How To Ai: Christopher Mims, Columnist at The Wall Street Journal & "Bold Names" host.

Speaker A: Justin.

Speaker B: I'm Justin Brady. And how do you AI? I know that sounds silly, but there's a purpose. AI is obviously disrupting everything. It's in every page, it's in every book now. But most observations about it are completely wrong, in my opinion. And in the opinion of today's guest, the author of how to AI, columnist of Wall Street Journal's Keywords, and of course, of course one of the co hosts of, uh, Wall Street Journal's Bold Names, Christopher Mims. Thank you for coming on the show, Justin.

Speaker A: Very excited to get to talk to you for however long we have here. It's a pleasure.

Speaker B: Only five hours.

Speaker A: Are we Joe Roganing this?

Speaker B: Yeah, I'm here for it. Um, toil. This is a word that comes up. I like to say humans inspire. AI perspires. I love how you talk about AI and its true role in our organizations. A lot of people think. Some people think, like, AI is a fad, believe it or not, there are some people that think that. Some other people think that AI is just going to disrupt everything. We're not going to work. Of course, you and I, I think, are aligned on this, that both of that is complete crap. Like neither one of those are even remotely true. But talk to me a little bit about toil, not replacement.

Speaker A: Yeah, toil is, you know, the work that we don't want to do and which people are getting smarter and smarter at doing with AI and that, you know, so in that scenario, AI is a supplementary abilities. And that contrasts with, um, you know, some of the kind of hype that's happened over the past couple of years where it's like, I'm going to hire a bunch of AIs instead of people. Or you've even had CEOs. You know, infamously, the head of Klarna was like, we laid off all these people and uh, we're, we're just going to replace some of the AIs. And that, you know, that hasn't worked out. You know, as Elon Musk famously said when he, um, kind of accidentally on purpose over automated Tesla's original production line and then had to walk it back, he said, and I quote, humans are overrated. Or, sorry, that's probably what he would say now. At the time he said humans are underrated. Um, but, but I think as we see knowledge work get automated, the lessons from robotics and automation in blue collar work, in um, you know, logistics and in manufacturing still apply. Right? This can enhance what humans do in the long run. It will fundamentally transform what jobs we're doing. But, you know, it is a supplement to our abilities, and in the short term we really can use it to, um, take on a lot of toil. And there's some great recent examples of that where you see the major AI labs leaning into that and frankly doing things where, you know, I feel, I'm. I'm feeling pretty happy. People have said, how could you write a book on an area that's changing so quickly? And, you know, I tried to write a book that was about where, you know, I was going to skate to where the puck is going to be. And so there's a lot of stuff in my book, how to AI, which I think frankly kind of predicted, where we are now with a lot of the tools that are available. And of course, the book hasn't even come out yet, so it's funny to say it predicted

Speaker B: so.

Speaker A: But anyway, I think I've got at least a year that the book is still relevant.

Speaker B: Yeah, yeah, uh, I've read it and hopefully it's like you said, hopefully it's still relevant when it comes out. It comes out in, uh, like a month or maybe.

Speaker A: Yeah, it comes out January 27th.

Speaker B: Right. Okay, cool. So we're recording this January 19th. For those that don't know, some podcasts recorded ahead of time. You mention in the book a lot, and you just mentioned it now that it's. Humans, uh, with AI are going to do a lot. They're going to get further. Humans with AI are going to be unmatched, un, um, like superpowered teams. But AIs by themselves if left alone. At least now and in the foreseeable future. Probably not the best situation.

Speaker A: Yeah. So, um, there are very basic tasks which you can fully outsource to AI. Now. I mean, my favorite example of this is people don't realize how many millions of invoices are, uh, processed every day. And every invoice, you know, is in a different format. And, uh, you know, no matter how good your OCR is, you know, doing the accounting exercise of reconciling what's on that invoice with your books, uh, you know, up to now has taken just enormous amounts of, uh, back office labor, some of which has already been outsourced. That is something that you can basically fully automate, except for some of the exception handling with AI today. Right. There are companies that you can go to. That's all they do. And so that's a great example of like, at the low end. Yes, absolutely. There are things that AI can fully replace kind of low Very low level human knowledge workers. You move up the sort of, uh, the stack of skills and, and it's a very different situation. AI makes people more productive, which is of course how in the short term it is hurting um, the job market, especially for entry level folks. Like I am hearing, you know, this isn't just like articles that we're reading. I mean I'm hearing this firsthand from people who are telling me, you know, we are not hiring entry level workers anymore, we are not hiring interns. Absolutely. Yeah. So this is something that, this a piece of reporting that I'm working on. So I don't want to reveal too much, but like I'm hearing this more and more and it does frighten me. Um, but the, but the mechanism is not that AI replaces humans, it's that it makes existing people, uh, more productive. So one of the things I talk about in the book is, you know, AI for like sales teams, for example, and there are a handful of tools now that uh, you know, they record every single sales call that the people in your organization make. And, and then they give people feedback about what's working and what's not. And uh, you know, I have been told by folks that those make their existing salespeople enough more productive. And this is measured of course in terms of like how much sales they're actually making, that they're like, we're growing but we're not going to hire any more salespeople.

Speaker B: Sure.

Speaker A: Because it allows, because of the AI dividend.

Speaker B: Right. Uh, it allows people, the existing people to be a lot more productive.

Speaker A: Yeah. And of course you see this encoding massively now. Absolutely tremendous encoding right now.

Speaker B: Yeah, absolutely. Um, there is like ah, an element to your book. This is a pattern that AI, uh, because I want to talk about job disruption, because this is the number one topic usually when AI comes up is everyone, everyone's going to, all their jobs are going to go away and robots are going to do all the jobs. And then of course when you ask anybody about their job, what can a robot do? Your job? Oh no, it's far too nuanced. Uh, oh, there's a, you know, it's like, which I think is an interesting reaction because people know their jobs pretty well. We don't know what goes into other people's jobs that well. So. And I want to talk to you about your own job being disrupted by AI. You talk about that in the book. But before we get there, one of the patterns is that AI organizations, uh, that embrace it, tend to create more value in the organization that really didn't exist without AI. And so would that also apply to jobs? Will we be seeing AI create a whole bunch of human jobs that we couldn't have foreseen will exist?

Speaker A: Yeah, absolutely. I mean, this is the thing about AI. It's both very much like every technological revolution that has come before, and then in some important and nuanced ways, very different. So the way that it's like every technological revolution that comes, that has come before is, you know, take the automobile, right? The automobile, uh, you know, disemployed all of the world's buggy whip makers. Right. And, uh, you know, it wasn't great for farriers, Right. Which is the term for the person who makes. And shoot, makes horseshoes and puts them on the horses. Right. Um, you know, it wasn't great for animal husbandry and that kind of thing. Um, but obviously, you know, at the dawn of the age of the automobile, nobody could imagine that, you know, in a, in like two, three generations, their offspring would be, you know, web developers or, you know, making apps or a million other things. SEO specialists.

Speaker B: Right.

Speaker A: Um, so, so that's just the nature of technological shifts. There are disruptions.

Speaker B: Right.

Speaker A: Like, the Luddites weren't wrong.

Speaker B: Right.

Speaker A: Like the, the, the, the automated mills were taking away their living. It was tremendously disruptive for them. Um, you know, uh, the way that AI is different, um, is that it is in some ways moving more quickly than past technological revolutions. And then also it's the first general purpose technology that can teach you how to use it. Right.

Speaker B: Oh, interesting.

Speaker A: Right? That's an Ethan Moloch, uh, Insight Moleck. Right. Who wrote CO Intelligence. Excellent book. Everybody should subscribe to his substack. If you, if you have a high tolerance for, like, you know, long disquisitions on the latest papers on productivity, which I do. I love it. I eat it up every day, every time he publishes. But yeah, like, you can literally ask AI, like, okay, well, what are some good uses for you? Or we've also all had the experience of going beyond like one or two questions back and forth, and it starts to reveal its, um, abilities to us, which of course are constantly expanding. And, you know, we can get more into depth than that. But, but, but anybody who's played with it for more than, you know, a half hour knows this.

Speaker B: Yeah. Um, one of the really interesting examples of this that you talk about, you talk about Clorox and some of the ways that they're using AI. Ah. Which most people are probably like a cleaning Company, they're using AI. You talk about a toilet bomb or something. Um, what's very interesting, though, I think their biggest use case proves this out on how AI will add more jobs, not take away jobs, is how they're using AI to predict and model future buying behavior so that they can sell more stuff and get more stuff on shelves, whereas beforehand they didn't really have a good technology to do that thing.

Speaker A: Yeah. So here we have to distinguish between generative and non generative AI, which I think is the terminology we're starting to settle on now. So generative AI is kind of just everything that came after the chat GPT moment, November 2022, when it, you know, came out and shook the world and everything. And then of course, also the, the other type of models which are different, the diffusion models that do, um, uh, image and video generation and are disrupting marketing and Hollywood now. Um, and it's kind of just a coincidence that those both came out at the same time, but that really baked everybody's noodles. But anyway, um, you know, Clorox is using pre generative AI, non generative AI to really help them with, um, their production and logistics planning, which is something that every company that has to manufacture stuff has to do. So they have to incorporate, uh, predictions about demand with, uh, knowledge about their ability to supply that demand. And, you know, typically this is a, an annual or a semiannual or even a quarterly process that companies go through where they try to match up all these different variables and make sure that they're making the right amount of stuff and they deliver it to the right place so that there isn't excess that wipes out their margins, et cetera. They have shifted to, uh, something that uses deep learning where, uh, you know, traditionally companies were not using that to help them with demand prediction. And, and, you know, I think one of the reasons that Clorox did this is they have really been on a roller coaster ride. They were, they like, they were huge during the pandemic. Everybody's like, let's buy Clorox Wipes, you know, because we're worried about cleaning every surface and our groceries and everything else. And then like, demand fell off a cliff and their stock, stock, uh, price crash. And so they kind of learned a bitter lesson from that. And now they're like, we need to be really good at demand production prediction. We need to be among the best. And, uh, you know, lots of folks are doing that. And I think it's a great example of how the vogue for generative AI, which makes AI Accessible to all of us, has kind of created a lane for the folks kind of in the back, um, to use every other kind of AI in other types of uh, human endeavor. Like planning for logistics, for example.

Speaker B: Yeah, um, logistics. And um, the planning and logistics are fascinating. But then I also want to, as far as careers, I also want to flip this and look at it the other way, which is, uh, industries that may reject AI and may not want to be disruptive, may not want to embrace new technology and will not get the benefits of what like Clorox is doing. You kind of cite an industry that may fall into this trap and that would be construction. With advances in technology, you'd think construction, with modeling and sophisticated technology, you'd think, think construction would be more efficient, but it's actually less efficient now than it was before. It hasn't embraced technology like it should. And I don't, I wonder, I do wonder. You don't indicate this in your book. I'm not putting words in your mouth at all. But I do wonder if their failure to embrace technology is that one of the reasons they can't fill jobs and so does an industry's rejection of AI. Is that kind of the self fulfilling prophecy? Like if you don't adapt, if you don't embrace AI, you, you are going to be the one that has your jobs displaced?

Speaker A: Yes.

Speaker B: It's a big question.

Speaker A: Yeah.

Speaker B: I mean, tell us with your crystal ball.

Speaker A: Yeah, here's just to take a step back and then we'll talk specifically about construction. I think the thing that I had, uh, uh, uh, uh, a strong sense of when I was writing the book, but now I have a very strong conviction about was if you don't embrace AI, ultimately somebody who does is going to displace you. They're going to take your job. Not directly. Okay. That's not how economies work. But the firm that you work at, for example, might fail because they have not taken advantage of the productivity gains that AI allows. And so they're going to be out competed by a firm that does. And that is the way in which somebody who uses AI is going to take your job. I think that's really, really important. And I think that like it's, it's not a very happy or feel good message, but it is something that I personally feel a great deal of urgency about spreading and, and I think it's one of the sort of core messages of the book. Like this is how you kind of begin to adopt the mindset so you can begin to use these tools and the Good news is it doesn't take a lot. Right? You can, you can, there are fairly big productivity gains to be achieved with fairly small tweaks to like how you use AI in your day to day work life. Um, and, and frankly we are still at the early adopter stage. Like we're not at the part of the adoption curve where it's like the, the early majority or certainly not the late majority. Um, so there's a lot of low hanging fruit. That's the good news. Um, but you know, construction is a great example of industries where it has been extremely difficult. Not just AI, uh, but his, but, but really since the 1970s, which is when production, productivity and construction started to go down, it's been an industry where it's extremely difficult to adopt, uh, automation or even digitization. And, and, and not for lack of trying. Right. Like we've had multiple big startups fail that we're like, we're going to just make buildings or parts of buildings in factories and prefab is going to be the solution. There are a lot of other, you know, challenges here. Like it's a very over regulated industry. If you ever want to go down an Internet rabbit hole.

Speaker B: Good point.

Speaker A: Just look on YouTube about like why uh, elevators cost four times as much in North America as they do in Europe. Like when you are done going down that rabbit hole, you will be like, I want to, I don't care where you are in the political spectrum. You be like, I want to torch every single one of these regulations because the end result is something that was designed to make buildings more accessible. Uh, you know, to people in wheelchairs has had the opposite effect. And now we have fewer buildings, wheelchairs. So um, it's an industry that has a lot of problems. And I think one of the reasons that I highlighted it is that AI has this ability to allow industries which previously could not digitize at the same rate as other industries to take advantage of potential productivity gains. Because AI makes different uh, types of digitization so much more accessible. Another great, two other great examples are medicine and the law where because AI provides this very friendly interface, you can have a conversation with it or it can be built into existing tools that you're already using. Suddenly you can do things that before would have required, you know, a huge team of coders and then you would have had this enormous technical debt that you would have to maintain forever to keep the system from breaking. And there's so many of these kind of little niches, uh, some of which turn out to be gigantic that um, AI is helping industries achieve productivity gains that they just couldn't before. And ultimately, when you look at what they're doing, you're like, well, that's just software. And it's like, yeah, but now you can build that software and not have to be a technical person, or you can interact with this system and not have to be a technical person. Or a handful of engineers can build a system that, uh, previously would have taken a huge team of them.

Speaker B: Yeah, I like that you bring up law. That's actually a really good example. I'd love to get your pushback or acknowledgement or reaction to this, but law is right now, I think most would probably argue, in most cases, unaffordable for a lot of people. Like, if something happens to you, uh, it's very expensive to defend yourself and get a lawyer. It's very expensive. And a lot of that is if you peel back the onion. With my very limited understanding of law from my suits watching, uh, is paralegals and a lot of really hard work digging through case law. A lot of stuff like that. A lot of hours. And so I do wonder if very soon law will be one of those things that is incredibly accessible to everybody. You get sued by a giant corporation that has 20 lawyers, and you hire a small firm in Boise, Idaho, that has one lawyer, maybe two, and you can defend yourself because they have AI that can sift through all that case law, saving them thousands of hours. Are we going to see, um, like, law is one of the great examples, are we going to see actually way more demand for legal services because the cost of those are going to come way down.

Speaker A: Yes. Great question. The answer to what you said is 100%. Yes. This is happening now, um, is a trend that is going to accelerate. So let me Just paraphrase the CEO of LexisNexis, who I interviewed the other day. This fact really boggled my mind. So, LexisNexis, of course, for those who don't know, since forever has been, you know, the way that lawyers research case law. Right. So. So, I mean, this has been around so long. You know, my own mother, who was a patent lawyer a million years ago. Oh, I'll never forget. You know, she was, like, so proud when she got trained on it. And so, like, there she was, like, sitting in front of the system. Remember those monitors where the text was green? It's like the matrix. Those are old Hercules monitors. It's black and white, but it's green and black.

Speaker B: It's like when Oregon Trail was cool and it was on those monitors.

Speaker A: This is like the Oregon Trail level of that technology. So lawyers would sit there just punching in. They would have to learn this weird syntax, and they would be punching in different keywords because the only type of search you could do was keyword search. And, um, that's how. That was the genesis of LexisNexis, right, as a digital product today, this thing, which has always been for legal research, has always been. Your paralegals are sitting there all day long researching the relevant case law for your. Whatever suit you're doing. Um, now, the CEO of LexisNexis told me lawyers are using their system more to draft documents than they are for legal research. And that is a change that has happened in the past 12 months. So what's going on? Right? You go in and where before you do the research. And then this is the nature of knowledge work, which AI is disrupting. You know, as another observer recently put it, you know, most knowledge work, it's like inputs in documents out, right? That's what lawyers and paralegals are doing. So they would read all the case law, they would integrate that with their experience in this area of law, and then they would generate documents. Um, well, now, you know, LexisNexis is helping them generate this. And because it is grounded in their database of actual case law, it is. It's not able to hallucinate or if there is anything where you're like, is that real? You can, like, click on the case. In other words, it won't cite a case that doesn't actually exist in its database. So this whole problem that people have had of, like, ChatGPT generating, you know, briefs for the court, and then the lawyers get sanctioned, sometimes fined because ChatGPT made up some case law, it can't do that. Right. It's grounded. And so. Exactly what you just said, right? So you're, You're. You're a sole practitioner in Boise or whatever, and you're handling, you know, any kind of law. Could be corporate law, could be contracts, could be labor disputes, could be wills. Think about how many people have not done. I bet, like, 20% of people who need to actually do, um, you know, a proper will or estate planning have done it. Right. Because it's not. Yeah, estate planning sounds like you're rich. It's like, no. Like, you die and, like, your kids have to sell your house from. It's a nightmare, all of that you would want a lawyer to take care of. Very few people have done it. It's a terrible mess. Then when People pass on that kind of thing is being not automated, but partly automated by these systems. So you are a lawyer in Boise. You know the law in your state. You, um, know what the needs of your client are because you've done the client facing thing that only a human can do still for now. But you're using LexisNexis to help you draft that document and think about how much faster it makes you. Absolutely. You can start to help more people. And what you see then, as any economist would guess, is the cost of that particular service goes down so it becomes more accessible.

Speaker B: Yep. And you mention in the book someone who is actually, you mentioned another example too of someone who's using it to like transcribe and note, catch and do depositions. Because then there's like a. I don't know if like a. I can't remember how you describe this, but basically their AI will tell them if someone answered a question. I'm not a lawyer, but I was like, oh, that's an interesting problem. I didn't even know existed. Apparently in a deposition, like people will try to avoid answering the question and sometimes they're pretty good at it. And so this will tell this lawyer that they did answer the question. It'll check that box. Or no, they haven't answered that yet. And this lawyer's like, oh, gosh, I thought they had. I totally almost got hoodwinked. So absolutely fascinating. Um, but I want to switch gears a little bit because obviously law is one thing. But you mentioned Hollywood earlier and you mentioned one example of a illustrator. I think that is like kind of losing a few jobs to AI and on one hand I think it's cool. On the other hand I think it might be a recipe for disaster for Hollywood M. And I don't know the answer here, but AI in my idiot explanation, um, I always tell people it's like an m. Incredibly amazing copy paste tool. It can't, you know, and this, this will lead into your job too. And why your job is, at least for now, AI proof. But are we going to start to see. Are we going to start to see if people are concepting and creating art for Hollywood movies using AI? Aren't all Hollywood movies eventually just going to start being the same storylines and same plot and we're going to have, I would argue already we have not that much creativity in Hollywood and maybe if they start relying on AI, it's going to get a lot worse. Or am I missing something?

Speaker A: Yeah, three data points on this. Um, one of the current experts on this Believe it or not, is Ben Affleck. He's been giving some great interviews.

Speaker B: I've heard about this. I need to listen to those interviews. I've heard he likes is blowing people's minds.

Speaker A: Yeah, he's a very smart guy. And, um, you know, he's observed that, you know, ChatGPT is not a good writer, but he's like, imagine, you know, you're, you're, you're constructing a plot and you need a particular plot device. You know, you're like, okay, we need something like, why would somebody send a message? But it gets delayed and they. And somebody else only receives it like two days later. And you can just go into, you know, your LLM of choice and start to have that conversation. It's like, oh, well, what about this? Or whatever. So it can be an assistant for you. Right. In those respects. Um, you know, it's, it's, it's truly not a good writer. Right. Because it is going to tend, uh, toward the, the mean of what everybody else has created, which is.

Speaker B: Yes.

Speaker A: Stuff that already exists. And that's just not what I mean. Think of like a really out there, but amazing movie, you know, that, uh,

Speaker B: like, um, Inception or something.

Speaker A: Yeah, Inception or the, or DiCaprio's latest. Um, which is just such an insane film. It's good because there's nothing like it. Okay, see that film not with their kids, because it's crazy. Um, but AI is never going to create that. Right?

Speaker B: Um, I agree.

Speaker A: AI is really great at creating things that have already existed. That's why it's great at vibe coding. Because if you ask it, like, hey, create me like a podcast client. Well, guess what? There's a million podcast clients in its library. Um, so that's data point number one about what this is doing to Hollywood. Data point number two is it is already being used in full production to create things that otherwise they couldn't. A great example is recently there's a hit, ah, series on, um, Amazon about one of the angels. I don't know, it's like a biblical series. It's been a huge hit. It turns out there's, there are scenes in it where they're trying to depict the genesis of the giant, uh, Goliath from the, you know, from the famous David and Goliath story. Mhm. And like, I didn't know this, but he has this crazy, like, mythical thing, fallen angel storyline or whatever. And the director of that used generative AI to create these insane scenes where there's like, angels and these huge expanses and you know, for a show that's just like, you know, episodic show that's going to appear on Amazon, he said this allowed me to do things to be more ambitious with those scenes than I would have otherwise. Now, is the rest of the show generated by AI? No, it's just regular people, you know, uh, talking to each other with cameras right there in their faces. So you can see how it's already supplementing in areas where, you know, we're already pretty used to like, CGI and you know, AI can do that. So is that going to eat more of that kind of stuff? Uh, that's going to appear in, in episodic shows at least and eventually movies, probably. And then, you know, the third thing that I would say about what's happening in Hollywood is it just demonstrates this tension between short term job loss and then long term productivity gain. So in the short term, you know, I talk to like concept artists who are losing their jobs because people are just like, let's just use a, uh, an AI image generator to come up with concept art. You know, it's never going to appear in the final production, but let's just get some ideas from it. Um, and that's going to, you know, move up the value chain. It's already starting to appear in final, uh, production stuff. And in the long run, you know, what people in Hollywood say is, well, this will actually empower a whole new generation of creators. Because imagine if you could be a writer or you are, ah, an indie director just starting out and suddenly you don't have to have this huge CGI budget for your short or your movie or whatever that you're going to do. Wouldn't that be great? So, you know, the X factor is we all have limited attention spans, but, you know, they could get redirected. I mean, the thing that frightens me a little bit, somebody said to me recently, somebody very smart said to me, you know, I really love. I can't remember what franchise it was. Let's just say it's Star wars, okay? Because Disney recently did a deal with OpenAI, which makes me think this could happen. And they were like, when I'm just like putting on a show in the background and not really paying attention, I would be totally happy if Generative AI was just endlessly coughing up more stories about the characters that I always, that I already care about because I'm half paying attention anyway. And like, to me that's chilling because it's like a complete removal of the human from the productivity process. And I Imagine a future in which people are just like endlessly consuming AI slop and loving it. I hope we don't get there. This is the thing about, it's like any new technology, it's like, you know, like splitting the atoms. The best possible example. It's like on the one hand, it's like cool carbon free energy forever. And maybe suddenly someday this is how we get to the stars. Because it's the, the, the heart of like a new type of rocket engine, which is a real thing that's happening by the way. Nuclear powered rockets are coming or we wipe out all of humanity. Right. It's just like AI is on that level.

Speaker B: Yeah, it's a balance. Um, and it's like, it seems like you talk about this in the book. It's a balance. The people who use it as a tool are probably going to be the winners and, and the people who get used are the tools. So, um, it's something you and I already see examples of people, uh, just people I know that are kind of over relying on it and just completely blindly trusting it and doing this hard human work that has been done by critical thinking, creative people. And then they hand it off to AI and say, what do you think about this, AI? And they come back to their team and says, AI says this. And I'm like, dude, what, what are you thinking? Like, I already see this happening. Um, I want to end on maybe a few things about how AI may make us actually more human if we have the time. But you talk about why, at least right now your job is probably not going to be disrupted by AI. I mean, it can't be now. Maybe in the future, who knows? I don't think so. But I want you to talk about that as a final note.

Speaker A: Yeah, so I, I have said, and this is since the earliest days of being a journalist because I really started during the early, early days of, um, I'm gonna date myself now. But like, you know, the web disrupting, uh, print. Right. Like when I, when I first started like a web, a blog was called a web log. Now kids don't even know what a blog is. I told my, I was like, kids, you know what a blog is? And they're like, is that, they were like, is that like a vlog? And I was like, no, okay, forget it.

Speaker B: Wow.

Speaker A: But since the earliest days of this, um, you know, I've always said my job is to, is to, you know, uncover information that's not on the Internet and put it on the Internet. Obviously an LLM cannot. No matter how smart it is or how many deep research tools it has access to, it doesn't have access to the real world. And it's not going to convince anybody to, um, uh, pick up the phone, uh, and have a conversation with it yet. Um, so that is the sort of fundamental reason why journalists, at least journalists like me and investigative journalists, you know, are not going to be, uh, replaced by AI. Now, is it, uh, a potential productivity booster? Yes. I use deep research, uh, tools all the time. They're incredibly good at helping me drill down very quickly to, you know, find out even very obscure things as long as the information is public and already on the Internet.

Speaker B: Right.

Speaker A: Um, you know, help me find sources. Um, and, uh, you know, and I, And I use them in innumerable other ways. I mean, they have. I used to sit at my computer doing interviews, you know, taking notes. Now I literally just, uh, plug in my earbuds and go for a walk because the meeting's getting recorded and I am going to get that full transcript later. And I don't even need to read the full transcript anymore because latent in my memory is like, oh, didn't that person say something about this idea? And I just asked the chatbot, hey, where in this, uh, transcript did this person talk about this? Or. Or what were the. What were all of the things they said about this concept? And it'll pull it from three or four different places in the interview. So it has completely transformed the way that I interact with even the data that I am extracting from other people's brains. And. And frankly, it's just made my life a lot better because I'm not sitting at my desk all the time anymore. AI has freed me to be what in the parlance is called a deskless worker.

Speaker B: Yeah. And is that. That's kind of what I was hoping we could end on, which is you have I. I've been kind of thinking about this a lot as, like, will AI be able to make us more human? And the one example I was thinking of is, like, cars, uh, some Teslas, now you can actually talk to the car. It'll do destinations for you, and it'll plan routes for you. That's one ex. But I love your example so much, which is that you are just taking a walk through nature and talking to people because I empowered you to do this thing. And before, you were stuck kind of dividing your attention between the person you're interviewing and essentially hammering out as many notes as possible to try to capture everything. Does that concept scale? Does AI start to remove the friction of technology because it's kind of working in the background on our behalf and we can go breathe some oxygen or as the kids say, touch grass.

Speaker A: Yeah, I mean, let's just like paint a picture of the not too distant future because we're getting there. Okay. The hardware is there. There's, um, I just saw today, there's like this little ring has a microphone on it, and you use it to talk to your AI, which is in your phone and in your earbuds, or obviously Meta's trying to do it and Apple and Google and everyone with smart glasses. But I think that, like, let's just, let's just like paint a, uh, picture of, of like Jane Doe's day in the future, because I think this is going to be really illustrative. So she wakes up and, um, you know, hopefully like all of her devices are nowhere around here. Like, I don't, I don't want this to be plugged in all the time. I don't want brain implants. I hope people don't volunteer for that. But, okay, you wake up, brush your teeth, have a shower, uh, uh, you're not checking your feeds. I hope for your own mental health. Then you start to, like, put on your devices, you start to become a cyborg, right? You put on your smart glasses, you put in your earbuds, you put on your little ring or your watch or you get out your phone. And now you're equipped to talk to the on device AI, and of course talk to the AI in the cloud. And you know, your assistant's like, hey, good morning. Here's what's on your schedule today. And you're like, cool. Oh, I have that 10:30. Let me move that. And it just does it for you. Okay, we're not that far off. Google's getting a lot better at this with their personal intelligence, which they just released last week. Claude is getting great at it. Um, on device. OpenAI, uh, sorry, anthropic is getting good at it on device. Uh, OpenAI is going to have to match, you know, feature for feature. Microsoft's, uh, going to try. We'll see how far they get. Um, and throughout your day, you're going to have this assistant, which sometimes unprompted, is going to be literally in your ear, like, hey, what about this? Or you're going to activate it and you're going to be like, add this to my shopping list. Add a reminder to do this. Um, oh, I need to do, uh, you know, I need to make an appointment for my Dog to get, you know, go to the vet or get groomed or whatever. Can you take care of that for me and let me know when it's done? And so this, you know, Spike Jones and in the movie her, this assistant that is in your ear and is always accessible and is being judicious in how it, um, commands your attention. And I think that is going to be absolutely ubiquitous. It will be the dominant way that we interact with our devices. So I agree with you. There's a lot of potential bad outcomes coming, but we do have the potential to return to uh, our own humanity. I think this is going to sound crazy, but I think like people are going to read and write less, but it'll be good. We are, I mean we are, we are fundamentally. Oh yeah. So we are going to return to an oral culture. It's happening already. I mean the bad side is if you look at all the numbers, um, you know, kids are, are less literate than they used to be. They have trouble reading books and stuff like this because they're just like scrolling tick tock or whatever. But I think that we are going to have more conversations, we're going to have more conversational interfaces, we are going to type less, we'll still read because there's, there's no better way to think than to write. And there is no better way to be thoughtful than to read deeply. But those things will be more of a choice. And so all of this endless, like, I gotta look at my phone, I'm gonna look at my calendar over here, I gotta read this email. Potentially all of that becomes something that is more respectful of our intention, which is delivered orally because of course, you know, the bitrate for speech is really high M. As long as you have a system that is above 95% accuracy in transcribing it, which we're now at and which can truly understand and has that error connection correction mechanism which is like, oh, you said you want to accomplish xyz. Is this actually what you want? And you're like, no, tweak this. Yes. And can function like a real human assistant. I mean, I really think we're going to get to a place where every single one of us is going to have a full time dedicated assistant that we feel we can't live without.

Speaker B: Yeah, I totally agree. And it's so many, when you start doing the mental math so much, so much of your day that you find irritating is the toil stuff, is the filling out the forms, is the repeating the same stupid process and pattern over and over again is that scheduling that thing, like, so much of it is that. And so in my mind I'm like, man, I feel like AI is going to be a net good. I feel like people will be able to live out their human purpose as toil love. It gets replaced. But maybe, I don't know for sure, maybe I'm the crazy one, but after reading your book how to AI, I actually feel a lot better about the future. Christopher Mims, columnist of, uh, Wall Street Journal's Keywords, host of Bold Names, which you should all subscribe to, and of course, the author of how to AI, which is out very, very soon. Christopher, uh, what is the URL, the website the easiest way to find you to buy the book?

Speaker A: Yeah, um, I think probably the easiest way to find me is just like, search for how to AI Christopher Mims, and a bunch will pop up. Because I have a newsletter, I've got videos I'm doing all over the place. It's, uh, I'm doing kind of like daily videos on LinkedIn and Instagram and everywhere else. So just search for, you know, how to AI Christopher Mims, and a bunch of stuff will pop up. And, um, it'll really give you a flavor for kind of like what the book is about. And um, you know, I really, truly, I mean, I know I'm biased, but I really, truly tried to believe to write something that will give you general principles and will stand the test of time. Like if AI has like a Das Kapital, like, I'm hoping this is it.

Speaker B: Yeah, I will say, like, one of the things I like about it is just very. It's very, um, there's narrative and story, but it's also very practical. You're like, here are a bunch of terms that people are using. So if someone is like, knows a lot about AI and engineering side, they're going to learn story and application. If someone kind of knows story and application, they're going to learn the technical side. So it's a pretty good book. Uh, and I do suggest everybody read it. I've of course read it. Thanks for sending the PDF ahead of time. Christopher Mims, author of how to AI thanks for coming on the show, Justin.

Speaker A: Thanks so much for having me.