Can Truth Survive AI?

It was the world before Chatech EPT in the world after Chatech EPT. There is a real unanswered question about whether or not truth will exist as we've grown up knowing it for generations. Once AI is embedded in every part of life. As video that's generated by AI wholesale, often is completely indistinguishable from actual video from real life. I think table stakes are anyone who can determine fake from real today. Let's assume that six months from now, that skill will be gone. So most news consumers just kind of go, Phew, it could be fake. Right, and that's a dangerous place to be. The future of truth is human.

The robots can help. Who should? Steve, welcome to the show today. Excited to talk to you. I wanted to ask what made you decide that this was the moment that required a book that's not just about AI and misinformation and media, but about the future of truth? So the journey of the book actually has a great story. One of the things readers sometimes forget is that books take five years, four years, but not the whole journey. So early on, the book began in COVID. And I was at NYU getting my master's degree studying truth. And while scientists will tell you that AI has been around for a long time, in terms of society, it kind of didn't exist.

And then it did. There was the world before chat GPT in the world after chat GPT. So I was midway through my master's program when chat GPT kind of went bing. And I went to my advisor and I said, I kind of think the thesis has to change, because truth is going to be really different now that there's this robot in the loop that I I mean, I knew it existed, but it was always theoretical for me. And that framed the book because there's the world before chat GPT in the world after. And truth, and I say this gently to your listeners and viewers so they don't go racing away from your podcast. I mean, there is a real unanswered question about whether or not truth will exist, as we've grown up knowing it for generations, once AI is embedded in every part of life.

Well, let's dive a little deeper into that, because that's really at the heart of the issue. Here, how will AI affect truth as we know it or did know it prior to chat GPT? So let's break out the two words. There's subjective truth and objective truth. And objective truth is easy. I mean, one plus one equals two. We did, in fact, land on the moon. I mean, they're facts are facts. Allegedly. And then there's beliefs. And as we get to know each other, my guess is we'll find there's some things we agree with and some things that we don't. And my experience with humans is that the things we don't agree with are way more interesting.

Because if I'm a curious person and a learner, as I am, I'm going to say, wait, how did you get there and explain that to me? And so part of, I think, you know, what the book set out to do, I mean, as I started doing the research, it was clear to me that any of the chapters in the book could be their own book. I could have written a book about autonomous warfare. I could have written a book about arts and AI. I could have written a book about love and AI. And every one of those books would have had a relatively small audience of people that are either, you know, own a dating site or, you know, I was interested in taking a journey.

And and the journey would begin with, what is this thing called truth before AI and where is it going? And would end gently saying there's some forks in the road we're coming up to. And we should pay attention to those because if you know, when you get to the driving along and you get to the fork in the road and you just flip a coin, whatever, you know, those your life changes in dramatic ways, depending on whether you go left or right. And I think that readers I'm hoping will say to me, I read the book and I made some choices. And I don't want to give people an answer to what the right choice is.

Yeah, well, let's look at some of the issues because what immediately springs to mind is the power of generative AI, certainly with images and creating believable text around it. That's one level, but it is an entirely new level as video that's generated by AI wholesale. Often is completely indistinguishable from actual video from real life. And part of what I find interesting with this is that the vast majority of people believe that they can differentiate and identify AI generated content whenever they see it. And yet several universities have run studies where they show AI generated faces back mixed in with real faces, real pictures taken of human beings.

And the success rate of people to figure out which ones are fake and which ones are real is is slightly better than random guests. It might as well be an absolute guess in the dark, which tells me that people are not nearly as good at identifying AI generated content as they think they are. And the video products get better and better every day, which seems to be a clear and present threat right now to what we see as truth. I think table stakes are anyone who can determine fake from real today. Let's assume that six months from now, that skill will be gone. And a couple of examples. The Warner ran, there were videos of missiles hitting a skyscraper in Saudi Arabia.

Completely fake. Completely fake. There are other videos on TikTok of dogs on diving boards. And if you read the comments, they're fascinating, because the comments go like this. Are this real? Do dogs really dive off diving boards? To which somebody says, I think it's AI. And then somebody else says, I don't care. It's lovely. Because I mean, at the end of the day, you don't need to know why their dogs are on diving boards in order to find the image of a great Dane doing a backflip charming, because it's entertainment, right? So the war video is not entertainment. And the fact that there's fake information out there, really bad.

And just by way of example, the shooting that took place at the National Correspondents' Dinner. So I was researching this this morning to see if that one image of the shooter with his shirt off on the ground was AI or not. And there's a very detailed article that says it's fake. Wow. The article is in the New York Post. And there's no other article in any news organization that backs that up. So my spidey sense is someone that's been a career in the news business is, if that were really true, you would see it in the Wall Street Journal, the New York Times, writers, some other places. So I'm saying at this moment, I think the post is wrong.

But that's a lot. What I went through to get to that is a lot of work. So most news consumers just kind of go, could be fake. Right. And that's a dangerous place to be, where I think you make this argument in the book, that it's one thing if we don't believe in certain things we see. But it is an entirely different issue if suddenly we're not really ever sure what's real, what's not, and can't, just on a subtle level, can't trust anything. It really puts civil society and democracy at risk. Yeah. And that's, well, let's put politics aside for a minute. Sure. I mean, Sam Altman, who is not a dummy, however you feel about Sam Altman, made the decision to terminate Sora II.

Yes. Sora II was amazing as a technology. I could say, you know, Nick, here's a picture of Nick, and I want Nick to be making a video about why my book is amazing. And Sora II would literally spin it up. And when you've stumbled into it on TikTok, you'd be like, wait, I didn't say that. I didn't say any of those words. And you'd send, I guess, a cease and desist to TikTok, which would say, at that point, section 230, free speech, tough luck. So the fact that Sam understood that A, there was no money to be made in manufacturing fake video, and it put them at risk and they pulled it down. Was I think it's not going to solve our problem.

Someone else, you know, there's plenty of other sites today where you can make really compelling fake video. But it's not going to go away. No. And Sora is now taken down so people can't go out there and try it. I was very active on it when it first came out. I mean, you don't understand what it was capable of unless you tried it. And this used it. And started doing stuff with it because what really struck me, I mean, I have experimented with a lot of these platforms and grok, for instance, when I have grok create a video of me, and I will give it the tools to work with, it always gets the voice wrong.

Whereas Sora, I mean, it looked like I was in this studio right here, the voice is indistinguishable. There aren't, there were not weird artifacts with six fingers or something like that, where suddenly you could tell it was faked. I mean, it really, really fooled people. And that existed, it's going to pop up in another form. I mean, there's just no getting around that. But here's where I think your audience has power, which is to say, you know, if you go to the supermarket and you're on a lactose-free diet or gluten-free diet and you read the ingredients in a food product, and it says gluten-free, you take it home, and it makes you terribly sick.

And the manufacturer says, yeah, we just made up the numbers on the side of the food. We really don't. Like, we don't feel the need like some number of our customers are just going to get sick and we really don't care. You don't, not only do you never buy that again, but you probably reach out to the FTC or the FDA, if you're, you know, presuming that there's anybody left at the store. I think we as consumers aren't going to get wise very fast about saying to different AI platforms. I want to know what your policy on truth is. And I'd like to have you just explain to me how you process a question that has a truthful answer.

And my experience, particularly with chat GPT, not as much with anthropic at the moment, but is that when it's wrong, it's so gleefully wrong, so charmingly wrong. And if you think that you've taught it something by saying, no, no, no, you're wrong on that fact. And here's a link where I can prove, if you think you're actually making the underlying database smarter, you're not. Because it's not using when you're those corrections aren't being fed into the larger database. No reason for that, technically, could be. Well, sure. And it's so confident when it is wrong. And I think, first of all, it's rare that somebody is active enough and is asking about something where they have enough expertise to first identify that maybe there's an underlying assumption from chat GPT that's wrong and to call that out.

And then you have the, so most people aren't going to do that. And then the second piece that you identified there is that, yeah, we can't change the model with those corrections because of the constraints. I mean, to be fair to the creators of these models, they can't let just anybody pump in corrections because people would put in false information. And we that opens up a whole other can of worms. Oh, but you're asking a great question. So I just came back from a week in Vancouver at the Ted conference where the CEO of Reddit stood on stage and said with great pride that they're feeding all of these models Reddit data.

Well, Reddit Reddit says unfact check as it comes. Yes. So, you know, I mean, I'd like to think that in the future, I could use an LLM in which people gain credibility by participating. And when I correct something, you know, Wikipedia, yes, is built that way. I mean, you can the corrections on Wikipedia are awesome. And, you know, if you have something wrong on your site about Wikipedia, having been there once, not very recently, it's a huge pain in the ass to get it fixed as it should be. So, so, and if you have incorrect information to Wikipedia, you get flagged and then you're not able to do that punished.

Yes. Yeah, yes, it punished. So, so part of what I want our audience to feel the optimism that I share is I don't think this, I think we've gotten to where we are very fast. And I think there's some great big kind of societal and philosophical questions to talk about. And part of the joy of writing the book is, you know, I mean, you get to talk to like, you know, I mean, I've been friends through NYU with David Chalmers for a long time. But doing an hour interview with him about, you know, whether or not we live in the Matrix is just delightful. And by the end, I was disturbed enough to say to him, how do I know you're not an AI?

I mean, and, you know, there's that moment we are, we're all on the verge of facing where we talk to a customer service rep or a digital person on a Zoom call. And we find out that that caricature on the other side is in fact a robot. You know, right. Well, and it's interesting to me being in this space because people have flocked to podcasts over prior to GPT coming on the scene. The ongoing theory is that because they are much less produced, it's harder to just edit something to get it just perfect when somebody's gonna chat for an hour or two hours. And you really do get a sense of where they're really coming from.

And the feeling from people is that there's more truth in that. And that is the drive towards podcasts, whether you're watching them, listening to them or however you consume them. And it is not the most comforting thought to think that suddenly the if the video version starts to get faked, then we would lose that truth. Even if it's there's certainly plenty of errors and inaccuracies that are spewed on podcasts all the time. But we're trying in general like moving towards more truth, at least knowing where somebody's coming from. And I, yeah, I don't like the idea of video stepping in there and a fake version, fake conversations that are produced by companies that have real biases and their own separate incentives.

So the book has a whole series of different journeys and interviews with amazingly interesting people that are pretty diverse. And part of my advice to readers is if you get to a chap, I mean, it builds to a conclusion at the end. But if you get to a chapter and you say, you know, I'm just not interested in the future of work, I mean, Andrew Yang's great and all, but I've heard everything else. I mean, you know, no points off for skipping a chapter. Right. But I do think that the larger question has to do with the confidence that these algorithms have, and increasingly, I think as a society, our willingness to accept their numbers as truth.

And I'll just give you an example. So let's say that Nick has decided he's going to go out and buy a house. And to buy a house, he goes into one of the three or four big companies that do credit scoring. And he types in the signs up and he types it as the name and he says to experience, what's my credit score? And Experian comes back and says 520. And you say 520. I mean, I pay my credit car bills and I didn't go bankrupt. And I kind of think I'm more like 780. How would you debate that with Experian? Well, with Experian, I believe I don't know I've actually don't know I've never been in that situation.

Their algorithm is is invisible to you. How they came up with 520. Could it what you know, intuitively, it's it's probably an error. Yeah, it's or it's a credit card that you lost track of that you owe $3 on that's kicking back as unpaid and whatever. But but but there's no human in I you know, I tried to get off of an Experian product. I just couldn't get anyone on the phone. Yeah. Well, I wasn't 520. But but but the same thing happens, you know, particularly during COVID schools race to use these algorithmic. Are you cheating on this test product? And they did all these things where they watched your eyes and they looked to see if you were looking off camera on where you're checking your phone.

And they then came back and said to the schools, these remote tests we found to be fraudulent or failed. And students, there's a bunch of big cases where students were told, I'm sorry, but you failed that test because this third party service we use determines you were cheating. And and and what the school administrators said in in the reporting was, you know, when when the robot says, your flag is cheating. You're kind of like says cheating next year name. Yeah. And those and so increasingly machines are labeling humans, whether it's an and and, you know, I'll just give you one other example of kind of where we're heading if we blindly trust the algorithms.

I was driving up state New York New York a couple of months ago. And as I was driving Google Maps, kept putting on the map a dollar store logos of the dollars I've never been to a dollar store. Yeah, not a customer. Why was it putting on the maps dollar store dollar store dollar store? So I got to the party I was going to found out that they needed paper plates and napkins and found myself at the dollar store. Yeah. Well, I wasn't paying Google for Google Maps. Dollar store was clearly paying Google for Google Maps. And that little bit of advertising was exceptionally effective. Right. You know, so so what the what I what I found myself coming to is that there are all these places in our lives right now where these platforms are already driving our choices and increasingly, I think we're going to find it's harder and harder to opt out of those things.

Yeah. I mean, the Google example makes me think about where these AI companies are at at present and having talked to a few economists about it. They said, you know, there's so much VC money coming in, historically incredible amounts of VC money. And so they can make the models cheap for people to use and try to really whether they're always successful or not is up for debate, but try to make them helpful to people so that they will use them. But there is a point in the future where they have to start making money somehow to repay all this investment. And the choices they face are hugely raising those monthly subscriptions so that everyone would not be able to use it, which creates a kind of have and have not access to these intelligences or some form of advertising in the way that you just just described with Google Maps.

And that creates all these incentives that aren't aligned necessarily with our goals or what we think is even going on with the tools, because it wouldn't take much to subtly, subtly start slipping in things like the dollar store into chat GPT results. And that's a, that's a very slippery slope when you're dealing with a private organization. It's scarier for two reasons. One, because advertising at least is annoying. You see it and it gets in the way of the thing you're trying to get to. But the point of which the answers are are proprietary to the funder. And they feel no obligation to tell you we've been paid to tell you this answer, then all of a sudden, we've abandoned Google with all of its flaws.

And we've replaced it with this algorithmic thing that says without absolute certainty, we're not going to show you the choices. We're just going to give you the answer. And the answer is someone who paid us to tell you that. And the perfect example of that is, you know, as a lifelong New Yorker, I remember when Uber arrived in New York. And, you know, just from the very beginning, it was so delightful, you could push a button on your phone and a car would be in front of your apartment, then you could get in this very comfortable car, it would, it was clean and they wouldn't talk to you and you'd have to put a credit card down and, and the drive to wherever you were going with $9.

Right. And you knew as that happened, you're kind of like, something's not right here. Like the cab is $35 or $22. Like, like, it was so much cheaper. And, and now I was in a, I took an Uber the other day from my apartment downtown and it was $57. Right. Because once they put everybody else out of business and damaged the taxi cab business, then they were going to charge you not just what it costs, but a premium. And so the coming apocalypse of cost on, on AI is absolutely, you're absolutely right to wave a flag about that. And the thing about the investors is, and I say this, you can beat me if you like, I mean, they just don't give a shit.

They truly don't. Sure. I mean, I mean, they're the same VC companies that backed Uber. There's, there's a direct line there and a lot of tech firms that have followed that same model, which is to flood the market really cheaply because there's so much capital available, and then put people out of business. And now we own the market and we raise the price astronomically. And what's anybody going to do about it? I mean, and I'm the only answer to that might be an IPO. I mean, maybe an IPO gets enough money in to pay off the investors. So maybe, maybe. But, you know, but that only happens once. I don't think you're going to see, you know, I think it's a race between anthropic and open AI to see who, who gets IPO first.

But, but, but let me talk about like some of the people in the book that are delightful and interesting. It's a whole chapter in the book about the future of news. And I spent a lot of time with Esther Dyson. And Esther is a early web one oh, you know, visionary. And what ship this is exactly what we were just talking about as. I mean, she talks about just paying and paying attention to incentives, which is, you know, understand how these companies are being paid. And, you know, and she's very, you know, it's fair to say, you know, the news business has never been pristine. There's always been yellow journalism.

There's always been the national enquirer. There's always, but, but the difference is, and this gets down to young people, the difference is that you knew the difference between holding the New York Times and your handing holding the national enquirer. So if some celebrity had an alien baby and you bought that at the supermarket, you knew where to put that in your news consumption brain. What's what's changed now is that, for example, TikTok, I mean, you don't get to make any choices about what channel you're watching, or, you know, whether or not you want entertainment or news or information. And it makes a purposeful excuse of blurring those things together.

And, you know, the debate among young people is, are they better consumers of news because they come to it with a certain cynicism and some suspicion than we did? Or are they being overwhelmed by misinformation? I think the jury's out on that. This episode of The Next Daily Show is brought to you by Zapier. If you've ever felt buried in repetitive work, copying data, moving files, sending follow-ups, you know it's like death by a thousand mouse clicks. Zapier has always been the tool that fixes that. It connects over 8,000 apps, Google Drive, Slack, Notion, Gmail, MySpace, you name it, so your tools can finally play nice together.

But here's the big shift. Zapier now lets you create AI agents with their chat GPT integration. Think of them as tireless teammates who never complain, never take lunch, never get bored of doing the boring stuff. I've actually made several of these little agents myself. No Python, no JavaScript, just pure vibe coding, which is my favorite kind of coding. Because even though I have no idea how to code, I just talk to the AI, tell it what I want, and almost magically, it works. For example, when a podcast guest books a time, Zapier can send them a personalized email from me, preparing them for the show, create a draft of show notes, update my calendar, and even prep a social media post automatically, all of which frees me up to focus on the important stuff, like taking credit for the amazing work my agents did.

And now, Zapier just took it up another level for developers. Zapier MCP is available right inside chat GPT, which means you can connect those 8000 plus apps and trigger workflows just by writing what you want to happen. Literally, you tell chat GPT, send this file to my team in Slack, update the spreadsheet, and draft an email, and Zapier plus chat GPT, figure out the right tools, do it for you. Getting started is simple. Head to the Zapier chat GPT MCP server and add the tools you want chat GPT to access. Follow the steps in the connect tab. And if you're an admin on a chat GPT enterprise account, you can enable MCP across your entire workspace.

So if you're ready to stop wasting time on busy work, join the AI revolution and make a little automation magic of your own. Try the Zapier chat GPT integration using the link below. I want to turn slightly because one of the things I haven't, we haven't highlighted just yet is that the book is not all negative and how all this can go wrong. I thought you did a wonderful job of balancing the two things on the upside of all these new technologies versus the downside and that there are different places that we could go, whether it's law or medicine or the arts or education. Let's, let's talk about that dichotomy a little bit, because sometimes it's hard for people to hold those two thoughts in mind at once, right?

They get stuck on that society is going to be damaged by AI or they get stuck on that. Wow, these tools are really helpful to make me more effective at work and it's really wonderful. But both things can be true at once. And so I just wanted to dive into that gray area just a little bit. Yeah, so I'm not a led by any means. I use technology all day long and own every device you could imagine and I spent my whole career buying and building tech companies. I think the question, I guess I'll give you a couple of examples. So I have a friend that is a senior partner in a very large law firm and he said to me, one of the things we're thinking about is how do we train junior associates when the work that they would ordinarily do would be checking periods and commas on 500 page documents, which is kind of sludge work, but gets them into the law firm and they get to meet partners and they get to know and then they grow up and become lawyers.

But if there's no entry point because the entry point is now a machine, what do you do with interns? What do you do with summer interns? And I'm like, hmm, that's interesting to me. The other thing is and I think a lot of this comes down to speed, which is if AI really does put tens of thousands or millions of people out of work. And the idea is, I mean, you're not going to retrain truck drivers if if trucks become autonomous vehicles, you're not going to retrain truck drivers to be ballerinas. You know, sure. So the Andrew Yang answer is, you know, basic income, UBI, and that's an interesting philosophical debate.

We touched that on that in the book. But I guess the thing I don't know and I think about a lot when it comes to art is do consumers really value human made things over machine made things. If I go to the movies and I see a film and the actors are all AI, but I find the storyline and entertaining and I get to go to the movie with my wife and we buy a cop car and have a nice evening. Is that does the label on the film that says human actors, human actors? Am I gonna other than the small number of political people who are going to say, I want to support humans? You know, when you go to the supermarket, you don't ask the supermarket where the food comes from who picked it, who manufactured it, how much is onshore food?

I think that there are things AI is going to do that will make the world dramatically better quickly, magically. And that's the metaphor of the book is I began my first career as a magician. And so we talk about the power of magic and people liking being fooled and all of these things. I think that things being more efficient and cheaper will feel appealing. And things feeling sloppy and manufactured will not be. But here's where I there's a lot of thoughts about this. I write a lot. I write every week. I write a blog post. I write every day. I've written three books. For AI to be able to consume Steve Rosenbaum written material is relatively easy for it.

So when I have an idea for a blog post and I sketch out a article and then I put it in chat GPT and ask it to clean it up and make it sound like me and not correct it in the way it wants to correct things. That works perfectly well. If I'm 17 years old and I say I have to write a paper about this and it doesn't have any of my voice in its database as a young writer because I don't have a body of work yet. How do I get to develop my writer's voice at 17 years old when it's just so crazy easy to have the robot do it? Well, you've touched on a number of interesting examples there. No, it's okay. So if we start with art and I don't think there are easy answers there because you mentioned a movie that's completely AI generated.

Well, movies are already a complete fabrication. They're inventing a fictional story. They're creating, I mean, if we go back a little bit, they're building sets that don't exist in real life to make it seem like they do in this piece of art that we're watching. And then as special effects have grown, there's been more CGI involved. And as long as the storytelling is strong, people still seem to respond to that. I mean, if you ask people, I find, and this is just anecdotal, this isn't research based. But when I talk to people, they certainly say, no, I wouldn't want to watch something that's just completely AI generated.

And yet, as long as the art is compelling, the story is compelling and tells something that speaks to what it means to be a human, I do think people would show up to watch that just as they do when there's a lot of CGI involved as long as they can still have that human experience where they go, they watch it, consume it with people that they know and like. And it's a fuzzy line, but it does seem like these using AI as tools through the guidance of the human artist is where we would find the most artwork that will resonate with humans. At least that is my hope for the future. So you said something important, and I just want to make sure I don't miss it.

So I'm the executive director of the Sustainable Media Center, which is an intergenerational organization that has a board of Gen Z leaders and an X plus one board as well, people have a certain age. More and more of the Gen Z community is shifting toward in person gatherings, and putting the phone down and being with friends and really valuing human connection, which I think is great. So I don't think, I mean, it's worth saying out loud that COVID, there was a moment, six months into COVID, where I remember thinking, I'm never going to see people ever again. I'm in my apartment, I'm on zoom, I'm wearing a mask when I go out, I'm eating dinner, sitting at a table outside, just cold.

I mean, there wasn't any evidence or confidence that we would ever be back in person at conferences or in hotels or on airplanes. And I'm, you know, so we already had this near near-miss experience of what it would be like to be fully digital. Yeah, I don't think anybody liked it. Sure. I mean, I did for six months. I thought it was great. I got a lot done, it was super efficient. And then I remember thinking, yeah, no, this is terrible. I missed just people deeply, being around groups of people deeply. Yeah, I tell people I like people in 3D. I mean, I like I like you fine. But when we meet in person, you'll be in 3D and I will like you better.

Yeah, sure. I think we've already, I mean, part of, I guess, what I want listeners to hear is, we have a moment now to say to these companies, no, I prefer to have human experiences. And what I worry about deeply is that we are now so overwhelmed with digital information and inbound inputs that AI will come along and say, listen, let me take that off your hands. I'll book your vacation. I'll schedule your restaurants. I'll, you know, I'll go to your kids school meeting for you. Like all this stuff that's overwhelming you, I'll do, I'll be your personal robot. I know what you like. I know your morals and your values.

I can, I can, hey, there's a new book coming out. Steve Rosemary, you should, I bought it for you. Yeah. 30 day return. And that for a moment as humans, we will go, boy, this AI thing makes life so much better. And by the time we realize how much we've given up, it'll be too late. Yeah. Well, let's, let's touch on the writing example that you mentioned with the 17 year old. So I think that is a really interesting touch point that a lot of parents can certainly identify with. I know I can, where I do have a lot of writing that I have been able to upload to the models. And when I want to quickly produce something, just like you mentioned, I can say, Hey, I don't want you to change the heart of this.

And I want it to make sure it stays in my voice, not yours. And you have enough information to do that. And it will help me get to the finished product more quickly. But they are those ideas that I am generating are cleaned up by the model. And it makes me more efficient, more productive. It actually frees up more time. I'm spending, let's say 20 minutes editing instead of two hours editing. And that is time that I can go spend with loved ones and it makes life feel better. At the same time, having children who are growing up, there is an incredible temptation there to have it, the models, do the writing for them, because it's so easy.

And it's hard work learning to write and finding your voice and wrestling with ideas. And who wants to spend 30 minutes figuring out if what is the exact word in the sentence that will communicate what I'm trying to say. And yet that's the stuff that teaches you how to think critically, how to communicate effectively. And that is a real danger for people that they're not entering into this in mid life, but they're growing up with it. Thoughts on that conundrum. So so I have great sympathy to this, where parents are in this moment, because there's no right answer. I mean, all my friend Jonathan Heit wrote the book, The Anxious Generation, which is selling like hotcakes because every parent is like, please just tell me what to do.

Put the phone in a drawer, like set limits, whatever. And similarly, if you're a parent and you say to your kids, don't use AI, it's bad. Then they're not being prepared for the future. Right. If you say to your kids, and the schools are all over the place on this stuff. You know, I mean, my younger son is a middle school teacher. And you know, do you use AI to read? So when students submit to you, do you use AI to read student papers? You know, you're grading 180 student papers, you feed it all into AI and say, you know, give me some guidance about which students need help. You know, and say to the AI, let me know if students are submitting AI written material because they are.

Sure. And the moral panic folks say, look, they said the same thing about, you know, Wikipedia, don't use Wikipedia at school, right? Said the same thing about calculators, you'll never learn math. All I'll say about the Wikipedia or calculator example is my research says that this moment in time is different. Yes. But I also think part of what's challenging is if you're intellectually stimulated, AI can make you better. What I worry about, and this has to do with the base programming, not the technology, it never says you got this wrong. It never says, I don't have a good answer for that question.

It seems so hardwired in favor of being ubiquitous. And I find my interactions with it pretty confrontational. I often think of it as a talented, but inexperienced intern. Because it says with, I mean, it writes these very bold shaped phrases that sometimes you have to, I mean, you can't scan it, you have to actually read it. Because sometimes you're like, holy shit, that's just wrong. And then strangely the other day I googled, I looked on YouTube. I don't remember why, but I was searching for my name, I guess I was looking for reviews of the book. And someone had put up a whole career journey of my life, produced by AI that Google, or the Microsoft thing that pretends to be a podcast with two fake hosts.

And they created a graphic of all my work going back to MTV through today, and how it was all related. And it was clearly robotic and super relevant. Like, yeah, I've never seen those pieces fit together that way. Right. So the big philosophical question is, if you take the day to day battle of should teachers do this, should students, parents, then you say, where are we going to be in 10 years? I don't mean to be a bummer. We could be less human. Yeah. Yeah. Yeah. You said it a minute ago, because the hard part, I mean, you know, so I went through months and months of months of editing on my book.

And getting notes back from the publisher and debating whether this was clearer, this was clearer, we're going to take truth with the capital T, and truth with a small T, and the hairs how we're going to use the capital T, and here's what we're really intellectually demanding, hard, frustrating stuff. Now, I'm sure there are mistakes in the book. I hope readers don't start sending me notes and saying you misspelled this. I don't think they will. But like, happily, there was no AI to replace the battle. It wasn't a battle with my editor, but it was, you know, there was friction. I would argue that friction was healthy.

But we're not very far away from, you know, when my, when a publisher, I won't say my publisher, who I love, Matt Holt at Ben Bella, but like when a publisher says, you know, we can replace this editor role with AI, and we'll save, you know, a salary of $64,000 works fine. That's a tricky moment. That is a tricky moment when that happens. And those publishers are then in a sticky spot, the other ones that haven't done it, because it costs them more to do the same work. And so, there's pressure on them to take the same steps in order to just be competitive and viable as a company, whether it's the right move or not.

Something that popped into mind. Yeah, we'll go ahead. Go ahead. I'll give you an example. Filthy is charged. I've been in the video business a long time. So as I'm putting together promos for the book, I know that I want a shot of the book cover. I want it to float back. I want to land on its spine and then open up with pages. And then I want words from the book to pop up above with a narrator. Ordinarily, recently, I would go on to Fiverr or Craigslist, and I would say, I'm looking for this and a bunch of people would pitch me. And then I'd look at their work, and then I'd give usually two or three of them the same assignment, little entry point assignment.

And one of them would flail on me in Vanish. Another one would do an okay job. And a third one would be really good. And I would hire them and they would become work partners with me. And I would pay them hundreds of dollars or thousands of dollars, but not. So this time, I said, I ought to eat my own dog food. So I'll use AI to build this promo. And I used Adobe Firefly, which I found to be completely ineffectual. And I gave it notes, which you'd ignored. And, you know, it, it, nobody gets an A for running a good promo to get me to try it. Just didn't work. But I ended up using chat GPT. And it produced a perfectly good it.

So, so those three or four people I would have interacted with those humans are not in my pantheon any longer. I mean, I will use them again for other things. But I mean, that's like a real world job that got replaced by something I paid $20 a month. Yeah. Yeah. The flip side of that, though, is that because I've been in that exact same spot, trying to describe with words some sort of visual I want created, whether it's a static image or a video or whatever it is. And it can be extremely helpful to generate a rough draft on one of these models and go, I need it to look like this. But I want these seven details changed.

Can you go do that? And suddenly that makes the process better. The flip side, the risk is exactly what you stated, which is that suddenly the model gets so good at it that you don't need that person anymore. And then they're out of the loop. And so you are more effective as an individual. But it's really gosh, it's raising the bar on the skills that everybody needs to have to be an effective economic machine on their own. And I think even with the school example, some of them popped into my head when you were talking about it, the there's this thing called the DBQ, which is it's a certain question type and it's only taught in AP history classes.

And I've always thought it was a wonderful tool that should be accessible to everyone, whether they're taking an AP class or not. Because what they do is you're given three to four source documents and then asked a question about conclusions you can draw from these documents. And often the there's conflicting evidence in the documents. And what the student is asked to do is to analyze the assumptions of the authors and what their perspective is where they might be coming from and why they would have phrased something exactly how they did. And then try to tease out what we could confidently expect to be true from this combination of documents.

And I feel like skills like that are going to be ever more critical in education and really what we should be pushing towards kids. So they're not taking those answers from chat GPT at face value and are pushing back and asking it questions to back up its conclusions. Someone ought to build a version of an AI that is more of an inquisition. Well, to be fair, I've had it do it to me a couple of times. I'll say design this thing for me. And it'll say, well, here's the description of what you want. Go to Canva and use these tools to build it. And my reaction generally is, no, you go to Canva. Oh, well, we don't really have a deal with Canva yet.

Well, sort that out. Yeah, because, you know, I mean, I would love if it would say, here's the layered version of the graphic you want. It's in Canva, you can now go and adjust colors and shape and style. But like, there's definitely a battle going on right now between the humans and the robots about what we want it to do. And when it says no, I mean, I take great umbrage to know. Yes. I've had that same thing. Also, there's the same thing where it'll say, oh, I've created this Microsoft Doc Word doc. You can just click here and like five times out of five what it delivers is just garbage. But it doesn't ever say, I've done the best I can, but I think you're going to be on hat.

Like, why it's not more honest, says the man who wrote a book about truth. Um, it not only is it often dishonest, but if you ask GPT or or anthropic, how it feels about truth, it, it doesn't take responsibility for it. Yes. Yes. I am with you on the on the building. It'll show you a quote, show you a quote of that will be completely fabricated right inside a bunch of other quotes that are accurate. Yeah. And we'll flag the one that it fabricated. That that's um, we should be worried about that. We should. I have done research work with it where it's citing let's say nine papers and two of them are completely invented, but they have the same MOS citations.

And unless you're going and clicking and reading each one to confirm that the research it brought back to you is true. It's you're talking a lot of work, a lot of intellectual investment in order to sort out and sift through what is hallucinated and what is our actual sources that it's drawing from. Uh, on an example though, um, because you mentioned creating a LLM that we could interact with that that pushes back more and that is more truth oriented. An interesting example, I do know a lot of people are trying to do things like that. And with, uh, we've been experimenting at test prep gurus with a individual GPT that is just made to discuss college admissions odds at different colleges, because one that we have all these kids that come in and they say, Oh gosh, you know, I, I fed my transcript and X, Y and Z into chat GPT and it told me I'm a great candidate for Stanford.

And because it's so gleefully wants to gleefully be helpful, right? And always affirming of whatever your goals might be. And it's very hard to walk a family off the ledge and say, well, you know, the acceptance rate at Stanford is three percent. And that includes all of these families that have donated money and the athletes that they want to recruit, it's probably less than 1% for a normal kid. So there's no one is just the shoe in at Stanford to begin with. Second of all, you don't have a perfect GPA or perfect standardized test scores. And, and that's okay. But it's not what Stanford is necessarily looking for.

And when we created the separate model that was more where we really had to wrestle with it and say, no, we need you to be truthful about the odds at a given college. It's been much more useful to families than using the generalized models. So I do think that is coming. And people are working on that in these specialized spaces. I don't know if it's a blanket solution to what we're talking about because that requires also requires work to go find these specialized models. But it is something that a lot of people are wrestling with. And I'm hopeful that we won't the models will move in that direction over time because people are pushing for that and want that.

I, you know, by the way, I love that story. And the marketing guy in me asks the question, do consumers want the euro you're going to get into Stanford version more than the actual like, I mean, that's a great question. You know, when you go into the store and you put on the jacket and the salesperson says that you look great in that mind, that's your color. I mean, nobody ever says to you, really, that's a shitty color on you take it off, which might might turn out to be entirely true. But the salesperson is getting paid on commission. So it depends on, I think the consumer and who that salesperson is.

So there are salespeople that be like, no, this thing would be better for you. And they're looking for a long term relationship. And then there's that person that is like, Oh, everything looks good on you, especially this $2,000 code that really looks good on you, Steve. You know, I have that in my closet. I brought that I bought that. Yeah. So here's the here's my hope of the book. Nobody wants to go buy a book that says AI is going to destroy the planet, we're all going to die. Nobody wants to buy a book that says we're all going to get rich and be powerful. If we put money into like what the book tries to do and what I've heard, I mean, the early blurbs tell me that I achieved at least some of it.

It's fun. It's a journey. I if there's some academia in it, but it's not an academic book. There's no test at the end. And I'm hopeful that people that start at the beginning and make their way through it will at the end say, you know what, I give a better sense of where we're going. And I have more feeling that I have control. And when I start to see, you know, one of these big brands I interact with, you know, tell me via AI how to solve a problem or buy a product, I'm going to understand that they have a, you know, as Esther says, there's an underlying incentive that they're operating under.

And pushing back is acceptable and maybe necessary. And while the book doesn't say it explicitly, I'll say it for your listeners, I don't think that politics is going to solve this. I don't think government's going to come in and build guardrails. I just, I mean, I've been in tech a long time. And the speed of technology relative to the speed of government is just like they're not in the same stratosphere. By the time the government writes a bill and tries to build in a guardrail, you know, chat GPT is going to be on version 12. And I just don't think that you can legislate us out of the box. Well, along those lines, before we went on air, we were talking a little bit about comparisons to social media and what we have learned from that over the last 20 years.

What can we learn from what has happened with social media and where this might be headed with AI? I should take some responsibility for this. So early in my career, I got my hands on one of the very first game quarters that had a reverse facing lens. And I had this epiphany that turning the lens around would mean I could photograph myself. And I pitched and sold in the room a show that became MTV unfiltered. And this was before YouTube existed. And we went on the air on MTV and we said, if you've got a story to tell, call this 800 number. And we answered the phone and we talked to young kids and we fed ex people cameras, we fed ex back tape.

And we were the very first user generated video project. And I remember vividly being excited in fact, we talked to the Fast Company magazine with great pride about democratizing news. And never did we sit down and say, you know, could anything bad happen here? Yeah, like, like, is there any out that we were democratizing news? How could that not be fabulous? And so I sit here today and see the spread of misinformation and racism and misogyny and all of this stuff. And I think, yeah, we didn't really run the play. Like, what if people that want to cause harm take our idea and take it in a different direction?

I don't know we would have done anything differently. But I wish we'd asked that question. And so social media in its early days was delightful. You got to see your high school friends and your college friends and their baby pictures and their pets and their vacations. What's not to love about that? And then it moved along. And a couple of the platforms said, what can we do to be more engaging? And Scott Callaway says very publicly engagement equals enragement. Yeah. And so, you know, I mean, I don't know how many viewers we have on this podcast. But presumably, if we scream to each other and call each other names, you would have more.

And and I'm not interested in doing that and you aren't either. And that's good for us both. But but social media has ended up being really harmful for young people. And until six months ago, we had, you know, we had Francis Hagen's early Facebook whistleblower data. But a lawsuit in Los Angeles that just got settled, settled that got one three weeks ago, basically was the jury saying, we believe the current state of social media is harmful to kids. And that's a really big deal. Yeah, I believe. I mean, I remember when cigarettes were prevalent in society. And when I tell young people, you know, there used to be a time when you could smoke cigarettes on airplanes.

They look at me and go, well, that's insane. Who would ever do that? Like, well, there's a smoking area in the back, like that's stupid. It was it's steel tube. Like, you know, so I think we're going to see social media have much more limits on how it treats young people. But it's going to take five or six years. Right. And so with that learning in my head, I look at, I look at AI and I say, you know, and my friend, Tristan Harris talks about this with great precision. We know what happens when you let tech do whatever it wants without any rules. And, you know, AI is in a position to be dramatically more dangerous.

This episode of the next family show is sponsored by granola. One of the worst things you can do during an important meeting is try to take notes during the conversation. You spend half the time typing, half the time listening, all the while not doing either one particularly very well. That's why I've been using granola. Granola connects with my calendar. So my upcoming meetings are already organized and ready to go. When the conversation starts, I can open the live meeting display and stay focused on the person I'm talking to granola captures the conversation and turns the transcription into structured notes that are very easy to use afterwards.

Instead of ending the meeting with a page of scattered fragments, I have the next steps decisions made and follow ups clearly laid out for me. I can also add my own notes and combine them with the context of the conversation for interviews, team meetings and strategy calls. It lets me stay present without losing the details that I'm going to need later. Learn more about granola using the link below. I'm a big fan of Tristan's as well. What does he suggest regular people do with all these changes coming with AI so that we don't wait 15 or 20 years for as it took with social media for these finally a court ruling to validate these concerns about all these negative effects associated with social media?

So Tristan's take is primarily legislated and his argument is compelling which is essentially that in a capitalist society commercial entities job is to make money and if you don't create rules around how money is made, I mean I jokingly will say to people I'll be out at a dinner and I'll say to one of my friends, listen I'm short on cash can you loan me 20 bucks? Then like sure and I'll pay you back 10x that in a week and then a week later I'm going to come back to you and say can you loan me another 20 bucks? I'll give you another 200 and at some point you're going to say to me after we've done this four or five times how are you doing that?

I buy and sell heroin so you give me 20 bucks I buy 20 dollars with a heroin I go out in the street I sell it and I give you back your 200 bucks well at that point that return on investment becomes unacceptable and I'm no longer you no longer I banker so I'm hopeful particularly in terms of social media that the people that are earning the profits from those 20 dollar loans are going to begin to see that money as tainted and similarly that some of the return on investment on AI will have some of that same pale float over it but but you know we're in a hurry to have that happen and here's the other thing too that I think you have to be careful about like the media has certain kind of simplistic answers to complex problems so the simplistic answer to AI is human in the loop right we need human in the loop so like we don't want you know autonomous weapons we want human in the loop and my answer to that is when I'm in my car and somebody jumps out the car jumps up the light and is cutting off my autonomous vehicle I do not want the car to have to ping someone you know sitting in the Philippines in front of a headset saying Steve's car needs the brakes pep pressed I want the AI to press the damn brakes sure thank you very much so I don't think human in the loop I think human in loop is simplistic and dangerous yeah because it's not nuanced enough to fast enough to handle all of the actual interactions with reality that this technology is going to have and I do think that's important to keep in mind and something that that you seem to keep front of mind as well which is that we don't we're not able to accurately predict the future with a crystal ball we don't know exactly where all these tools are going to go I mean I was just thinking the other day on how when I was doing an MBA and we're talking 20 years ago they said emphatically with all the confidence of a chat GPT hallucination that we were moving into an age where you would never receive a marketing message you didn't want anymore you would only receive the right message that you're looking for at the right moment you wanted it marketing was going to become everybody's best friend and 10 minutes on Instagram or TikTok and you're flooded with stuff that you're you don't want to see and so we clearly haven't reached that I was watching Netflix last night and I refused to pay that for the commercial free version and not only are the messages insanely irrelevant but they're repetitious as hell they're like claw your eye there I mean maybe the answer is they really want your extra six dollars a month maybe that's really it because they couldn't be any less effective right yes the whole permission marketing promise the Seth Godin permission marketing promise turned out to be a well-meaning failure yeah I mean I I'm a big fan of Seth's work and I think it there's a we run our company that way and it works for the people that gravitate towards that but as far as it working everywhere for everyone there's reality is too complex and I guess the broader point I'm trying to get out here is that with AI as it grows and changes and its interactions with reality there's no way to know exactly where all of this is going but it is really important to ask these questions just like you said with the self-facing camera to just try to think through and game out well what are some of the possible risks how do we get to the future that we want versus the one we don't want and this is happening really really fast which right raises the risk of entering into one of these features that we are less desirable simply because we're just going to let it run its course well and and let's also you know let's let's name the names I mean Sam Altman and Elon Musk both seem to be kind of ethically bereft I mean you know they and they're argument and and and some of the other AI leaders as well is essentially we're on this you know we're on a a space race version of it China's either going to win or we're going to win and we don't really know what it's going to take where it's going to take us and by the way might be really terrible right but but but we're still going to allow it to to go as fast as it possibly can with no guard relative you will and I mean it's worth wondering you know Sam is a human being and so he dreams at night presumably and wakes up in the morning with some things on his mind and the question is are those things educational are they societal or are they economic because if they're economic then you know and and we're as we record that they're in the midst of the Sam and Elon core case which is fascinating it's like you think one of those guys would have been smart enough to realize that getting into court is not a good this is not a good place for them to be but I guess though in the same way that you're thinking about I mean I love the story you told about trying to have an AI that answers the question of what my likelihood to get into that school is you know if every person building a new product is is thinking that way which is what does my customer need they don't need bullshit they need an actual answer then the technology goes in a really good direction yeah yeah well and to come full circle on what people can do is one I think voicing some of that where they where they can whether that be I mean if the LLMs are scraping Reddit making that voice heard on on Reddit and I had a chance to talk recently to Malo Bourgon who is the CEO of the machine intelligence research institute and he mentioned that there is real power in reaching out to local representatives because they don't receive much input from normal people on what their big concerns are and the more we can voice these concerns to representatives at least there will be an attempt to put guardrails in place to steer these technologies in a more positive place for the average person out there so I probably spend an hour a day I maybe I don't even know what to some amount of time a day on Chechi P.T.

and I'd say it's 50 50 annoyance to delight yeah and and the and the delight piece is complicated so like I just by way of as you've now figured out I mean I I tell stories right so so my brother's birthday was yesterday and every year I make him a birthday card and I go into Photoshop and I do something funny and I take some pictures and I piece them together and I you know it's probably an hours of work this year for the first time I told Chechi P.T. what I wanted the birthday card to look like and I fed it in and it delivered back in its first pass something that was pixel perfect the words the language the tone the color the funny my my personality and I had two immediate reactions which was one I just got an hour back in my day and two I feel a little guilty yeah because is it as heartfelt is it as much me if I mean I gave it a little bit of a prompt but not not enough to feel like I could really consider the authorship of the prompt a replacement for the hour eyes and by the way I mean no I didn't go get paper and pencil and cut paper out and you know I it wasn't a the old things that I did weren't crafted you know I wasn't yarn and and cardboard but I did feel authorship and in this case I didn't feel authorship and that leaves me a little trouble because I'm an author I was creative and so like if you're the person that makes candles and sells them on Etsy and you make each one by hand and it's beautiful and then you figure out that you're selling so many you can't really make them all and you go on to the web and find a Chinese manufacturer where you can buy them by the crate load and it ships them to you and you then wrap it in your own you know cray paper and put a little ribbon on it make it look like it's Steve's candle shop but really now it's Steve buying mass produced candles from China and packaging them up I think not being a crafts person is a is a worthwhile concern particularly if you get joy from making things crafting things absolutely I mean for my fourth book if I say to chat tpt you have one or write a book about podcasting and here are my five chapters and go ahead and write it and it does which it could and will you know I don't think my book agent would like that right well and and you probably wouldn't like that I mean on on that note just one last story I did just as a experiment the other day so I had all the chapters in a book ready to go except for the last one and I thought you know I'm just gonna give chat GPT I want you to discuss these six elements in the final chapter go write it and what came back was absolute dribble like it repeated a lot of stuff from before it didn't really make sense and it can't do that just yet but that moment is coming and I don't think that's going to be a great place for us to be because the your book is your book is a strong book because of what I am paying for when I buy your book are all the hours you spent getting the capital T truth versus the lowercase T truth correct for all the thinking that went into it the interviews that you did for your the totality of your experience and I do think that's going to be a tricky place to navigate when it can produce it in an instant I still don't think it's going to be as good as what Steve can produce even though it's going to take a lot longer for him to do it so what AI does is really good creation looking backwards right so if you're trying to summarize a thing that already exists it can do that yeah if you're trying to look forward and say here's the future and I'm I'm going to propose it based on some things about the past it's going to do that poorly and you know not to put too much of a button on the conversation but I mean the future of truth is human and the robots can help and we should use them but at the point at which we start abdicating our responsibility to figure stuff out and invent new things and make different kinds of songs and different kinds of art be politicians that are super tall or super short or super like whatever conservative liberal whatever they are like like the human part of being on the planet is not going to be replaced by chai chi pita and I do think special purpose robots yours for you know I mean the thing that's interesting when you say you can reach customers is yes because you know everyone's birthday you can go into the database you can say they're going to be applying for college a year from this date and I can pick zip codes where they're likely to apply to schools and they've got disposable income to pay for a like I mean I can I can I can sketch out your customer with precision and and the day after they apply to school they are no longer a customer yeah and you need not waste marketing money on trying to reach them it's the same thing with you know families that are that are pregnant like think you can sell them diapers and and cribs and you know all this baby stuff and then they have the baby and you don't need to sell them stuff in but they're all I think I would argue that for lots of marketers you know hamburgers hot dogs soda airplanes cars hard to know when someone's in the market for your thing but I think that is a nice button on the conversation where you said the future of truth is human and I very much hope that is where we are headed because that's that's the world I want to live in that's the world I want my kids to grow up in this has been an enormous amount of fun thank you so much thank you so much Steve Steve if anybody wants to find the you online where's the best place to find you future of truth.com perfect okay I'm I'm only on there about 75 times a day okay okay perfect perfect and there's going to be a group or building a truth group and so I didn't tell the story but I did an event that had a truth dinner and I don't know what I expected they asked me to put it together I figured it would be 15 people I walk into the room that it's set for 80 people and it was sold out and it was absolutely fascinating and they all were like you know that was a great night I was like there there is something about that word that I get to sit in the middle of and be kind of watching the circus go on around me that's really fun and I guess maybe that if I'm not trying to instigate anything it's that yeah yeah when you're doing good work and the the book is fantastic and highly recommend it to everybody out there appreciate it all right thanks Steve okay everybody until next time ask questions don't accept the status quo and be curious