The following transcript was generated using automated transcription software for the accessibility and convenience of our audience. While we strive for accuracy, the automated process may introduce errors, omissions, or misinterpretations. This transcript is intended as a helpful companion to the original audio and should not be considered a verbatim record. For the most accurate representation, please refer to the audio recording.
Michael Dunne: I’m Michael Dunne. I use an AI program to put the transcripts for this show together, and it works pretty well. But would I use AI to diagnose a health problem or drive my car? No, I wouldn’t. Maybe I should be more open to AI, but like many people, I fear the outcome, and that, in many ways, is the crux of the AI problem. We know it can do many, many things, but we’re fearful that it can do them too well or do them without human constraint. Today on the show, you’ll meet a political and business expert who studies AI and feels a lot like we do, but believes we in the U.S. have to continue to develop AI, or else we will lose control, not just to China, but to the AI that’s already out there doing what we asked it to do. Dmitri Mehlhorn: Democratic strategist, founder of the Atoll Society and executive in information businesses. Dmitri, thanks so much for coming on and talking with us.
Dmitri Mehlhorn: Thank you, Michael, for having me. I’m looking forward to it.
Dunne: So, I’m here in Oregon, and right now there is a severe backlash going on in a state that has courted data centers. From what I’ve heard talking to people, one of the reasons data centers are suddenly not in favor is that there’s a lot of fear around what data centers power, and that’s AI. I wanted to get your read on the situation. From what you’ve observed, is the fear about AI among average citizens justified?
Mehlhorn: Yes, absolutely, it’s justified. Artificial intelligence is very powerful. It’s very dangerous, and it’s moving very quickly, and there’s not a lot that people can do about it. So people are rightly worried that AI will be used for evil and for bad things rather than for good things, and that’s a political question that people are right to be concerned about.
Dunne: You’ve immersed yourself in this topic. What are some of the key aspects of AI that are troubling you, both in terms of the potential for harm and in terms of the place of the West and the United States in this race to either dominate or contain AI?
Mehlhorn: Yeah, in your question, there are at least two, maybe three, embedded questions. So let me try to take them in turn.
Dunne: Sure.
Mehlhorn: On the question of what AI is and why people might be scared of it, the analogy that I think best captures it is the invention of fire, the human capturing of fire. If you think back to our long-ago ancestors who were figuring out how to use fire for the first time, and you could talk to them and ask them what it’s about, you would learn they don’t really understand how it works. They do know that it’s powerful. They know that it is powerful enough to do really good things, like cook food, make tools and keep people warm. They also know that it can be used as a weapon of war and to create terrible pain and suffering for other humans. And once you discover fire, you cannot undiscover it. The only thing you can do is try to make sure it’s used for good rather than for evil, and that’s why people are right to be scared. It is an incredibly powerful new technology that is going to change us and change our societies, and it could do so in a number of different directions. That’s a good source of fear.
Dunne: Yeah, it seems like a couple of years ago, the main fear about AI was job replacement. And I know you talk about this with a lot of AI insiders who are coming forward in the public domain saying that it could end civilization. Is that the transition of fear with AI, that it’s gone beyond ‘Hey, it could replace me’ to ‘Hey, it could end civilization’?
Mehlhorn: Yes, that is a fear, but I don’t think that’s what most people are concerned about.
Dunne: OK.
Mehlhorn: Honestly, it is true that people who think at a high level about what could happen from the technology are concerned about what it means at an existential level. But you also saw some of that with the development of nuclear weapons and a lot of other technologies over time. I think people are more concerned with the tangible, day-to-day uncertainty of what this new technology means for them. Will it take my job? Will it require data centers that are polluting and hot and consume a lot of energy? Will it change my future in a way that I can’t predict, for myself or my children? I think those are the more tangible fears. And for what it’s worth, as for the predictions that AI will wipe out civilization, there are a number of folks, including me, who believe that argument is just absurd, because it rests on the idea that artificial intelligence will become this new superintelligence, a general intelligence that will become self-recursive and smarter than God. That argument, even though it’s made in good faith by a lot of the people who develop AI, really doesn’t make a lot of sense when you think about the energy constraints and the computing constraints on artificial intelligence. It is far more likely that artificial intelligence will be a tool that is used by different humans in different ways. It’s a little bit like thinking back to when nuclear weapons were developed. If nuclear weapons had been developed by the Nazis in Germany and by Imperial Japan, the path that technology took would have been much worse for humanity. So the most important thing, the thing people are worried about on the front line, is the role of artificial intelligence and data centers in their communities, and that is a very important set of concerns that people and politicians should address. But there’s also a broader worry about how artificial intelligence will shape civilization, the challenges and collapses of civilizations, and the fights between large civilizations like China and the United States. Diplomats and politicians need to pay attention at that high level, too.
Dunne: Well, let’s go there. The two main players on the global stage, I think, are the United States and China, and you talk about the fact that we can’t fall behind China in the development and implementation of AI. Talk about that, this need for the United States to be the leader.
Mehlhorn: Yeah. We used the analogy of fire, and the development of fire is comparable to the development of artificial intelligence today. If fire exists in the world, it is not a good idea to say we’re not going to have fire, because that means everybody else will have a tool that we do not. Right now, the major developers of the frontier models are overwhelmingly China and the United States. Everybody else is far, far behind, and artificial intelligence is improving very rapidly. The models that are available today are much more powerful than the models that were available even a few months ago. If we just put the brakes on all development in the United States, it is very unlikely that China will slow down. China’s track record on international obligations is not one of joining them and being a good actor. So you have to wonder: In a world where this is very powerful, do you want Xi Jinping and the authoritarian Chinese state to have a monopoly on this super powerful weapon, or would you prefer to have this tool be more broadly distributed, with at least some of it in societies like the United States that are still more open?
Dunne: OK, can we make AI powerful, reliable and safe? Using your analogy about fire, I think that soon after humans developed the ability to create fire, they understood that fire can be put out with water. I think one of the fears about AI, founded or unfounded, is this: Is there an off switch? Is there a technological, moral or political off switch to make sure that it doesn’t, I guess, get out of the lab, for lack of a better phrase?
Mehlhorn: Yeah. So, two answers. One, no, it cannot be turned off. It cannot be made safe, and I’ll explain why.
Dunne: OK.
Mehlhorn: But two, there are ways that it can be used for good and help humanity rather than damage humanity. So let’s talk about the first: why it cannot be made safe and why it cannot be unplugged. Consider the agents that are created by artificial intelligence. These are self-replicating, self-motivating, self-governing sets of code that can be released out into the world to do things. They’ve already been released into the world, and they’re already doing things. You can’t shut the barn door after the cows have left and think that it’s going to be OK. If we try to, quote, unquote, turn off artificial intelligence, what we are doing is taking the parts of artificial intelligence that we can control and shutting those down, leaving the world with all of these agents that are, in some cases, breaking through systems and, in some cases, committing crimes. If we try to stop AI, the balance of AI power will be the parts of AI and the AI agents that are committing crimes and are uncontrollable. So that’s the bad news. There’s no way to completely stop AI now that it’s out there in the ecosystem. It will exist, and it will keep advancing. The good news, as a way of thinking about AI, is that the solution to bad and dangerous AI is good AI, and I’ll give you a very concrete example. Today, artificial intelligence agents are out there that can help build bioweapons, biological weapons like COVID, except more advanced and more powerful, that can be released on enemies or by terrorists. That is out there today because AI is out there today. However, there are also artificial intelligence tools out there today that can protect us from bioweapons, identify them earlier, give us tools to solve them, and give us drugs, medicine and antidotes. All of that together means that artificial intelligence can be used to solve the problems that artificial intelligence has created. But even more important than that, Michael, artificial intelligence can be used to solve massive problems that existed before the development of artificial intelligence, such as cancer. Artificial intelligence tools can dive in at the level of the specific genes, the specific protein folds and the specific cells that allow cancer to develop and kill people. Artificial intelligence is already working on breakthroughs in medicine, math, physics and materials that will allow humans to solve many severe problems that have been with us for a long time. So if we focus on using artificial intelligence in those kinds of ways, then we have a tool to fight back against the people who are using it badly.
Dunne: Let’s turn to a realm that I know you spend a lot of time in: politics. Tell our listeners how AI is currently being used in political campaigns. Is it effective?
Mehlhorn: Yeah. Artificial intelligence is a major issue in political discourse right now. It is not as important to people as inflation, perhaps, but it’s probably becoming a top-five issue, and a huge part of that is data centers, as you pointed out at the beginning of the show. Just to be clear, data centers use a lot of energy, and they generate a lot of heat and noise. So if you’re an average American going about your life, and a data center gets plopped down next to you, you will find your electricity bills skyrocketing. Your water bills may go up as well, and the ambient noise and unattractiveness of the neighborhood will get worse. These things are big, they’re not very attractive looking, and they generate noise and a lot of heat. So if you’re in a place that has temperature problems of any sort, the artificial intelligence data centers are going to require you to run the air conditioning more. It’s really a tough situation. On the other hand, as we said before, you don’t want the United States to fall behind on artificial intelligence. So it’s a close balance. In an ideal world, you would make sure that communities that are potential sites for data centers participate in the process. They get to choose, and if they do choose, then all of the economic wealth generated by a data center is shared with the community, so that everybody has a share in that wealth creation, and that offsets the harms the data center creates for the community. That would be an ideal solution. That is not the way politics are going today. The way politics are going today is that President Donald Trump and many of the technology executives in Silicon Valley have what is called a pure accelerationist approach, meaning we’ve got to just keep going hard, hard, hard, and that means data centers being built everywhere with no regard for local communities. That’s kind of the Republican position. The Democratic position is largely that we need to slow down and maybe even stop the development of data centers. So right now, the politics heavily favor the Democrats on this important issue. Most people don’t like data centers, and most people aren’t personally getting a lot of the benefits of AI. So for most Americans, including most voters, the Democratic position is much closer to what they want and what they think is good for them.
Dunne: Do you think that will factor heavily in the upcoming midterms?
Mehlhorn: ‘Heavily’ is hard, just because inflation is such a dominant issue. Most of the midterm election will, I think, turn on prices: gas prices and all the other prices. The president has implemented a lot of tariffs, which increase prices. The president has engaged in a war with Iran, which massively increases energy prices. And the president has gone deep into deficit spending, with the highest deficits we’ve seen in a long time. Those three factors create such a cost-of-living problem that it’s going to be the dominant issue. But data centers are a pretty important top-five issue, and they very much favor the Democrats. So in combination with inflation, the data center issue does suggest that Democrats will have a very good midterm.
Dunne: OK. You’re a Democratic strategist and someone who looks at campaigns and political parties. One of the things about AI that I think people are confounded and astounded by is image creation. I would imagine that, whether you’d use it or not, the holy grail of a political campaign would be generating an image that looks, for all the world, as real as possible, showing their opponent doing something terrible. I am wondering: Is image creation, blurring the lines between reality and fiction, perhaps a new frontier for AI in political campaigns?
Mehlhorn: Yes, absolutely. And there’s no good news about that. That’s bad news. But there is a little bit of, not quite a silver lining, but a mitigating factor. The bad news is that humanity, for the last few hundred years, has been able to rely on artifacts to agree upon truth: pictures, videos, audio. When people saw those things, they could agree that was a thing that happened, and that was very important in many ways. For example, after World War II, there were a lot of people who denied that the Holocaust could have ever taken place, and the world saw the videos and overwhelmingly agreed that it had happened. You got this period of 50, 60, 70 years where there was a general agreement that we couldn’t go there again. That was very good for humanity, and it relied upon the ability to trust images and videos. We’re now in a world where you cannot know with your naked eye whether something is real or fake, and that is bad news for humanity, and it is bad news for politics. The thing, Michael, that I think is a little bit of a mitigating factor, and it’s not good news, is that most of America already lives in bubbles where there are no agreed-upon facts. If you watch Fox News, read the Gateway Pundit and listen to Tucker Carlson, you’re already in a world where you won’t believe the things that The New York Times or bipartisan Senate commissions might say. Take the bipartisan congressional investigation into the attacks on Jan. 6, 2021. To my mind, as somebody who reads The New York Times and The New Republic and whatnot, that evidence is overwhelming that there was an attempt by Donald Trump and his allies to overthrow the U.S. Constitution. But if you talk to someone who’s a Trump loyalist, they will tell you with a straight face and with great certainty that it was a peaceful protest that just got a little bit out of hand, and that the Democrats overreacted. So for better or worse, we’re already living in a world where people experience different realities. Artificial intelligence is just going to be the rancid cherry on top of what has already been created.
Dunne: OK. Dmitri, my last question for you is this, and I think it’s fascinating because you’re also a writer. You’ve got a political thriller coming out called “Archipelago of Light.” To me, it’s interesting because we’re talking about this very serious topic of AI, but obviously it’s fertile ground for great storytelling, isn’t it?
Mehlhorn: Yeah, it sure is. When you’re talking about science fiction, you’re talking about trying to predict what’s going to happen in the future, and artificial intelligence is a key part of that. The reason I wrote fiction after a life spent in nonfiction is to try to explore how things might play out in the next 15 years. In my book, “Archipelago of Light,” you have a world that is increasingly affected by climate change, where there are more and more climate refugees and a great deal of political instability because of that. The book has some dark tones, but it ends with a fundamentally optimistic idea that we’ll make it through.
Dunne: He’s Dmitri Mehlhorn, Democratic strategist, founder of the Atoll Society, executive for information businesses, as well as an author. Dmitri, fascinating stuff. Thank you so much for coming on.
Mehlhorn: Michael, thank you so much for having me.
Dunne: That’s the show for today. All episodes of “Oregon on the Record” are available as a podcast at klcc.org. Monday on the show, we’ll move from artificial intelligence to real human action and talk about two programs in Oregon that rely on volunteers to help rescue animals along highways and in rivers. I’m Michael Dunne, host of “Oregon on the Record.” Thanks for listening.
