Cspan, 45 years and counting. Powered by cable. Host this is neil chilson, director of a policy. He formally served as the chief technologist of the federal trade commission joining us here to talk about Artificial Intelligence and campaign 2024. Good morning. Lets start with the Abundance Institute trip to the abundance is that it is a new Mission Driven nonprofit with our explicit goal is to create the cultural and policy environment that allows emerging technologieslu including ai to germany, to bloom and to flourish. And so we think that is a major road to widespread human prosperity and we are excited to tackle the big problems that we think are in the United States right now. Host so much so the major director of policy specifically to that. How much come that reflects the concern they have. Guest and i we think huge potential to create widespread human abundance. This new set of tools, ai has been around for a long time. You and i talked about in the past. But this new set of generative ai creates a new set of tools that empowers individuals to create new and powerful types of medications with each other. We think its one of the big tools. As ai continues to be used throughout the economy we think its going to great a lot of opportunities for growth, Economic Growth and personal growth. Host before this job released in your resume you at the federal traden commission, position, chief technologist. What was that . How does it inform what to do now . Guest the federal trade commission is that the general consumer protection, and one of the competition agencies in thei United States. Chief technologist jobs to bring anderson of technology to the policy space. I spent a lot of time looking at reports but also cases and investigations that the ftc was doing to bringng that lens of Technical Expertise into the policy worked of the federalas trade commission was the. Host if over go to topic at hand, the london abun, how is it funded . Guest it has a wide range of foundation, individual and corporate dono. We are very news actually some of that is still being figured out but we Just Launched in april but thats sort of the space within the. Host does money come from the ai community . Guest on the back to 100 sure on all the details of what companies but i think that a lot of the ai communities would be interested in doing what were doing kernel when it comes to the topic of the election, joe institute put out first a series of reports forthcoming about the election. Whats the over all orn at least the things youre concerned with these reports . So almost as soon as chatgpt launched inn october 2022 there was discussion about hower this technology might be used to impact elections especially in u. S. And a lot of concern about the 2024 election. A lot of speculation it wasnt clear what is happening. There were reports of some incidents that came about and what we decide to do was launch an observatory, a tracker that would track all media mentions in u. S. Outlets of ai, ais ud in elections. We had been running that usingor the keyword tool that tracks all of that. We launched it in january 21 about here before inauguration and weve released our first periodic report last week, 180 days out from the election. Out of the 7500 articles that are track identified, we spotted three incidents of uses of ai in elections. Me some of those are interesting and we can talk more about what those specific incidents are. The goal here is to track all the way up after the election so that after the election with a very good ground truth, a database that says like what actually happened here a lot of the speculation is ai will fundamentally transform our election. We seen a lot of concern from academia, from politicians saying things like this is our last human election, or that we need to get this under control or we will not have a democracy anymore. E. We want to really get a good database of information that says what kind of impact this is actually happening right now. Host republican specific examples to what degree do you think that ai has the ability to undermine or sway an election . Guest theres a couple of different vectors that can happen. One of them that people were early on very worried about would be it would be used to hack election machines. It turns out that vector from all the people i talk to is very unlikely. The election system, the mechanics of the election system in the u. S. Are very robust. They are also very distributed. Its a kind of attack service thats very difficult to get around. Ai doesnt change about that much. Yet the way my change things would be around disinformation or misinformation. The question is how much does it increase misinformation or disinformation in elections . I dont think any watcher of news or observer of u. S. Politics is unaware that there is a lot of misinformation in Election Communications already. People are generally quite skeptical of information. On the margins it will increase the amount of speech because these tools to make it easier for people, individuals including individuals to create very sophisticated content, but theres no reason to think that speech will be out of balance, like more lies and truths. Ai has a lot of potential to help people who are in marginalized communities who want to speak to politicians, or politicians want to speak to them, to connect them for things like automated translation. Theres a lot of weight ai could help increase the robustness of democracy. There are some ways they could threaten it. Thats why were doing this research, two at the end of this election cycle will have really good set of data that says like what was the practical impact. Host this is neil chilson joining us of the Abundance Institute. If you want as questions concerning Artificial Intelligence and the upcoming election you can call the lines. 2027488000 if you live in the central time zone and easter. 2027488001 if you live in the mountain pacific time zones. You can also text as your thoughts at 2027488003. From the first part of the report one of the things that came back information you called because it was a robocall it took place before the New Hampshire primary. Set this call up for us. Guest this is a call that went out to about between 500025,000 people in New Hampshire. Its not quite clear how they were targeted it was democrats, republicans are some mix. This robocall was the voice of joe biden saying if you vote in todays primary, the republicans will win, orou Something Like that. And so the implication was that it was trying to suppress voter turnout. That robocall without, i think hundreds of thousands of people voted in that election and the secretary of state of New Hampshire since investigated this and said there was no evidence that it suppressedre votes. In fact, republican primary that year which was on the same day had record turnout. The democrats had a very average turnout for a year in which the democratic candidate is running unopposed. So it doesnt appear to have had a suppressive effect. In fact, in a weird twist it was actually created by democratic strategist. He says to promote the idea and the concern around ai. Its an interesting case. Host hears that at that took place just before the New Hampshire primary. This is a New Hampshire president ial preference primary. Republicans have tried to push nonpartisan democratic voters to participate in in the primar. What a bunch of malarkey. We know the value of voting democratic with our votes count. Its important that you save your vote for the november election. We will need your help in electing democrats up and down the ticket. Voting this tuesday only enables the republicans in their quest to reelect donald trump again. Your vote makes a difference in november, not this tuesday. If you would likee to be removed from future calls come please canal. Wo host thats the culprit is forced to call itself what did you think about the ability to reproduce the president that way . Guest this technology is amazing, right . I think that it is pretty incredible that we can use these techniques to take a small sample of somebodys voice and replicate it. And yeah, it can be used to lie, right . There are some concerns about that. I think the call itself as without it didnt have a huge impact, but it is something that voters should be aware of if theyre getting unusual calls from people that sound like a certain person, that they should Pay Attention and double check especially with your local Election Officials about whats going on. Host lets take a call from north carolina. Youre on with neil chilson of the Abundance Institute. Hes formally of the federal trade commission, and joining us today to talk about the 2024 election and when it comes to election matters. Go ahead. Are you there . I think we are having a little transmission problem but we will try that call, so if youre on, keep going, or if well get back in a second. When it comes to this idea you found out about the video being one part of those concerns of information, what other things did you find a came back from the search that youhe did . Guest two of the other incidents were an incident where there had been a deepfake created of taylor swift, right, that had some of her holding up a sign that said trump one, biden lost. I think there were two different entities. One was a video, one was a picture. It turns out it appears that those, especially the photo, couldve been done with traditional editing tools. Its not clear at all that was an ai created problem. That was one of the incidents the kalb report as potentially in ai problem. The other when we found in the time to look at was that there were trumpmp supporters who spreads these ai generated photos of donald trump embracing African American women and showing support from the africanamerican community. This and was a deepfake created by w ai. There was a callin show guy in florida who admitted he just those tools to make that. Those with the other ones within the window. We also talked about a significant one that happened before we started the tracker that was the use in a ron desantis add of some faked pictures of trump and fauci together. That has happened before that, did using a tool and though any campaign at. I think thats the only one we seen so far of use Ai Technology in such an explicitly in a campaign at. Host a highlight in the story from npr saying during the ad it was the pictures of where the president appears to be hugging dr. Fauci that has been created trade to i think thats right. Host amongst other pages as well that were real. Guest exactly. Whats interesting in the case is you would think as a be the sort of thing that would be very hard toot notice, but because of your highprofile candidates, obviously theres a lot of scrutiny on the ad they put out. It was identified relatively quickly that this hadhi fake images in it. The Desantis Campaign, run by a sort of calms branch of the Desantis Campaign. The Desantis Campaign is to morsel not sure that this strategy benefited them the way that they hoped itfi would. I get another incident in the tracker of the 7500 media reports we saw this is one of the 304 that keeps popping up, heres the photo, bbc publishing thehe photo of former president with africanamerican supporters. Supporters. Again deemed as false but the bbc pickingth that up is low. On thats one of the conversation we will have when it comes to 2024 matters and Artificial Intelligence. Letsig try nika again in north carolina. Good morning. A caller so im currently studying Artificial Intelligence program in college, and theres a major issues that i see with it i could affect our political landscape. Mostly its facts that our political leaders are uneducate on ai. They have no idea how it operates, how its created, the treaty, the data says used for and so on. Has to institute looked into trying to create Training Programs for our congress so that they understand the concept of this technology and little better when theyre making policies . I dont want another ai winter to happen as it is a get my degree tractor i i dont want tt for you either. The. Caller to ai winter is interesting. Ai has gone through several periods when theres tons of excitement and then sort of thy sovereign people, gets integrated into technology and nobody thinks of it as ai anymore. They think of its as computers. Your point is important. Getting regulation right in this case is critical and to do that you need to know, you dont you do everything about the technology but you need to know what i like to call the policy relevant characteristics of ai. By institute has created some educational materials. We have largely used him at the state level for state legislators where a ton of the legislative action is happening. Congress moves slower, has a lot of competing other interest, and so we are making those inroads in congress as well. They are also being talked to a lot by people from industry and from advocacy groups that all turned to bring comics up to speed on this stuff. Each with their own political angle, and so we do think its important to get principal educational materials that will help make the right decisions around how to regulate in the space, absolutely. Host eric is in california, antioch california. Hello, good morning, everyone. I was just wondering, looks like election will come down to like five or six states and now the states i was wondering could ai possibly tell you what individuals you have to concentrate on and be able to tell my algorithm how to influence of thosego individuals as to make them vote the way you want . Guest well, its a greatat question. Political strategists have for decades at this point been selling tools that can help identify what our potential voters in the state are persuadable. It turns out its really hard to change peoples minds. So tiny. How to do that, ai doesnt have a particular advantage on that. Most of the techniques we use in targeted political advertising are statistical techniques that you could call the Artificial Intelligence because they use computers and the use algorithms not what this new generative ai does not create a new set of tools for identifying people who are particularly persuadable. I i dont know that thats the g difference. Ai does create a way to customize messages, the generative ai customize messages may be to the issues that the politician and the targeted voter are most of the sort of intersection of what the most care about. Maybe there are some tools that help politician reach our candidates can reach people who they wouldnt otherwise be able to connect with. I dont think that the targeting tools have really changed all that much under this new wave of generative ai. Host how much background and technology you have to have to use ai sophisticated like the one we saw with biden . How much training do you find it . Guest not a ton. It turns out the person who created the Biden Robocall was actually a new orleans street performer that this democratic, democratic democrat strategist at had to do this. I dont think he had any formal technology training. A lot of these tools are sort of point and click and the really powerful and really interesting. I encourage listeners to try out some of these tools and become familiar with what they can do and what they cant do. That familiar with the tools will help people understand their limits, and also Pay Attention to what theyre hearing and maybe bring a skeptical eye to it with some educated background. Host do you see a future where some type of transparency is needed when you look at an ad and see if ai is being used in the ads that has to be identified in some way . Guest there is a lot of bills looking at that. In the political space thats, there are some bills that are actively looking at that. We have disclosures in political ads are pretty for various things. The challenge is that as i mentioned before once ai works we dont call ai anymore. Ur your cell phones phone si algorithms built into h it. So if you take a big you with that, does that count as an ai created at . I think theres some difficult definitional questions around what would count as ai. The safe advice would be to say if i am the lawyer for a candidate i would you say hey, put this disclosure in because we are not really sure. If you had that disclosure on every as it doesnt add anything, right . It doesnt add information for the voters. I think there are real tradeoffs and it crowds at some of thehe important messages candace want to get out to the public. I think there are real tradeoffs in having forced disclosures in that space. But it is certainly something that congress is looking. Host neil chilson for thisg conversation of the Abundance Institute. Let year from date in south carolina. Hello, dave, good morning, pedro. Listen, why is the basic deceit thats being practiced by ai purveyors, why is this deceit not prosecutable . I mean, this type of interference in our election and our control of our government is unheard of. I dont understand why does that have this areho allowed to practice what they do. Guest its a real great point, dave. Especially in a commercial context is already criminal. If youre lying about your product prices its criminal. In the political space, political speech is highly protected by the first amendment. And so missed truths or something, and is also very often quite subjective. However, there are laws both that the state and the federal level against lying about such things as when the election is aware of your polling place is. It doesnt matterg what tool yu use to do that. If you use pen and paper or if you use some sophisticated ai. If youre lying to people in order to suppress voters and suppress the turnout by lying about the process of the election, that is a crime and can be prosecuted. Host we have viewed off of exit asking, it seems if china or russia decides to fully pull the trigger ai the rest of the western will have no choice but to do the same. Do you agree with that . Guest im not sure if this in the context of an election. Unless there ably to influence elections. Lets take that. Guest if you take in the elections, again, its hard to know exactly how ai would be used. In many cases the goals of foreign adversaries in elections is not to get determine a specific outcome. Often it is to undermine the American Trust in the electoral process. For that you really dont need sophisticated techniques. In fact, the sketchier than techniques look, may be the better they are sort of, they trigger somebodies along and say hey, look like some is going on really sketchy in this election. I do think this new generative ai doesnt really have the ability to undermine our confidence in elections. If thats the goal of many of these thirdparty actors, i think and i really doesnt change it too much, is there a way to measure how much sophisticated ai is no sense of first made the scene . Guest lots of benchmarks these Companies Use to test their models against. These test, lots of different things. Many different capabilities. They are different for video or audio or text, which are the three main areas in which we have generative ai right now. What were seeing is this stuff is getting much, much better. Chatgpt Just Launched a new version earlier this week. It is very impressive a can do things like live translation between two people here has a lot of potential for helping people out in their daily lives. Its much more powerful than three, four, five years ago. Host christine illinois for our guests. Good morning. You are on. Caller good morning. Data for taking my call. I had a question any comment. The comment is regarding one of your statement earlier in the broadcast when he said that this too has the capability to increase speech. I was just one if you could comment on what he meant by that. The other question is with regards to regulation. We already know that technology typically things like regulation the following at a very slow pace nine actual speed of technology. So things like the example where this guy thats not very technical you know, did a great ad, what happens when the real professionals do it . You know, especially in the campaign space. I mean, what kinds of things that Interest Rate can go wrong . Guest great set of questions. I think those are quite related. When i was talking about increasing speech, what i meant is that it empowers people who dont have sophisticated studios four cameras or Editing Software to generate something that is persuasive and compelling, and professional looking. In that way it is a sort of level center. I do think what that means iss the people whose voices have been more marginalized or upstart candidates who are trying to take on a big incumbent, i think it levels the Playing Field a little bit between those. Thats what it by increased speech. If you look at how these tools are being used now, especially the image, just explosion of rich, interesting images i think is an example of how much these tools can increase the ability of someone to express themselves. I think that holds true in the political space as well. I expect to see a lot of content around politics were people using these tools to speak. I think that is sort of scary if youre sitting at s the top and youre saying i dont know about all these people having these two powerful tools, that might be a threat to those of us who are already sophisticated in our tools andf techniques. But i i think ultimately its a good thing for democracy because thats what we want. We want people to go up to and get theirlves ideas out there. As to your second part of the question, i cant i think this is more of a level center that it is a level of for the people who are we have sophisticated tools. Thats what we seem in the job space where we learned from generative ai is its helping people who are new to jobs or who are inexperienced very quickly move up the learning curve to become people are much more, putting out much more sophisticated work product in a much shorter time, even if they