AI's Promise and Peril — and How to Simplify: Jim VandeHei
AI, Energy and Climate Podcast
Jim VandeHei, CEO of Axios and author of the new book Simplify, joins host David Sandalow to discuss his year-long obsessive use of AI tools like ChatGPT and Claude, highlighting both their transformative potential in healthcare and writing and
Key takeaways
- Jim VandeHei has used ChatGPT and Claude for hours daily as an 'AI lab rat,' inputting extensive medical data for himself and his wife to achieve results he found superior to human doctors.
- VandeHei warns that dozens of insiders from DeepMind, Anthropic, and OpenAI have recently quit in public protest over alignment risks and the potential for AI to destroy humanity.
Main topics
- Personal AI adoption and prompt engineering
- AI safety, alignment risks, and insider warnings
Notable quotes
"I've spent the past year using AI obsessively... it's way better than most people think."
Conclusion
VandeHei's dual focus on AI's transformative potential and its existential risks underscores
Transcript preview
Speaker 1 (0:04) Welcome to the AI Energy and Climate Podcast, a special series from the DSR Network hosted by David Sandelow, Inaugural Fellow at Columbia University's Center on Global Energy Policy. Join us as we talk with leading experts to explore the intersection between these critical issues that will impact the future of each and every one of us. Speaker 3 (0:31) Hi, this is the AI Energy and Climate Podcast. I'm David Sandelow. This summer, I read two articles that I thought were excellent and got my attention. One was about how to use AI, in particular, how each of us can use AI tools most effectively in our day-to-day lives. The second was on AI policy, taking a somewhat broader view and recommending AI policies for the U.S. government. Both articles were written by the same person, Jim VandeHei, who is the founder and CEO of Axios. I'm an admirer and regular reader of Axios and also of a book that Jim Vande Hei and two of his colleagues wrote called Smart Brevity. In my view, it's a brilliant guide to communicating and in particular to communicating concisely in ways that will have an impact. So I reached out to Jim VandeHeide, asked whether he might be available for a conversation on the podcast, and I was delighted when he said yes. It turns out Jim has a book that he is releasing this week called Simplify, which I've had a chance to read and highly recommend. It sounds some of the same themes as Smart Brevity, but in a slightly broader context. Jim and I had a chance to meet and talk a couple of days ago when we covered a lot of ground on topics related to AI. I hope you enjoy our conversation. Jim Vande Hei, it's great to meet you. Welcome to the show. Excellent to be here. Thank you. It's great to have you. And so just to start, I am a smart brevity fanboy. I think it brilliantly sets out fundamental lessons for communication, and I assign it to all my students. If any of our listeners haven't read Smart Brevity, I highly recommend they go out and get it. I'm also a big fan of Axios. Read it regularly. It's an invaluable news organization. So thank you. for that. I'm thrilled to have the chance to talk. Excited to chat. So you wrote two articles this summer that got my attention and led me to reach out and see if you might be available for a conversation. The first article was about how you use AI. The second article was more about policy issues on AI. And then you have a book coming out called Simplify. And so if it's okay with you, I thought maybe what we do is start with those articles and then go to your book. And starting with that first article. I thought the first article was very interesting. You said in that article, I'm quoting, you said, I've spent the past year using AI obsessively. You then wrote, it's way better than most people think. And when you said AI is way better than most people think, I wonder, say a word about what you meant by that. Speaker 4 (3:10) Yeah. In that piece, I talked about how I've kind of made myself an AI lab rat. Because I run a media company, because I write, because I think AI is going to be that transformational, already has been that transformational, I really wanted to understand it. So I think I was early to utilizing it, especially ChatGPT and Claude. And I've used it for probably on average hours a day now for well over a year. And what I've discovered is there's a big gap between people who utilize it all the time, particularly for business needs, and people who use it passively. If you just use it passively or once in a while as a search engine, it's interesting. It's better than search. It can sometimes be very clever, sometimes blow your mind, sometimes frustrate you. But if you use it very persistently and you really fine-tune it, that's where you start to see the magic that I think people out in the Valley see. And I'll give you two examples. For health, I've really, because I wanted to be a lab rat, I'm doing what I not necessarily would advise other people do, but I put every piece of medical information ever known about me into ChatGPT and done the same in a CLAWD agent for my wife, who has a much more complex medical situation that she's dealing with. In both cases, I've spent a lot of time fine-tuning kind of where it draws its answers from, like really trying to curtail it to medical journals that I trust, literature that I think has been validated, and really fine-tuned all of the results that I'm plugging into it. In both cases, I have found hers and mine to be way better than any doctor I've ever spoken to. in terms of its ability to really put pieces together, to keep track of all the complex pieces of our medical history, to be able to even think about things like supplementation or workouts as it relates to maybe things that I'm dealing with or that my wife is dealing with. And I know that the AI companies, because of the... because of the kind of make or break nature of health and the fact that it's highly regulated. They put a lot of time early on into really making sure that that information is even more airtight than some of the other stuff you might come across when you're using an LLM. So that's one area. The second is just in me using it as like a thought partner, sparring partner, like how to be a better writer, thinker about the different things that I write. And there, it required a lot of work on the front end. Like I inputted. hundreds and hundreds of columns I've written, every speech I've given, every video transcript of any interview I've ever done. I've written multiple books. I put a lot of that into it. I've given it access to my notes function, which is where I just keep my random brain farts that I have stored on my iPhone. And then I really worked on fine-tuning it to really understand how I write and how I don't write. And also how to think about fact checking in kind of a multi-step way so that it's really doing a deeper set of fact checking than it might otherwise do. And in both cases, I find it to be like just astonishingly good, astonishingly helpful for the work that I want to do. But like I said, I had to put in that work. And had I not put in that work, I'd probably be like everybody else and be like, oh, it's fine. It's interesting. But I don't understand what all the hype is about. Speaker 3 (6:34) I found a lot of people are concerned that AI is going to substitute for thinking, that if people use AI as much as you're describing using a gym, that somehow their writing capacity is going to erode, their thinking capacity is eroded. You've been a writer for your career, but you're saying that it's really useful for you. Could you just speak to that concern? Speaker 4 (6:51) Yeah, I mean, again, I would differentiate between me as a 55-year-old CEO who's written a lot and thinks a lot and is paid to write and to think and somebody who's in college who's just having their brain formed in how to think. I do think there's a real danger that if you outsource your writing, outsource your thinking to any machine. that you're just going to inevitably, you're going to be dumber, you're going to be duller, you're going to lose the ability to fact pattern recognize, or just to think deeply about things that might perplex you or perplex the world. You know, I'm 30 years past that. And I kind of, and because I'm kind of aware of the pitfalls of AI, I think I've been able to navigate that. And in fact, I'd argue I'm a lot smarter because of it. But I do think I'm in a very tiny. in a tiny set of people who are probably getting that benefit right now, mainly because I have the time, I have the business instincts, and I also have the business need to really be able to understand it. But I think that's a legitimate issue. You know, I go to colleges and I'm of mixed minds. I'm pretty frustrated when I talk to most of the universities because I find that I don't know that they're doing a great service to young people because I think there's two different things. There's one thing, don't use ChachiBT to cheat or write. If I were a professor, that is exactly what I would tell my class. I would say, I want your brain, I want your writing, I want your thinking. That's different than equipping somebody who decides to go into marketing or decides to go into business analytics or whatever they go into. That's different than equipping them to be able to use the technology as a power tool. when they get into the workforce so that they're ready to enter the workforce as a super well-equipped young professional. And I think that's where a lot of the universities are falling short. They're saying, oh my God, this is a scary thing. We don't want you cheating. It uses too much energy. Therefore, like it's bad, bad, bad. And it's both. It's like anything. It's everything in life, but certainly any technology, it's good and bad. And this one. could probably be so good it cures cancer and so bad that it does really catastrophic things to humanity. I'm very eyes wide open about that. But I think that any young person, God, this is a time to think. This is a time to reflect. You're going to deal with some of the weirdest things possible. Ethical dilemmas that you and I would have thought were pure science fiction five or 10 years ago are going to be staring these young people in the face, right? Like what happens? when you have a machine that could offer therapy superior to any human therapist, or enter into a relationship with a thing that's not a person that seems to know you better than you know you, and seems to make you feel a way you want to feel. Like, those are really perverted, tough things that could have good outcomes and could have bad outcomes, and they require... real depth of thinking. And I think that's why you go to high school and go to college. It's to learn how to think and then learn how to be a professional. And so I think there's a lot of work to do. I mean, I'm sure we'll get at this. Like, I feel like government and business and academia have really let the American public down. I think we had this real opportunity for this to be a real golden moment to come together. master a technology to make sure it does good for workers and good for all and not just great for a few rich, super powerful people already, to be able to show that government could rise to the moment and make sure that it's working for everyone and not just working for the few. But instead, I feel like we frittered away