which says where you stand depends on where you sit.
So even if you have an organization that is like ideologically aligned and within any organization, there are people who have ideologies, politics, that, of course, plays a part. We’ve talked about jawboning and things of that nature as well. But just even if you’re aligned, you have a sort of viewpoint based on where you currently your job is at that company. So you talk about in the book, for instance, the civic integrity team, you say the Civic Integrity team, which was a team which was job, was to find bad content to make sure that the content online integrity teams are removing bad content, making sure that the content is good. And then, you often saw that the civic integrity team often saw the policy team, I think talking about the public policy team here as week two willing to bend to political pressure.
But then you offer a bit of a defense of the public policy public policy team saying they are dealing with real world ramifications of these decisions. And you come away by saying, this isn’t an argument for always choosing business over principle is an argument for establishing your values before the crisis hits. So again, this idea of principle, which is a theme throughout the book, but people within an organization have different roles. You have you want people who are out there finding bad content and taking it down and finding adversarial behavior and stopping them. You also want people who are there saying, I think we should value expression.
We need to think about the ways in which if we take this content down, we’re also going to be taking down this content. And is that a harm to society or are the harms so great that we think that it’s worth taking down the speech? There’s other people who are looking at what’s going on in. What do people in other parts of our country or other parts of the world? Think about this. So when you have to bring it back to the social reckoning, when you have whistleblowers, when you have convenient leaks to Congress where that shows how bad the tech companies are, I would just posit that we take that with a grain of salt, because oftentimes that’s coming from one perspective.
It doesn’t show the perspective of a team like mine, which was trying to say, let’s try to do a balancing out. Let’s try to think about these other important elements of speech. You’re only hearing about the harms that supposedly companies are engaging in, and maybe there are some harms there that may be the case, but we should take it with a grain of salt, because if we just buy into those narratives, I think we’re shortchanging the overall debate that’s happening within companies and really defaulting to a very cheap, simplistic narrative. And one other thing I might notice that when you have these kind of leaks or situations and that these the bad research, the research inside the company where the company says, well, we know there might be some harm associated with this activity, I worry that leaking that kind of content so that can just be used against them in the court of public opinion or in some cases, the court of law.
That makes companies less likely to actually do that research. You want people doing red teaming saying, what are the potential downsides of this, this technology, of this policy we’re thinking about implementing? But if that research is going to be leaked out of context and all the other information, all the other perspectives are not leaked, then it’s just why do the research at all? It’s probably worse for society that we don’t have companies thinking about, like how their products could affect people, but if it’s a source of liability, well, why would they do that? So I actually think it may make things worse by sort of saying, how dare you know about this bad stuff and then not do anything about it? I think we ought to be awfully careful, because it could actually make companies want to be less transparent and do less research into the potential outcomes or impacts of their products.
Adam Conner: I’ll be a little contrarian in the sense that I think, you know, Katie talked about in the book and she’s correct, right. Impossible trade off, lots of debate. And you kind of run into this kind of other opinion here, right? Which is, you know, we’re working, you know, outsiders are working, the refs their jawboning. Look, there is definitely real jawboning that occurs. I think the canonical example would be after Twitter labeled when a President Trump’s tweets in 2020 and he threatened to repeal section 230. And then people forget this, vetoed the NDAA because it did not include a repeal of section 230 that had to be overridden. The first, I believe, override of an NDAA we’ve ever had, a veto it ever had, like that is a level of very coercive jawboning.
You know, I think some of the things that were elevated in the Supreme Court case that the AG’s later lost, The Economist and the canonical example of jawboning, I think, really speaks to the, you know, not to defend all the actions there, but like, you know, these aren’t necessarily clear or understandable rules. You know, I dealt with a lot of politicians when I worked for Facebook, would email me asking about a piece of content because they didn’t understand it. When I had to explain, here’s why I could stay up or come down. They had to ask questions because it’s not obvious to the people who don’t work in it. And I think we have fashioned an outside regime to regime in these companies that kind of is supposed to look like the law, but the law is also meant to be questioned and explained.
And, you know, I don’t necessarily think it’s jawboning to try and understand what was clearly, in many cases, an internal deliberation to come to that to, you know, I think too often what gets called jawboning is like asking questions when it was very clearly like a messy internal discussion, too. Right? So I tend to separate those into kind of, you know, that kind of peace and unfortunately, right. We don’t have the kind of external independence that like a judiciary branch or others would have, which is something they tried to do with the Oversight Board at Facebook. But, you know, I just think that ends up being kind of a complicated thing.
And so when we say yes, absolutely, people are trying to work the refs, but that’s also kind of how a legal system is kind of set up a little bit. And so, you know, there is some of that natural give and take that within the kind of person that these entities and other pieces. So, you know, I do think that the, the illustration that the messiness clearly happens inside often translate into the belief that then that’s kind of a universally understood declaration should be questioned, you know, from these companies, because they don’t really like being questioned. And the answer is like, of course it should be questioned because it’s something that’s not particularly easy to understand.
And so I do think that is thing, you know, that has I think, again, I think lowered in salience over the last year, but because of the step or two, because the steps many of the companies took to I think maybe we argue, appease the new administration. But also, like I said, so much of this content becoming, you know, kind of short form video and short form video in particular from maybe already elected officials, others have kind of decided to exempt from some of these rules generally.
Katie Harbath: Well, and Adam, you made a really good point in your Substack today. Thank you, by the way, for that. Where you were like, and I talk about this in the book, too, like the very important role that civil society in think tanks and folks on the outside do play. And you made a really good point about sometimes it was easy to dismiss their recommendations or their concerns of being like, oh, you don’t understand tactically how this works like this. That’s there’s no way how how would we even do that? How would we implement that, etc.. And you make the very, very good point that there’s a lot we probably could have done of at least trying to find that kernel of, okay, what is that thing they’re concerned about?
And then how do we help apply that internally? Because we understand how these things work. And one of the things I’ve been trying to do since leaving Facebook, and I think all of us kind of have, is like help to people to better understand how these things might work so you can improve those recommendations. And I really wish that we didn’t do it good enough job inside the company of after we would talk to folks on the outside, we take it inside and then we would never circle back with them. I feel like or sometimes we try to at least be like, here’s where we took your recommendations in. Here’s what we were debating, here’s where we landed in.
Why. And I do think that even even if they would have disagreed with our decision, being able to at least help walk them through how that decision got made, I think people would have at least appreciated that ability to do so. And then again, continue to have the debate as needed.
Adam Conner: I think there’s a kind of a related piece to this, which is right. You you want to be technically correct as much as possible, right? You want to. But we also don’t necessarily all have the knowledge, you know, to, to get it to. So there’s some level of what assistance should we have been giving them to help them understand, to kind of get to better solutions? You know, I see this a lot, particularly right now in the frontier AI discussion. Right? I was just meeting with somebody at an AI company pretty high up and say, we’ve got to get everybody educated so we can have the discussion. And, you know, it’s like look like. So they understand these, you know, the frontier models look like the like actual people who truly understand this, but is actually like a vanishingly small percentage.
Right. And but you’re unleashing them onto the world and they’re going to have what we what you argue are kind of global consequences. And so in many ways, this idea that you can only way in if you have a full understanding is another way of just kind of retaining power among kind of a small handful. And, you know, maybe it was something that was reasonable or debatable in the kind of kind of slower. Now I want to say, you know, for a company that has billions of people over the course of years, but much slower growth at Facebook than OpenAI or anthropic has had, you know, society is not going to wait for everybody to get educated before they engage in it.
And I think to the extent that AI companies have to reckon with that, any kind of singular thing, it’s that they have to go into this discussion that they have chosen to kind of unleash these things onto the world, understanding that like, no, not everyone’s going to understand that’s a consequence of you unleashed in this world not only being a couple hundred people, a couple thousand people who can’t understand it. And we’re going to have to do policymaking in a kind of messier sense. They’re kind of given the kind of impact and stakes.
Jennifer Huddleston: I want to turn now to audience questions and answers. I’m going to start with a online question while we get our microphone runners in the room. For those of you that are in the room, if I call on you, please do speak clearly and directly into the microphone. And also please announce your name and any affiliations. But to start with, I’m going to go to one of our online questions and I’m going to paraphrase a little bit. But the question basically is around disclosure and questions related to that, particularly in the political context. But this has also come up more generally in kind of AI generated or AI manipulated, depending on the point of view of the individual and the piece of content.
When it comes to what, if any, kinds of disclosure or disclaimer rules, either by companies or by the government could prevent concerns about disinformation? And how might that play out, especially in the election and political context?
Katie Harbath: Well, we are I mean, so there’s a couple of things, I think when you think about disclosure, I’ll. You know, on AI right now in deepfakes, I think it’s, what, 31 states that have laws that require saying that AI was used used in the ad. A lot of states did not. I think Texas famously does not have that. And we saw that used quite a bit. And so there is a lot of debate around forward facing disclosure. But then also, what sort of data or ID can you have to be able to tell if something was used by AI or not? That we continue to see these companies continue to evaluate and look at. But another challenge to is that it can feel it. I really wish the solution was just label it right.
Let’s just tell somebody that this is the thing. The challenge is with that, right is like, if you’re on your phone, you could have all just labels and no content whatsoever. The other thing too, that researchers has found is that sometimes having that label, people don’t see it. They start to ignore it, or sometimes it has the opposite effect where it makes them dig their heels in even more. And so I think this is an area that is very ripe for more research around what is the appropriate type of disclosures, not only for the end voter or citizen that might be consuming it, but also for those that are trying to study what’s happening?
David Inserra: I think that’s exactly right. And it’s why I like when different companies try different degrees of disclosure and or sort of warnings ahead of time, because I think that’s where we don’t know exactly what. I’ve seen similar research that says we just sort of get the fatigue from seeing so many labels that like, does it actually impact your decision? At the same time, more information to consumers is likely a good thing. Like I would like to probably know if if you if you can easily tell me that something is clearly false. And that is helpful to me as a consumer of information. And so it’s finding that balance of how do we tell people pertinent, useful information without losing their attention in such a way that, yeah, it even it is even counterproductive, because if I just use a filter to touch something up, I have now, I guess used AI depending on your definition.
And now that’s like, well, all I did was apply a sepia filter or whatever it is. I just tried to make myself look nicer. Like, should that be labeled the same as someone who did a complete deepfake from the ground up? And then there’s all sorts of other questions about what if it’s, you know, it’s not real video, but it’s real speech. I think was was this Schumer I think there was content involving taking some content from I think it was Chuck Schumer, Senator Schumer. And it took were written words and they put it to they made an AI generated the content, but it was as written words. But it wasn’t his spoken words. I think it’s what it was. Okay.
So that’s clearly I did. But it’s not false in the sense that like it is accurately what he said. It just he didn’t get in front of a camera and say it. So how you do the labeling I think is something that companies are going to deal with. And that’s why I would prefer that probably be handled more at the company level. Again, because I think we if we sort of set a one size fits all policy, we risk getting that wrong and then we can’t easily switch that because the law has been set. Whereas with companies we can say, I like how X does it, I don’t like how threads does it or vice versa. But I do think that is something that we’re grappling with is how do we do disclosure and labels. And I don’t think there’s an easy, obvious answer.
Adam Conner: Yeah, I think people often correctly and understandably jump to political speech, which is in this country, obviously among the most protected forms, maybe the most difficult to think about. And so to me, it’s always helpful to think about where the kind of commercial speech analogs first as a principle. Right. Like do we think disclosure if you are paid, you know, being paid to promote something, right. Is that a reasonable thing that you as a consumer would like to know? Right. Is your influencer actually being paid or not? I think is a useful piece of information. Right? Even if you might not always pay them to label right similarly, right.
If you’re buying a product and it looks nothing like the thing you’re actually purchasing, right, is that, you know, consumer fraud is that is unfair and practices you. So I do think there’s some standards relative to that, that, you know, are still being developed in the commercial world, but like, might make some analog sense to go into the kind of political speech realm if we think that they make I think particularly paid disclosure feels like a space where that is a kind of campaign speech. Fights are kind of a always a big kind of issue there. So I think that ends up being, you know, kind of pieces within it that we may want to start with, you know, and as I think I mentioned at the beginning, right, regardless of whether it’s legal or not, you get, I think, a kind of factor from like truly deepfake, not even the kind of the example of a senator, but like the fake of your opponent paid for by the, you know, and so, you know, I think what is also interesting there is right, like that is just a level are we going to see diminishing returns writ large on that? Right. And, you know, it would be ironic if the thing that destroyed the attack ads, right, was, was deep takes in terms of like, efficacy. But, you know, I think that remains such a tricky job.
Jennifer Huddleston: All right. Well, go to the audience here. So.
Karl Polzer: Yeah. Thank you. Great discussion I’m Karl Polzer. I operate the center on capital and social equity. Thanks for letting me advertise it. And I analyze economic inequality. And I have for the bottom 50%, which I felt nobody did in Washington. My my question is a systemic one. So what something that kind of concerns me about my I use AI kind of as a research assistant, but I notice lots of stuff for free. Then they start charging my my concern I guess I like your views. What are the what’s the risk of AI carving up the internet or fundamentally changing this open resource? At the beginning it was an open resource, like a free range into a carved up, privatized space dominated by a few monopolists that can direct because they know a lot about the audience.
And the power of the robotics is very, you know, to manipulate is very strong. So what is that the risk and, you know, that would restrict freedom of speech and just systemically. Does that make sense?
David Inserra: Yeah. Yeah I’m happy to start with this one. I think at a high level, there are always going to be products that have different degrees of different degrees of the product. Right. Do you want the best product that’s out there? I think it’s likely that you have to pay a significant amount of money for that. But there are a fair number of of free AI. I probably this also gets into, I think, the debate around open source models and how the government, some different governments want to clamp down on those. And I actually think that’s why it remains actually very important for those to be to continue to exist. Because while there are concerns around how open source models could be used for, for various various bad purposes.
It also still gives the average person away. That’s not within, you know, the within one large entity or another large entity. And I think we so in general, I think we want competition and the competition, the solution to making sure that it’s not like just a product that just the rich can use is some that is first, first of all, in general, I think it’s democratizing, like the amount of information you can get even from a free model is still mind bogglingly better than what I could have when I was, you know, high school, in high school, trying to do my research products and using the Encyclopedia Britannica or whatever it is. But then I also think that there’s an opportunity here for a different companies to come along and open source models to potentially compete with those, maybe not the same level as the frontier models and the most powerful corporate models, but I do think it’s a powerful option that remains gives people that free option or an option that is free from government interference, or from if you don’t like one corporation or another.
And I think that should be considered in the debate. When we think about open source, which I don’t think I think we think about the risks a lot with open source, but we don’t think about the impact that will have on people who don’t want to use the large models.
Adam Conner: I think to your question, undeniably, AI is changing the way that the open web has worked for the last 20 years, and it is up and in the financial underpinnings of what has traditionally been this kind of. So it’s not about necessarily model capacity or capability talks about. Right. It’s about where is it. Able to draw. So this information. Right. And so this kind of vast network of websites that has kind of survived on the kind of contributions of people putting things out in the internet for free of search engines, you know, kind of directing people to it, but not taking it, you know, the kind of rise of AI reviews, the rise of kind of being pulled into chatbots as a search.
And really, you know, I think even more importantly, right, the costs that are associated are skyrocketing with hosting that information. Right? I believe there was a stat the other day, right, that now there’s as much bot traffic as human traffic on the internet, right. Because we’re all opening up, Claude, and launching out ten agents to go open web browsers to kind of do that. And so you get these websites that might have been passion projects, whether they read it or wikis or other things that are now kind of both providing free information, getting no traffic from it, which has traditionally been their source of monetization, and paying higher hosting bills.
Right, because now you’re getting extra traffic from bots. And so it is fundamentally shifting. But you’re seeing already, right, these information words are sledges, as I think Katie mentioned. And that has written about a lot trying to fill the void because correctly, some people are blocking the ability for AI agents to hammer them without modernization or even just to make their hosting bills not go through the roof. And it is bifurcating and fragmenting the web. And unfortunately, like even in a world where you say there’s no more, you know, paywalls or robots that txt or any of those things, right? That doesn’t mean the economics of work to keep these sites over.
So it is, I think, really you highlight a fundamental challenge. It’s not a clear answer, but it is a very real reality that they’re being squeezed now on a couple of different levels and what had traditionally been the source of the open web. And it’s not clear where it’s going to go from here.
Katie Harbath: And I would just add one other point there to where also thinking about how this impacts the offline world, like, I think a lot about small businesses that are going to start to potentially get inundated with people using their agents to either like, call them or make orders or stuff like that. And those people aren’t going to necessarily know and be be up to speed yet, and how they can use AI to talk to the AI agent. And so I worry about that too. And will that put people out of business that we weren’t these unintended consequences by having this technology out there so fast and not having others, even government services, etc., be ready for that, kind of that kind of traffic.
Jennifer Huddleston: So we’re running short on time, and I want to make sure that each of our panelists gets an opportunity for kind of a closing thought, and also to make sure that our audience, both here and online, knows where they can find out more about the work you’re all doing on this issue. So for my last question, for each of you, if you could have the public or policymakers understand one thing about the intersection of technology and politics, what would it be? And, Adam, I’m going to start on that in and work my way in.
Adam Conner: I think it would be that at least is my understanding from what they’ve said, both publicly and privately, how many of the kind of leading CEOs of frontier right near. I do not have a very clear understanding of where there either today or in the very near future, their model capabilities and their safety abilities to control or align are the kind of fundamental theory right now on AI models is the goes up and the kind of safeguards are supposed to stay in front of it. And it is very clear from everything we see that the capabilities are growing even faster. And if we have not crossed the line in which they feel comfortable and being able to control it, we are very close.
And if they think that with the information they have, which is much less than we have, but we are kind of hearing like what are the broader implications for everybody, policymakers to the public in that? Because I think so much of our debate is stalled on where on that line are we? And I think the future of that is understanding that as best we can and adapt and responding to it.
David Inserra: Just briefly, I think I would just encourage people to avoid or step back from the moral panic. There’s a panic responsibly, I think, would be what I would encourage policymakers and people to think about is, yes, there are transformational new technologies. Yes, it is changing the way we gather information. We speak business and a whole host of other things. But I worry that in the rush, in the panic, we’re forgetting the potential life saving positive and increases in speech that are available through new technologies, whether it be the older technology now, social media or the newer ones of AI. I think we should we should make sure that we are not losing the benefits, for we should not forget the benefits in light of what concerns that we may have.
And I think there’s a lot of positive that can be done. But even if you see it in something as small as like a data center fight, and that impacts the way that like we’re talking about speech online, what AI models can produce around these tricky questions. So I would just encourage people to think about the positive things, about the way that these technologies can be used for good, not just for ill.
Katie Harbath: And I think I would add on to that, you know, the trust deficit is big right now. Trust towards tech, trust towards our government, trust between those, those entities and the voters and everything like that. And I do, but I do think that if we were going to find good solutions to this, we need to be able to come to the table and talk about these really thorny, these really thorny issues and what this looked like. Like, we can talk all we want about wanting regulation around AI safety, but what does that mean? And those are the conversations we need to be having to. And it’s hard to do that when you are having a lot of this panic. And the last thing I will say to you, I dedicated the book to the people in the arena.
If people remember the Teddy Roosevelt speech where he talks about it’s not the critic accounts necessarily, but it’s the person that’s in the arena doing the hard work every single day. And I want people to know it can be sometimes easy to forget that there are a lot of people inside, not only the companies, but think tanks like Cato, other places, academia that are really doing that are doing this work. It still exists. It is still happening, even if it is not being talked about as much publicly by that. And I think that it is helpful for us to remember those people, and not just those at the very top of these companies.
Jennifer Huddleston: Well, thank you so much for what was a wonderful conversation today. Please join me in thanking our panel.
