The AI Expert Who Will Change Your Mind About AI - Josh Tyrangiel

0:00Study and play. Come together on a Windows 11 PC. And for a limited time, college students get the best of both worlds. Get the Unreal College Deal. Everything you need to study and play with select Windows 11 PCs. Eligible students get a year of Microsoft 365 Premium and a year of Xbox Game Pass Ultimate with a custom color Xbox wireless controller. Learn more at windows.com slash student offer. While supplies last, ends June 30th. Terms at aka.ms slash college PC. When you need to build up your team to handle the growing chaos at work, use Indeed Sponsored Jobs. It gives your job post the boost it needs to be seen and helps reach people with the right skills, certifications and more.

0:42Spend less time searching and more time actually interviewing candidates who check all your boxes. Listeners of this show will get a$75 sponsored job credit at Indeed.com slash podcast. That's Indeed.com slash podcast. Terms and conditions apply. Need a hiring hero? This is a job for Indeed Sponsored Jobs. Morning decisions. How about a creamy mocha frappuccino drink? Or a sweet vanilla? Smooth caramel maybe? Or a white chocolate mocha? Whichever you choose, delicious coffee awaits. Find Starbucks frappuccino drinks wherever you buy your groceries. We got hypnotized by social media, where for the last 20 years, it kind of showed up on your phone.

1:20It's easy. Oh, it was fun too, right? But it turns out there were some harms, and those harms were known by the companies that made social media. Some of those companies now make AI. And so if you were a passive consumer of AI in the way we have been of social media, we're screwed, man. AI is so much more powerful than social media. So you actually have to demand what you want from it, and you have to make a market for the good stuff. When I talk to the guys running the labs, they're working on technology that nobody has ever made before, that's volatile. They have huge competitors. They have trillion-dollar caps that they're looking at.

1:52They have investors who are saying, where are the returns? They have energy issues. They have computing issues. You know what they're not thinking about? Boy, I hope this works out great for the Republic. That's actually on us. And so we have to know what to do. We have to know what we want out of it. And then we have to insist that they deliver. Welcome to Open Book. I am your host, Anthony Scaramucci. Joining us today is Josh Tiringel. The title of the book is AI for Good. Wow, Josh, I got to get into this with you. How real people are using artificial intelligence to fix things that matter.

2:25and so i am a believer in what you're saying and so i appreciate you coming on the show because i think we have to proselytize this more but let's go to your career if you don't mind because uh you got an interesting career my friend your media technology storytelling intersection and the book reads very breezy josh so you know how to write okay but tell us a little bit about your background and tell us what drew you to the AI conversation, which I think is the most important conversation that we're having right now. Yeah. So I'm, um, look, I'm a career journalist for the most part, but I, I grew up in an era where it was very easy to move between writing, editing, digital and television and documentary.

3:11And so to me to stay in one lane is kind of boring. So I just kind of veer around the highway, wherever the most interesting possibilities are. I spent about a decade at Time Magazine. I literally did everything the place had to offer. I started writing the, you know, you remember those old three-inch obits they used to run on the front? So that was my first job. Ended up as the number two at the magazine. Went to Bloomberg for a while. And I think I actually, we actually met at Bloomberg many, many years ago. I was the chief content officer and ran Bloomberg Businessweek. Ran the TV over there.

3:45Was Dan Calaruso there at that time? Yeah. Okay. Great guy. I then left Bloomberg and went to HBO and vice, launched a new nightly news show called vice news tonight, did all the vice stuff on HBO. And ever since I've just been kind of going where the, where the best funnest opportunity is. So in 2023, right when chat GPT came out, a good friend of mine was then running the Washington post op-ed desk. and the way he just he called me and he said look it's a hellscape down here with ai it's like it's an information hellscape we have uh journalists who can't really quite figure out or explain the tech we have subjects who seem to be racing ahead without telling us what it's good for what it's doing um our readers are furious they're really angry that this thing is like showed up on their desks uh and they offer me the chance to do a column for a year and just kind of chronicle it And so, listen, I was like, sounds great, right?

4:47Let me dive into the deep end of this. Long story short, after about eight weeks, I realized it was not all that different from writing about money because everybody was focused on raising and spending and their future valuation. And very few people, you know, what they would do, and I know you remember this in those first couple of years, basically, they'd say, well, it's going to cure cancer. Or on the other side, no, it's going to end civilization. And I just don't believe that those extremes are really the outcomes. And so what I ended up doing is going hunting for reality and trying to find people who care about the things that normal people care about, you know, healthcare, education, government, human connection, and just diving in to see what could we do with this stuff?

5:32What can we do with it today? Not 10 years in the future. How hard is AI to work with? What are the possible gains? And I found a whole kind of AI counterculture of people who just don't get written or talked about enough so that's really what the book is okay so you know we have stanley kubrick we have the terminator james cameron uh we have elon musk uh we have this i mean i mean this did you read this sam altman uh article by Ronan Farrell. Oh, yeah. Okay, so we have, I mean, we have a group of people involved in this thing that to me, I would worry about some of these people. You know, maybe not Elon because Elon's worried about this.

6:21So that makes me comfortable, but I don't know. You tell me. So we're okay? All these different types of people? There's no existential threat? I wish there were more types of people, to be honest. You know, there are four to six people who run the big AI labs and who are really in charge of most of the advanced technology, that's not enough in my book, right? That's a concentration of incredible power in the hands of just very few people. And all those people, whatever you may think of their personalities or their morals, all those people are more or less alike, right? They're driven by business.

6:55Now, two years into it, they have investors and those investors want returns. They are driving the tech to a very specific place. So I'm not here to tell you it's all going to be okay. I'm really not. I think that to this point, the United States has kind of done what it usually does when something new comes along, which is abdicate the need to regulate, to try and get ahead and be competitive in the market. And that's great. Except for the fact that this tech is different. It really is different than any other technology that we've ever had before. And it's kind of hard for regulators, politicians, and even normal people to wrap your head around the fact that these machines make themselves smarter, right?

7:34So we think it's one thing one day. The next day it can be something totally different. And I can't tell you it's all going to be okay. I think it's a real civilization testing moment. And we're going to see how we respond. Which one is the best one, Josh? Best one meaning kindest, most moral, thinking about society? Yeah, which one are you using the most? Um, you know, I vary. I use Claude a fair bit. I think that it is the most direct and most honest and most likely to tell you three very important words, which are, I don't know when it doesn't know an answer, or I can't help you, or I should.

8:14It hallucinates the least amount of the ones that you've experienced. Is that fair to say? No, I actually think that all of them have brought the hallucinations down a fair bit. The question is, are they going to be honest with you about the access to information they have? And are they going to be honest with you about the things they shouldn't do, right? So copyright infringement, making things up just to satisfy you. And it waxes and wanes. So one of the hard things you get our heads around is that these products are living products and they change day to day based on the number of inputs they get.

8:49I would say right now, Claude and GPT are the most consistently good and give me what I want. Gemini is very close behind. I think Grok and Meta's Llama are further behind and significantly so. Some of the Chinese ones are pretty good. They're competitive on good days with GPT and Claude. When I look at this, when I use it, obviously, I feel like I have a 300, maybe a 400 IQ business partner. Uh, and I can ask it questions and it responds to me and then it listens to what I have to say. And then it also, you know, I can coach it to go in a direction that I want it to go in. And it's very, very flexible.

9:34Uh, won't that go exponential? I mean, a year from now, if you and I are together, are we in a gentic super intelligence or we in general AI at that moment or where are we? Yeah. So you're referring to AGI, which is artificial general intelligence, which is really what these guys are chasing. Right. And that's the got machine. That's the machine that knows everything, is capable of working on really advanced problems, a super intelligence. And part of the reason that they're so interested in it, you can imagine the market value of a company that has AGI because it can create all sorts of other companies.

10:13um i personally don't know when we achieve agi and i don't even know if they know when we'll have achieved it i'm not sure it's something that ordinary people are really going to care about in the short and medium term because to your point like it's already pretty damn useful um getting to that next phase where where people are communicating through their agi agents i don't know one one common thing about the guys that we just talked about their specialty is not not human nature, right? Their specialty is computing and financing and building. I don't think they all know how people use these incredible products.

10:52You and I use it to make ourselves smarter, to save some time, to explore the world a little bit. I don't think that's what they're building it for. So I don't think regardless of how smart the machine gets, I'm always going to want to talk to you. You are going to be a surprise to me. I'm going to get creative friction from talking to you might i have our our calendar bots schedule this conversation so that you and i don't like six emails yeah maybe but i still think that the the reward of human contact is being overlooked and i think they just it's just not the thing that they concern that they're concerned about if we went into that agi god machine um you know there are stories that you know these things uh lie uh they potentially could blackmail there are fears at anthropic that without a human being involved, maybe the AI in terms of a risk management simulation, a Nash game theory analysis may want to deploy nuclear weapons, which would create a nuclear holocaust.

11:51And none of that worries you? Oh, it worries me. Listen, I didn't do this book as a counterweight to try and offset the negatives. The negatives are real. The one thing that actually keeps us safe is this concept of human in the loop, right? And that means that the AI is just not off doing its own thing unsupervised. That when it comes time to make decisions, to do complicated moral judgments, there is always a human being monitoring the machine and overruling it. And so what I found actually in every one of the realms that I explored is that if you want a successful AI product that serves human beings in meaningful ways, you need a human in the loop at every phase.

12:33So it's not like you just turn the AI loose in a classroom and all of a sudden the kids get smart. In fact, the opposite happens. Same thing in a hospital, same thing in government. I mean, one of the ideas behind Doge was, well, let AI run everything. Well, it turns out government is for people and by people and AI misses a bunch of stuff. It can help us. It can make us more efficient, but not without a human in the loop. So the notion that you're just going to set programs free, I think is delusional. I think it won't work. I think it creates dangers. But the idea that you have this, the same way you use it as a sort of partner and assistant, that makes sense to me in these realms.

13:10Thank you for tuning in to Open Book. And if you haven't already, please hit the subscribe button below so that you're the first to know when our new episodes drop each week. We've got a lot more coming. And now back to the show. So let's say, you know, we get it right in the U.S. We have checks and balances on it. What about our adversaries or what about somebody developing this sort of technology where they take the code from Anthropic and they hijack it and they bring it down to Africa or they send it to China? Worried about any of that? Oh, yeah. Listen, I think and you saw the Trump administration is worried about it because they pretty rapidly changed course yesterday.

13:48is. So just to make sure the listeners are up to speed, the Trump administration put David Sachs in charge of both AI and crypto. David is a very smart guy. He's also an investor. And his attitude toward AI was, please go. Go as fast as you can. And by the way, here's some of my money so that I can make some profit on this too. That's all well and good. There's really good cases for why we should stay out of the way and not regulate, get as good as we can faster than our adversaries. A couple of weeks ago, Mythos, which is Anthropic's most advanced code tool, comes out. And it turns out, as you know, it exposes vulnerabilities in pretty much every possible cybersecurity route.

14:36And Anthropic voluntarily submits Mythos to financial services companies, banks, the government. We can't let this kind of thing loose in the wild. Trump administration didn't insist on any of that. Yesterday, I think they realized how fast this is moving and that it will be deeply unpopular if an AI product is in the hands of bad actors and takes down our financial system and takes away all the privacy controls that we actually need. And they signaled, maybe we do need to regulate this because it's deeply unpopular. So I'm definitely worried about it. Do you think Anthropic will resolve its situation with the U.S.

15:19government? I do. I think that there are two things going on. One, Dario Amidai, who runs Anthropic, very smart, very ethical for the most part, he had some lines. He didn't want the government using Claude or any anthropic product to make kill decisions on its own. So they had to resolve that, right? But everybody thinks he had no leverage. In fact, he has a ton of leverage because you know what's indispensable to the Pentagon right now? Claude and other anthropic tools. And so they are going to work, my sense and from my reporting is that they're going to work it out and they're going to work it out pretty quietly and pretty quickly because these tools are an advantage already.

16:01You know, we've seen with Palantir, which is relies a lot on AI and machine learning here in the book. I quote a general who says, look, the Ukrainian people are incredibly brave. They're courageous. They fight like hell. This isn't a war without iPads and Palantir. And so I think that the military is so dependent already on AI driven systems that this is going to work out. You know, and I believe that, by the way, because it's in the best interest of everybody and the cooler heads will obviously prevail. I guess the thing that you wrote that has me really encouraged, and I want to talk about the Cleveland Clinic, the IRS, the Indiana School District, and the Pentagon.

16:43What's the common variable that determines whether an institution will actually use it or it ends up gathering dust? Yeah, I mean, it's the great question of the book, right? And so one is talent, right? The talent inside an organization is the first thing, and not everybody has a talent for AI. And it can't just be a talent for computing, right? So the Cleveland Clinic, I talked to the CEO, who's a cardiac surgeon, a great one. We basically said, look, our theory of AI is that everything we do, including AI, has to serve patients. And so we're not going to bring computer scientists in. We're going to have doctors be the technology managers.

17:23And so they have a series of volunteers. And these people are already busy, but they volunteer to bring the technology into the system. And so Cleveland Clinic, they did just one massive thing, right? there's a very talented doctor from an emergency room doctor a very talented icu nurse and the two of them both were on what's called the sepsis committee inside cleveland clang now sepsis is the most deadly thing in america and around the world it's an infection uh it looks like about 30 other things and it's we haven't ever been able to really fully cure it right we we can the crazy part is we can cure it if we detect it right we detect it it's just an infection you treat it with antibiotics, it goes away.

18:06If you find it too late, it kills you. And the way the nurse described it to me, she said, yeah, it's like you see a body rotting from the inside out. You can smell sepsis inside a body. But because it looks like so many other things, it can often show up and kill somebody inside a hospital, right? The Clinton Clinic was losing thousands of patients a year, and it's one of the best healthcare institutions in the world. So they went on this mission. Some of it was training nurses and doctors to be on the lookout for it. and then they brought in a sepsis detection AI-driven tool. You let the tool loose on its own, and it does a couple things.

18:40It'll detect some sepsis, but it'll also beep all the time. It'll distract doctors from other care. And they're so busy in an ICU that they'll just turn it off because they've got other things that are more concerning. So for about a year, these two doctors and their teams worked with the AI team at a place called Bayesian Health. they tweaked the algorithm they tweaked the noises and the language and a year and a half later they'd reduced sepsis mortality by 40 percent in the hospital and so that's a couple thousand people who's who are living not entirely because of ai but certainly ai played a role and the recipe there which is the same everywhere is you got to have talented people who actually give a and are willing to do extra work to make sure that you get the best out of the AI.

19:29But there was a human being in a back room monitoring Bayesian when I was there. Lovely woman named Dana, looking at every time it pinged every flag, and she would determine, is it right? Is it wrong? Give the machine feedback. And Anthony, I mean, as you know, a Blade server isn't going out to help a patient, right? So if a couple minutes went by with a serious alert, she'd get out of her chair, she'd go to the bedside, and she'd call the doctors over and say, look at this. And they'd assess. And if it was right, treat with antibiotics right away. That patient is alive and saved. And oftentimes, no damage.

20:06And if not, for me, I thought it was the most encouraging stuff that you've written about. I think that's the messaging that, you know, I did an interview myself. I'm obviously pretty bullish. I had Peter Diamandis on the show for his new book. And I'm obviously very bullish on AI and the prospects of AI. And even for some reason, though, I'm not ultra worried about the Stanley Kubrick or Terminator thing. But should I be more worried? You seem like a pretty balanced guy, Josh. Yeah, listen, I am. I don't want to be delusional about the risks, but I also am genuinely optimistic about the use cases to improving the things we care about and improving our lives.

20:50My worries about AI are generally the same worries I have about people, right? And there are people in power who don't do the work to understand what AI is, who might be self-interested. If they're in charge of AI, it gives me pause. And so one of the things I've been really diving into recently more out of curiosity than a story on the horizon is like, what do the Chinese want out of AI? Like, what do they really want out of it? Because we have turned this into a kind of like US versus China, somebody has to win. And the more we know about what people want out of it, the more I feel a little bit calmer in some cases and less calm in others.

21:32And so if you're an ordinary person trying to figure out how worried should I be, don't look at the tech. Look at the people in charge of the tech. It's a much easier, more reliable way to understand where the risks are. Well said. Let's say I handed this book to a mayor or maybe I gave it to a hospital CEO or a school superintendent. Let's say I gave it to him over the weekend. What would you want them to implement on Monday morning after reading this? um that's a great question um i would want them to understand that the technology can help them do their job a lot better and serve their citizens and customers a lot better and i'd want them to understand that you can't that find the two most diligent caring obnoxious hard-ass people in your organization and make it their job to implement and implement it carefully implement with outcomes understand that they're just the number one thing I interviewed 70, 80 people for this book.

22:35The number of times I heard from people, there are no silver bullets with AI. I just stopped writing it down in my notebook, right? There are not. The tech is great. It doesn't matter if you don't implement with care and with skill. There's great lessons for public policymakers in here about the right way to do it, the wrong way to do it. So I think that's what I want them to know. The tech can help you. The people are going to be the thin that gets you to the end zone. one of the brilliant things about the book is that you're talking about the enhancement of human judgment. And you're talking about humans using the AI and not necessarily being replaced by the AI.

23:13And so let's say that I'm a human listening to this interview. What would you say to a human to get started? Because by the way, a lot of people in my industry, I know this is sort of crazy, don't use it. I mean, I've implemented it throughout our organization and I use it in a lot of my stuff. It's also a great teacher, but some people don't. So how do you get them to reduce their fear and or intimidation? Yeah. And mine too. I got the number of people who have come up to me and said, I can't believe you use this stuff. The hard line, to be honest, is do you think a defensive crouch is a strategy?

23:51In what point of human history is like curling up in a ball and saying, go away, ever done anything. And so I would say, learn the tools. They're actually not hard to learn. 20 minutes a day for a week, and you kind of get a sense of what an LLM can actually do for you. You don't need to get advanced into agentic or coding. Just understand what it can do, right? And then try to figure out, like, what do I not want it to do? What are the behaviors I don't want it to do. To me, the main issue is that we kind of got hypnotized by social media, where for the last 20 years, it kind of showed up on your phone.

24:30It was easy. Oh, it was fun too, right? I mean, both of us were on Twitter going crazy, right? But it turns out there were some harms. And those harms were known by the companies that made social media. Some of those companies now make AI. And so if you were a passive consumer of AI in the way we have been of social media, we're screwed, man. We're screwed. AI is so much more powerful than social media. So you actually have to demand what you want from it and you have to make a market for the good stuff. So that's my main thing is like when I talk to the guys running the labs and I'm saying this with some sympathy, they're working on technology that nobody has ever made before that's volatile.

25:14They have huge competitors. They have trillion-dollar caps that they're looking at. They have investors who are saying, where are the returns? They have energy issues. They have computing issues. You know what they're not thinking about? Boy, I hope this works out great for the republic. That's actually on us. And so we have to know what to do. We have to know what we want out of it. And then we have to insist that they deliver it. And that means saying, yeah, I want this. I've talked to numerous senators, congresspeople. They have no idea what I'm talking about. You know that tech is not their thing.

25:46We have to insist that people actually run on these issues. We have to insist that we get the best service from our government with these things. And then we have to not do some stuff too. And that really means, evaluate the AI. Is the AI there to remove human beings from a certain realm? Don't use it. Is it there to make better spam and porn and cyber? Like, don't use it. Trust me. All of these companies are monitoring our behavior. You get to vote with what you do with it. But if you sit it out, if you say it's not for me, none of that's going to work. It's 10 years from now, Josh. We're looking back on this moment in AI and your book.

26:27What do you hope this book will have changed about how we talk about it? Um, I really hope that it inspires people who care about things in society to pick up AI and start, start playing with it and start owning it and start saying, you know what, like I'm in a school district. Here's how we used to think about education. Here's how we used to think about technology and education. Let's do it differently. um you know to me the i was the most inspiring thing i saw in real time in the book is i was in indiana at watching a 10th and 11th grade class learned steel styochemetry which you know i got the the teacher sent me this google doc the night before and i was like oh god i have to go back to school this is gonna be a nightmare and basically what she'd done with conmigo which is a tool that con academy collaborated on with open ai she just decided when she was teaching during covid these kids are tuning me out.

27:21You know, they're sitting there, they got their laptops open. I know they're looking at multiple tabs. Now they're in back in class with TikTok. And she said, I have to reinvent what I do because guess what? Like the world is moving forward. They have all this new technology and I'm sitting here lecturing at the top of the class, like it's 1885. No wonder I'm losing them. And so she turned pretty much every lecture in her curriculum into a lab. She turned a quiz into a lab. And so those kids now, when they come to class, the computers are open because they're guiding them through a bunch of steps, but there's no silent note-taking.