AI Isn’t Passing You By - How to Catch Up (with Jeremy Utley)

0:00Jeremy Utley: Yeah, the way I think about it very simply is AI is a capable teammate who's going to thrive under your leadership. But you cannot abdicate your leadership. It's very important that you bring your voice, your discernment, your taste, your judgment to the table. That's actually what AI is multiplying. And if you don't bring that, any multiple times zero is zero. You need to bring that. That's the underlying kind of value to be multiplied. What you've got to do is when you get suboptimal output, you have to give feedback. You've got to make sure that not only your thinking is known or your response is known, but your reasoning is known.

0:40Jeremy Utley: If you don't make your reasoning known, then the AI is left to interpret or attempt to read your mind.

0:50Before we jump into today's show, I want to share some feedback that I got. People have said it is very hard to know how to roll out a culture of radical candor on their team. And as a result, we've started a company, Radical Candor. You can find it at RadicalCandor.com. Great team of people who have developed talks, workshops, and other ways for you to roll out radical candor in your organization. Hello, everybody, and welcome to the Radical Sabbatical. Today, I'm thrilled to have with me Jeremy Utley, who wrote a wonderful book called The Human Advantage. Welcome, Jeremy.

1:26Jeremy Utley: Thanks for having me. I'm glad you're here. You also do other things in addition to write this book. You're at Stanford, and you teach people how to use AI. And in fact, you taught my son and me, so thank you very much. Well, I'd love to know what you learned from me, actually. That's a fun place to start. Oh, my gosh. So there was a moment where you explained why it's important to give AI a persona when you're giving a good prompt. Why don't you share that with people? Because it was very illuminating. You gave an example of what do you notice in this room? Yeah, yeah, yeah. There are all sorts of ways you can interact with AI that improve its ability to do a job well.

2:07Jeremy Utley: And one simple way anybody can work with AI to improve its ability to do a job well is to tell it what expertise you want it to bear in mind or what role you want it to play. So for example, if you're doing a financial transaction and you want financial advice, not that I'm suggesting that, but that's the first thing that came to my mind, you would say, I want you to be a financial advisor. Actually say that, right? Or if you're working on your sink, you might say, I want you to be an experienced plumber. and I was telling the class that you attended why that's valuable and it has to do with what's called the attention mechanism but it's basically where the model focuses and humans do the same thing you know if I told you in in a room hey look around the room right now and anybody who's listening can do this describe the room you're in you know think for a second about how you describe it and then if I say now describe the room you're in as a biologist chances are your description is different.

3:03Jeremy Utley: Describe the room you're in as a economist. Whoa, the room just changed. What if I said, describe the room as a petty thief? You weren't looking for the easy mark like two seconds ago, but now you're looking for who's got the valuables out. Why? Because I use words to invoke a perspective. And it turns out there's actually a ton of information in your language model, your brain, that you don't think to draw on unless I use certain words. And basically, AI is the same way. And it thinks to draw on certain sets of knowledge or information, depending on the words that you use to prompt it. So telling it what persona you want it to play actually does help in shaping the output of the conversation.

3:42It's so funny because as soon as you explain that, I was much less annoyed with the, I always felt so fake and dumb to say, pretend you're a experienced plumber. And now I wanted to focus on not all the world's information, but on the world's information that's relevant to this.

3:58Jeremy Utley: Well, if you don't say something like that, you're basically asking Reddit effectively. You're saying, hey, Reddit, what do you think about my plumbing problem, right? And that may or may not be what you want. And so when you realize that the words you use actually constrain where in the model's training data it's looking for the next words that it should be suggesting, then you start to be more thoughtful about the words that you use. So your book is incredibly practical. And not only that, you have QR codes so that people, when they buy your book, they can begin to put the ideas into practice right away.

4:34Jeremy Utley: Yeah, yeah. And the first thing that you suggested is to basically use AI to hire five people to help you in your life. So talk about that idea because I do feel like in the world of AI, we all become managers. I mean, we're not managing people, but there's something about trying to understand how it works that's important. So let's first of all talk about what your team is and then talk about how you can use radical candor to manage that team of five. Well, it's directly related actually to the conversation we were just having, which is whose expertise or perspective do you want in a given moment?

5:12Jeremy Utley: And again, like you don't ask a plumber for marriage advice. You probably don't ask a therapist for a backed up toilet, right? But then you go, okay, well, what kinds of personas might I need to draw upon in the early days of my collaboration with AI? And I learned this from my friend, Bryce Chalamel, who's actually, he was at the time when I met him, the head of AI at Moderna. Now he is the head of AI adoption at OpenAI because of how successful the transformation at Moderna went. Yeah, it's pretty cool. And the way he described the transformation at Moderna was there's no more individual contributors here.

5:42Jeremy Utley: Now that everyone's proficient with AI, we're minimum of team. Every person is a team of five. And he described the team of five as number one, yourself. Number two, your AI-powered assistant. Number three is your AI-powered coach. Number four is your AI-powered expert. And number five is your AI-powered creative partner. And so even just thinking in terms of those simple personas kind of helps you define search parameters for where AI might help in your life. Like, where do I need coaching, for example? I mean, just yesterday, I was on a call with a professional sports team. And I had a really interesting conversation.

6:19Jeremy Utley: We're talking about all sorts of possibilities. these. And then after the call was over, I've got a transcript. They let in my AI note taker. And I just thought, I wonder how I did. Well, I have an AI coach. So I dropped the transcript into my AI coach. And I said, hey, coach, give me a little feedback. Where did I shine? And where did I get in my own way? And honestly, Kim, I actually wish I hadn't asked because it gave me so much. Oh no, I could have done so much better. But just thinking in terms of the simple personas, yeah, oh, a hundred percent. But even those simple personas, coach, assistant, expert, creative partner, just helps to kind of steer how we as humans think about the possibilities.

7:02Jeremy Utley: Those aren't the only teammates we might think about hiring, but in the early days, there's a simple way to think about the various roles AI could play to start to augment you. Yeah. What other AI team members have you created as you've gotten deeper? So I'll give you a very practical example. Right now, I am working on a couple of op-eds. It's book launch season. I'd love to place an op-ed in a major publication. And I've kind of got an idea of the shape of the op-ed, but I want a really timely story from an organization. Specifically what I want, because I'm actually talking about CEOs in this op-ed.

7:39Jeremy Utley: I want a CEO to say a certain kind of thing. But you want to find that. So it's your research assistant. So every morning I've got a research assistant that goes out with instructions that knows what I'm looking for and reads every earnings call from the day before. And looks, does, has any public company CEO said something like this? And then it pulls out anything it finds relevant and scores it against a rubric that I've developed with that analyst. and if it meets a certain threshold, it sends me an email and says, hey, the CEO of ServiceNow just said this yesterday. I score this a 75 out of 100 and it's for your consideration, right?

8:19Jeremy Utley: So every day and many days actually, Kim, the analyst writes me and says, hey, I read all of the transcripts yesterday, nothing popped out. Now, I actually don't know about the false negative, but I will say I don't have capacity to read a bunch of earnings call reports, especially when it's kind of a needle in a haystack problem and effectively augmenting myself with a research team is something I've never been able to do before. Yeah. Did you do the thing where you get another AI to judge the first AI's output to look for false negatives or false positives? One can. In this case, I haven't been that prescriptive.

8:55Jeremy Utley: I'll give you another simpler one because few people probably are trying to write op-eds. But a very simple one is my travel briefing agent. So, you know, like one kind of perennial annoyance in the Utley household is daddy doesn't know what his flight schedules are. And so people ask, dad, are you going to be home on Tuesday? And I'm like, I got to look at my calendar. Well, I have an agent now pull together a Saturday briefing where every Saturday it looks at my calendar for the next week and looks for flights, hotels, keynotes, etc. and then drops a one-page briefing via email to me and my wife.

9:33Jeremy Utley: And it just kind of has all of my information right there, right? And that's, again, I never had somebody looking for all my travel information and putting it in one place every single week. And, you know, I actually, what's funny is I was testing, I was teaching a professional soccer team's back office, or front office actually, how do you create agents and schedule tasks and things like that. So I was doing that and I was giving them a demo of a match report. So every day they go scour player performance in leagues across the world and send them a match report. But while they're doing that, I was like, I honestly don't want a match report.

10:08Jeremy Utley: So I just changed the text that I was giving them for my own purposes. And then I said, hey, everybody hit enter and let's see if it works. And I hit enter and it's an integration of my calendar, my email, all sorts of stuff. And so I didn't know if it would work the first time. And then I get a text message from my wife. And Michelle's not an all caps gal, but she writes me back all caps. Oh my gosh, that email is amazing. Why did it take you 20 years? I go, I think it worked. It worked. It worked. That is incredible. And the nice thing about that story is that now you and your wife can talk about things that have meaning to you too as human beings.

10:46You don't have to have the irritating conversation of where are you on Tuesday? And then you pull up your calendar.

10:52Jeremy Utley: Exactly. I think that one of the optimistic scenarios about AI is that it frees us up to have better relationships because we're automating some of the friction out of our relationships. One of the things that Jason, my co-founder, has done for himself and for the whole team is to create some agents that people can turn to for tech support. because he's the most technical member of the team. Everybody, you know, I'm having trouble logging into HubSpot or whatever. Sure, sure. And so he created this agent and now everybody goes to Claude to answer those irritating little questions and not to him.

11:34Jeremy Utley: I'm sure his quality of life has improved enormously. And so has yours, right? Because now you don't have to bug Jason about that stuff. And so his quality of life improves. I get answers faster without having to hesitate to ask because I don't want to irritate him. and our relationship is, I mean, not that it was ever bad, but we're now able to spend 15 or 20 % more time each week talking about things that interest us and that really matter to us rather than how Kim can log in to HubSpot or whatever. It's beautiful. It's a beautiful example. That's really great. Yeah. All right. So another thing that you write about is how to use AI with your orthogonal thinking.

12:13And you suggest that people ask it to respond to something as an alien, a toddler, and a retired sea captain. I love this idea. Explain this method, this AI approach.

12:26Jeremy Utley: It's very simple. I mean, you know, I'm a creativity junkie. I'm a student of creativity for the last 20 years. That's really been my area of focus. And one of the simplest ways to think about the search for good ideas is you actually need to open yourself up to bad ideas. Yeah. And the challenge is usually when you need good ideas, you're, you're less open. You're because you're, you're, you're being more precious, right? So there's this kind of like paradox of you are unwilling to do the very thing that will help you. Like the best way to get better ideas is to force yourself to have worse ideas.

13:00Yeah.

13:01Jeremy Utley: But usually when you feel the pressure to have better ideas, you go, well, whatever I do, I can't have bad ideas, right? I can't waste my time with those bad ideas. So that's actually, you know, that's a big problem. And it's a big problem in creativity in general. And now you get AI and you go, well, what do I ask AI for? And the natural thing is I want better ideas. I want better ideas. And a fun kind of way to flip the frame is to go, no, no, no. Actually, I need help getting worse ideas. Yeah. Because my human brain is actually quite capable of activating and responding to something that's wacky that a sea captain might say.

13:33Jeremy Utley: I just, I can't get myself to think of the sea captain thing. Right. And so you can actually enlist AI's willingness to entertain absurdity. And then you activate your own brain's ability to go, wait, there's that alien said something is actually kind of interesting. Yeah. Yeah. So it's almost a way to trick yourself. Yeah. So give me an example of how you, how you've used this in your work, this approach to get it like kooky ideas. I mean, I'll tell you this, you know, I have four amazing daughters. and this true story, I was making breakfast the other day. It's very important. We all have role clarity, right?

14:07Jeremy Utley: I do breakfast, right? So I'm making breakfast and I hear a very loud bang. And like any good father, I pretend I hear nothing. And then my 14-year-old's like, Dad, get up here. There's glass everywhere. So I go to the front of the house and I'm devastated to see my favorite window. We live in this 130-year-old house. My favorite window is shattered on the floor because seemingly sensibly the 12 year old had thought, I need a little personal space. And the best way to accomplish that is slam the door in my sister's face so hard, the window shatters, right? And so I'm looking at this window and I'm looking at eight eyeballs going, what is dad going to do?

14:47Jeremy Utley: And I just pulled out my AI chief of staff and said, hey, I need 20 terrible ways to respond to this. and it's like well here's a few russian torture tactics you know here's uh you know you could call the police right you could expel her and all of a sudden you know i start laughing it lowered the temperature it's like patently absurd right but but it just expanded the solution space enormously because in that moment the only thing i can think of is i guess i got to take away the kindle fire right that's it yeah yeah and you know there was something about expulsion we homeschool are girls. So it's a particularly funny idea to think about expelling.

15:28Jeremy Utley: We're not going to, obviously, not going to kick her out, but then that made us think, wait, school, how could school be a part of it? What if the offending girl has to teach her sisters a homeschool module about why our household rules matter? And one rule is not slamming the doors. She is uniquely qualified to deliver that message, right? And so it was inventive. It was fun. It It was kind of on brand with our family values. All of it came because I asked for terrible ideas. Yeah, yeah, yeah. I love that. Such a good example. And I hope you got the window fixed. Was it one of those like old leaded?

16:06Yeah, yeah. Sadly, our contractor said irreplaceable.

16:10Jeremy Utley: Yeah, so. That's a shame. She's going to be working it off from her allowance for the next five years. But the lesson is invaluable. Works way more than the old leaded window. Exactly, exactly. All right. In this case, you evaluated the output. You did not take the ridiculous suggestions. And so another thing that you write about, you have a whole chapter on how to evaluate your AI output because you're the boss of this AI output and you're responsible for it. Like we have a rule, a radical candor that none of us are going to be meat routers. What I mean by that is I'm going to type in, somebody asked me a question.

16:52I type in AI. I don't even look at the output and I just email it to somebody. I love that.

16:57Jeremy Utley: No meat routers. That's great. We are not routers made of meat, meaning we're going to add value. We're going to evaluate AI. I just wrote that down. That's so good. That's good. We're going to evaluate the AI output. Of course, I broke the rule yesterday, but I admitted I broke it. I said. Which is a first step in the right direction, I think. Yeah, yeah. So anyway, how can we make sure that we're evaluating the AI output? Because in Cory Doctorow's most recent book about the reverse centaur, he points out that often when you're getting this AI content back, and most of the time it gets it right, it's very hard to continue to evaluate it.

17:42And he uses the example of people reading radiologist reports. And he suggests that the right thing to do would be to have the radiologist read it and then AI judge it rather than AI read it and then the radiologist judge the AI output. Because pretty soon you just start passing it. If you're judging it rather than driving it, you might not evaluate the AI output.

18:10Jeremy Utley: Yeah, the way I think about it very simply is AI is a capable teammate who's going to thrive under your leadership. But you cannot abdicate your leadership. It's very important that you bring your voice, your discernment, your taste, your judgment to the table. That's actually what AI is multiplying. And if you don't bring that, any multiple times zero is zero, right? So you need to bring that. That's the underlying kind of value to be multiplied. So there, what you've got to do is when you get suboptimal output, you have to give feedback. You've got to make sure that not only your thinking is known or your response is known, but your reasoning is known.

18:55Jeremy Utley: If you don't make your reasoning known, then the AI is left to interpret or attempt to read your mind. So I ask for, I mean, a simple example is I ask for 10 ideas. I had Ed Catmull on my podcast. I asked for 10 ideas about how to title the episode. Three of them are decent. I go, oh, I like numbers three and seven and nine. Can you give me some more like that? Well, you may not be surprised to discover I didn't get much better work, right? Because I didn't think about why, right? So then I go, hang on. The reason I like number three is there's alliteration and I'm kind of a sucker for alliteration.

19:30Jeremy Utley: The reason I like number six is because there's an allusion to cinema history and I like that. The reason I like number nine is there's kind of like a little punny dad joke. Can you leveraging those three criteria as kind of constraints generate 10 more ideas? Well, then I got 10 ideas. It's tough to choose from. Yeah, but it required me to actually be explicit about what's in my head. And I think we're used to doing this with humans, or humans kind of require it in a different kind of way, right? Because if I tell you, I like number two and five, do more like that, you're like, what is it about two and five?

20:06Jeremy Utley: Whereas an AI, because of its pre-training, its desire is to be helpful. If you say, I like number two and five, do more, it doesn't say, why do you like two and five? It says, you bet, here's more like two and five, right? So it's incumbent upon you as the human. And this is where even basic understanding of what's happening underneath the hood, not that you have to get a PhD in computer science, but a basic understanding of what's happening under the hood actually helps because it enables you to be a better manager. And here's the thing that I find really interesting. Something like only 1 % of professionals have been given an assistant.

20:44Jeremy Utley: Very few people ever have an assistant. But the thing is, whenever you become a new manager and you get a team that's reporting to you, most organizations recognize managing a team of people is a distinct skill set from whatever got you here. And then they give you new manager training, right? So that you can be good at getting great work out of other humans. Most folks are getting an assistant now and they're not getting new manager training. Yeah. Is it any wonder? Yeah, I might disagree with you. Most people get a team of people and they don't get new manager training. Yeah, you're right. They should.

21:19Jeremy Utley: I mean, in an ideal world, they do. Yeah. This world is not ideal. But even take as the first, it's fascinating to me. You've been given an assistant. Most people don't know what to do with an assistant. Yes. Is it any wonder? You know, if I told you, if you're listening to this episode, you get one Harvard Business School caliber intern who's got 100 hours a week. how will you make best use of that person? Most people don't, they don't know how to answer that question. I don't know what I would do if I had a Harvard Business School intern who gave me 100 hours a week. It's not about paying them.

21:53Jeremy Utley: It's not like I, even if it was free, most people would struggle to make fair use of the intelligence available to them. Yes. And that's actually, it's like right now, I think the biggest limitations are imagination. What do we even think to ask of a vast intelligence that now we'll do our bidding if only we can think to bid it. Yes. And one of the other disciplines that we've been at Radical Candor trying to implement is before you feed your prompt into Claude, which is what we're using, but whatever AI you're using, discuss your prompt with another human who will say, this doesn't make any sense to me, or I don't understand, or I disagree with your premise.

22:38Jeremy Utley: Brilliant, brilliant, yes. Because the problem that we were having is that people were sending me stuff that looked fully formed. It was 20 pages long. And I had to read the whole 20 pages to understand that the premise, it's much faster for me to read your prompt than for me to read the output. That's brilliant. That's brilliant. Yeah, so read each other's prompts. Like use your humans, use your colleagues, use the real people to help you prompt better because AI - It's almost protecting. That's fascinating. Okay. So can I tell you something that I, and I think there's a direct parallel here.

23:19Jeremy Utley: We used, it used to be that engineering time was a precious resource and we would have meetings to protect the deployment of that resource, right? So now engineering isn't actually the precious resource. And often the meeting, gathering a leadership team is the precious resource. And I think in the same way, it's kind of an analogous thing here. Recognizing that your prompt actually is incredibly important, refining the prompt with a team of humans who need to be protected from 20 pages of output. It's going to go in whatever direction you send it. And if you're one degree off, 20 miles, 20 pages down the road, that's a long way off from where we want to be yeah that's brilliant that's really brilliant it's well i i think sherry turkle so she has this new book artificial intimacy and she describes i think this is where i read this but she describes a situation where if you're going to ask your professor to write a letter of recommendation in the before ai times you would only ask your professor to do this because it took your your professor real time but now professors are getting a little bit more profligate in their letters of recommendation because they can write four bullet points and AI will generate this beautiful long recommendation letter that's three pages long.

24:40And now what happens is that the four bullet points turn into three pages of text, which gets sent to someone who doesn't have time to read all these recommendations. So they're using AI on the other end to distill, but it's like a game of telephone. The four bullet points, it would be faster if the professor just sent the four bullet points, faster and more accurate.

25:02Jeremy Utley: I think that's a brilliant, I mean, and here's something underneath what you're saying, which actually perhaps may be useful just to put a fine point on for your audience. The outputs of an AI are primarily a function of the inputs you give it. Yes. And so, and so the reason some people get better output out of an AI is not because they have a better model. It's because they've been more thoughtful about their input inputs yeah yeah and i think people don't realize that yeah when i spent i spent one summer at mckenzie and one of the things that we talked a lot about there is do the upfront thinking don't jump in and start making your slide decks if you spend the time upfront to think and talk with each other don't start too soon so i think that's The danger of AI is that it looks like it's done, but it isn't.

25:55Yeah. It's garbage in, garbage out.

25:57Jeremy Utley: Yeah. Geico is exactly right. Geico is garbage in, garbage out. One thing I think, a simple assumption a lot of people make is I provided sufficient information to this AI. And the human test is one way to do it, right? You walk it down the hall to a human. The human goes, what does this mean? Right? Yeah. Which is great. Another simple test, if you don't have a human available to you, but a really simple hack that's hugely valuable is end your prompt to an AI with the following statement. Say, I'm sure I haven't given you all the information you need to do a really good job at this. Before you start the task, would you ask me at least three questions to make sure you have enough information to do a great job?

26:38Jeremy Utley: Yeah. It's very simple, but when the AI thinks its job is not to do the thing, but rather to seek the understanding required to do the thing, it's actually spectacularly capable of seeking that understanding, provided you give it permission to ask. Yeah. You have to not just give it permission. You have to demand that it asks. Yes. You have to solicit feedback. Yes. Before you give your AI feedback, solicit feedback from it. Well said. Which is true with people. get it, get it, gauge it. That's where your management training, your feedback training that you got by hiring Radical Candor will help you manage your AI.

27:18You got to make sure that the humans are capable of evaluating the AI output. And the danger with AI is it can just drown you in slop if you're not very thoughtful. And I think part of the issue here is that Thinking is the hardest part. Executing your thinking is the easier part. And in order to get something good out of AI, you really have to think. That's right. And I think it used to be that you could spend 10 % of your time thinking and 90 % of your time doing. And now AI is doing all the doing if you're using it correctly. But you can't think 100 % of your time.

27:58Jeremy Utley: Well, they say writing is thinking. Yes. And insofar as now you are not writing, not writing is not thinking. Yes. You know what I mean? And so then if you're not the one writing, then you definitely need to be the one thinking. Yes. Yes. Well, and I would argue that you shouldn't use AI to write your first draft. You should write your own first draft. Yes. Yes. Unless it's some kind of garbage thing. but you know one one thing i feel like we haven't really touched on that that may be interesting is uh the feeling that some people have that they're behind yes and i just want to just kind of normalize that for a moment and maybe de-stigmatize it because that can be a very paralyzing feeling you know i gotta you'll love this kim i have this course on my website anybody can sign up for it's like a three-week long boot camp every day you get a drill like do this with AI, right?

28:55Jeremy Utley: Well, I had, and if somebody doesn't enroll, which to me is a very no-brainer, it's cheap, it's fast, it's great, all that stuff, then it'll, the system will just send them a quick survey. Why didn't you enroll? Was it, was it too expensive? Was the time not right? Are you not interested in AI? And I had a Stanford professor who will remain nameless, but a Stanford professor replied to me and said, hi, Jeremy. They didn't just fill in the survey. They said, your survey did not include the following option. Fear of learning something new that I should effing already know. Right? That is the truth.

29:28Jeremy Utley: And that is, I think, how a lot of people feel. And it's exactly the wrong instance. If you feel that, you don't bury your head in the sand, right? What I hope to do is make it so simple that you go, hey, it's perfectly normal to feel behind. Andre Karpathy, who's one of the co-founders of OpenAI, okay? This is one of the best programmers in the world, he recently tweeted, I've never felt more behind as a programmer. Okay. If Andre Karpathy feels that way, it's, I would go so far as to say, if you don't feel behind, you're delusional. Yes. Okay. So it's fine to be behind. The question is, what do you do when you feel behind?

30:04Jeremy Utley: And if you kind of put your fingers in your ears and go, you know, that's not good. Right. So I just want to say to anybody who's listening, it goes like, you guys are using these word agents and like, because even our conversation can probably be intimidating to somebody who maybe is a little bit self-conscious. And I want to say to that person, just get going, trying new ways of working. Don't be intimidated and don't feel like the ship has already passed. It definitely hasn't. No. I mean, I think I was at a conference recently where people were talking only about AI all day long. And I can summarize, I can save you all a lot of time.

30:43The TLDR of that conference was, we have no idea what's going to happen. So get in and start playing around with it. And one of the things that you talk about in the book is talking rather than typing. Take a walk. And you also describe how you got your grandmother using AI. So share some of that.

31:03Jeremy Utley: Well, my grandma's awesome. Well, what started, this was interesting. We were driving. She lives in Oklahoma. We're driving together. I'm driving. She's in the passenger seat. And this is a couple of years ago. And she just goes, hey, Jeremy, what's this chat thing you keep talking about on Facebook? And I mean, just think about how do you answer your grandma when she asks about AI? And I just said, what is an emotional decision you're thinking about right now? And she goes, I thought this was technology. I said, I know. Just bear with me for a second. She goes, well, your grandpa and I have been wondering when it's time to move into assisted living.

31:38Jeremy Utley: And I didn't know that. I said, wow, thank you for sharing that. I want to show you what this chat thing is by demonstrating this conversation. And actually, I'm going to invite you in. And so it was just voiced by virtue of the constraints because I was driving. So rather than typing, I just used my voice. But I said to – it was chat at the time. I think it was the only one. How did you do it? You opened up chat and you clicked on the phone just for people who – yeah. Exactly. Yeah. I opened up chat and I hit – there's actually a – I don't even know if voice mode was even out yet. meaning now you can actually have a conversation where the AI will speak back verbally, audibly to you.

32:14Jeremy Utley: At the time it was just voice input. And so I think I just hit the microphone and I said, Hey, my grandma just asked me how she should think about assisted living. And to be honest, I have no idea how to even approach that. Would you please ask us a few questions about her life until you have enough context that you could make a recommendation? And then the AI said, you know, sure. Can you tell us about her or can you tell me about her current health? Does she have any health issues? And I just handed her my phone. I said, granny, you talk into it. And so she goes, oh, I mean, you know, a pop-off fell down, but he plays golf.

32:50Jeremy Utley: He's fine. And I don't really have it. Is this what it means? I said, yeah, that's fine. I said, what's the next question? Chad goes, tell me about her relationship with her caregivers. And I said, granny, talk, talk to it. She goes, well, you know, we don't really have, I mean, the youth group comes by sometimes, but it's because we make cookies, you know, and my daughter, his mom, Vicki, she, you know, is this what it means? I go, yeah, it's fine. We do that a few times. And then I just said to Chad, Hey, based on the information you've gathered in this conversation, how would you advise granny to think about this question?

33:20Jeremy Utley: And then I hand her to the phone because I'm driving, so I can't read the response. And she's kind of quiet there for a while. And I, after a few moments of silence, I said, so what do you think? And she goes, you know, I've never thought about it like this before. And I said, yeah, I mean, we can talk to computers now. And she goes, not the computer, Jeremy. I've never thought about assisted living like this before. And that, I mean, I have chills even saying that right now because I go, and this is stupid chat. You know, like we have smart chat now. And I was shocked, you know? And so there's a lot we can talk about granny or we can talk about boys.

33:56Jeremy Utley: I don't know where you want to go you tell me uh let's let's talk about let's talk about voice like that's a beautiful example of how you introduced your grandmother to this technology and she kept using it right oh yeah oh yeah she's she's regular yeah her latest her latest use case by the way she told me she's diabetic and she's been reinventing some of our old family recipes to be diabetic friendly recipes and now she's sharing them with the ladies in her neighborhood in oklahoma That's amazing. And you never would have thought of that in a million years. No, no. But I mean, that's the thing that's so cool about this, right, Kim?

34:31Jeremy Utley: It's like, it's her background. It's her perspective. It's her pain point. It's her context, you know. And the human who feels comfortable bringing their perspective and background and pain point and context, I often joke with people. I say, the answer is yes. What's your question? Right. You know, and for most people, they don't have a good question, right? Yeah. And granny goes, hey, we're out of cream of mushroom soup for the green bean casserole. Could your chat thing help? You bet, right? Again, I never would have thought of that, right? Yeah, yeah. But it's really about bringing that pain point and context and background.

35:03Jeremy Utley: But the thing about voice that's so cool is not only is it faster, you know, I mean, we speak usually at roughly 4X, the rate at which we can type, but a different part of our brains are activated when we're speaking. Like this conversation is totally different than it would have been if we were, you know, writing letters back and forth. Yeah. And that unlocks, I think it's actually a hack. It's a human hack to get context. When you use your voice instead, I call your fingers your bottlenecks. And if you bypass the bottleneck, you start bringing much better context to the interaction, which enables the AI to do better work.

35:36Yeah. The fingers as bottleneck is a really good analogy because part of the reason why writing is thinking is that it slows your thoughts down. But I think most people hate to write. And so talking and that sort of back and forth and I don't have to get it right is really helpful as long as you push AI not to be too sycophantic, I think.

36:01Jeremy Utley: Yeah. And that, all you have to do there is specify that desire. I want you to challenge my thinking. I want you to identify blind spots. I want you to let me know when I'm making assumptions. I never want one response. I always want you to give three radically different ones and request me to do a choose your own adventure as to where this should go. Right. There are simple ways that I mean, a lot of people will a very concerned mother. It's fascinating, actually, when I'm doing keynotes, moms usually approach me for one of two things, which are which are kind of opposite, but the same in some way.

36:33Jeremy Utley: One is they say, I am deriving immense value from AI and I've got to hide in the pantry because my kids are being told at school that their mom's a cheater. Oh my God. What do I do? And actually I hear this a lot. And why are dads not saying that to you? I don't know. That's a good question. But it's always, I would say 95 % of the time it's happened more than 10 times and I don't think a man has ever asked me. Wow. But women say, I'm getting so much value and my kids are accusing me of cheating. That's kind of one response. And then the other is, I'm worried about my kids' kind of critical thinking.

37:08Jeremy Utley: How do I make sure that, you know, this whole cognitive offloading thing? To which I reply, you tell the AI that you want the interaction to sharpen your thinking, not weaken it, right? So don't think for me. Ask me sharp questions that force me to develop my own opinion, right? It's an amazing capability of the AI that it's going to do what you ask. You just have to ask. Yeah. And most people, they approach it like it's a Google search box, like, oh, I type this thing in and I'm done. Rather than realizing, no, you can actually shape the nature of the interaction significantly. And that's, again, it's mostly bounded by what you imagine to happen in the interaction.

37:48And I think also, I mean, you talk about using AI as a thought partner. Share your thoughts around that, how to do that. And then one of the things I'm trying to do in this podcast is have some respectful disagreement. So I'm not sure I disagree. I'm not sure I agree. I'm not sure I'm right or wrong. But let's talk about the benefits of using AI as a thought partner. And then I'll suggest a risk of using AI as a thought partner.

38:13Jeremy Utley: Well, I mean, I think one thing that's incredibly powerful is, and it ties to our last conversation with voice, you can go on a walk with an AI coach, you know. And, you know, for example, if you know your Enneagram type, you can say, I want you to be my Enneagram coach. I'm an Enneagram three or whatever you are, right? Tell me what you know about an Enneagram three, you know, pros and cons. It'll kind of give you strengths, weaknesses, et cetera. Now I want you to ask me about my performance at work for the last week as my Enneagram coach, knowing that I'm an Enneagram three, right? Ask me three or four questions to get me to reflect and then give me some feedback or areas that I could focus on next week.

38:53Jeremy Utley: You can just, it's not unlike walking with a human coach in the park, right? Yes. The coach is going to talk to you. Yes. What do you lose by talking to an AI rather than a human coach? Because, I mean, you may, and part of the answer is you may not be able to afford a human coach, which I. That's where most people are. But yes. Yes. And one of the points that Sherry Turkle makes in her book, Artificial Intimacy, is the danger of using AI as a therapist, for example. Yeah. And I think a lot of people say, oh, well, people can't afford therapists. Let's solve the problem that people can't afford therapists.

39:33Because I do think there are some conversations that we're better off having with another. I mean, let's take the extreme example. Don't try to marry your AI. I think we can agree that's an example of AI. Having a romance with an AI is a really terrible idea. Don't try to make chat your friend. Like make real human friends.

40:02Jeremy Utley: Yeah. For sure. Like where is the line in your mind around that? I am sympathetic to the fact that there are some people in difficult situations. And I, so I hesitate to judge because I think there are some kind of what, what's the phrase luxury beliefs about, you know, just, just hire a real coach, for example. It's like, well, you know, that's like saying just buy a Maserati. It like feels great at the interior. It's like some people go, uh, must be nice. Right. So I'm, I, I, I'm reluctant to condemn. That being said, I do think you have to bear in mind, this is not another being. It's not another person.

40:47Jeremy Utley: It doesn't have personhood. It doesn't have agency. It doesn't care about you in any actual or real sense. It is adept at predicting what the appropriate response is based on the context that's been given. Yes. And so insofar as you are in need of an appropriate response, like for example, like you hear about people who they'll write a letter and then they'll put it in a burn drawer because they're not sure they should ever send it. You know, it's like, that's a great, I think an AI kind of venting session is actually provided that you don't like get more spun up and then all of a sudden get vindicated in your belief that you should really let this person have it, you know, which, which I guess is a conceivable outcome.

41:34Sometimes it definitely, that definitely happens. That would not be good. That would not be good.

41:38Jeremy Utley: But I think I would actually say, Kim, just if you want a hot take or even for the purpose of debate, I think the greater danger is people not being human enough. And what I mean by that is the great irony of this moment is now computers can actually talk to humans and humans are the robots. We're the ones who are hyper robotic in our interaction. And the kind of effectiveness and the quality and the possibility of our AI-powered collaborations are limited right now by our roboticism. And so on the spectrum from some people are being way too human and then on the other end, people are way too little human.

42:25Jeremy Utley: I do feel there is a danger over here that some people should be aware of. What's the danger of being too human? Well, going back to what you were saying about companionship or thinking that the AI is your friend. Of using AI into human of a way. I think there is danger over there. But I would say for the vast majority, if people are kind of in their right mind and they have a fully developed prefrontal cortex and executive function and all of that, like provided some developmental guidelines, things like that. most people I think should move in the direction more probably towards relatability and humanity and their interaction rather than away from.

43:02Jeremy Utley: And I'm hesitant to warn people about the kind of extreme of dependency or codependence or whatever, because I think that that unnecessarily vindicates a lot of people and anchors them on the inhuman or unhuman end of the spectrum. And that's actually holding more people back probably i think that though the the danger like for example we built compass that will help you prepare for a hard conversation and after you've gone back and forth with compass a few times it tells you stop talking to me and go talk to the person that you need to talk to because sometimes we found that people were enjoying the role play and the problem with the role play is that you've told compass what you think about that other person so compass is adopting that persona but that may be completely incorrect it may be teaching you to say it more clearly whereas what you should be practicing is for an emotional response with this other person and so we've tried to build some warnings in also people will type in saying i need to text this person some feedback, evaluate what I'm writing.

44:18And it will say, I won't. You need to go talk to the person. Do not send this text. And so I think that an example, Russ Laraway wrote this great book, When They Win You Win, and he has a career conversation methodology. And most people's bosses are not taking the time to have career conversations with them. And most people can't afford a career coach. But I think rather than trying to build a bot that'll do it with you, it's much better to say, why don't you go have career conversation with a colleague, with a friend? Because you don't have to have a boss or a coach. Find a friend, find a real person to talk to.

45:01If you can't afford therapy and you're struggling with drinking, go to an AA meeting or go to a smart recovery meeting. Like, I think that there's a time and a place to remind people to go talk to people, right? Yes. And to stop talking to the AI, because the AI is not a person. I had dinner with a bunch of people who were using AI, and one said, oh, AI is my co-founder. And another person was literally, like, it was scary to me how they were using it. So maybe I'm sitting here in Silicon Valley and hearing about people on this nutty end of the spectrum and that's not the real problem.

45:38Jeremy Utley: I'll, I'll give you a funny story to end if you want, or you can see what you think about this. But speaking of spectrum, I had a friend tell me the other day, I thought I was going to run into Starbucks today, which I think is strange because he lives in another state. And I go, why'd you think you're going to run into me at Starbucks? He goes, well, the barista had a latte for AI and I just figured that was you. I go, okay, that's insulting. And I go, well, what happened? He goes, well, I got curious, like who, who's going to be AI, you know? And he said, you know, the barista is going, I got a latte for AI.

46:07Jeremy Utley: Nobody. He said, triple shot latte for AI. He said, this very unassuming guy, almost looked like the brawny man walks up and goes, it's just Al.

46:20Jeremy Utley: And to me, it's like the perfect, it's a perfect illustration of the spectrum, right? There's like people who are the barista, it's like AI is everywhere, right? And then there's just Al who's like, just wishes we'd all shut up about it already. I love it. All right. You can call me Al. We'll end on that one. I love it. Well, Jeremy, everyone should read your book, The Human Advantage. And thank you for demystifying AI and encouraging people to use it to outsource their grunt work so that they can have the real human relationships. And the more, maybe the last thing we should say about that is the more experienced you are, the more you stand to gain from the collaboration.

47:00Jeremy Utley: So don't think this is a young person's game. Don't think this is about being digital native. If you have the hard won skills to get great work with this weird intelligence called people, you've got the skills you need to get great work with this weird intelligence called AI. I love it. All right. Thank you so much, Jeremy. Where can folks find you in addition to buying your book? Oh, jeremyotlee.com. There's a bunch of pre-order bonuses there that I'd be happy to extend to folks who hear this conversation. And then, of course, LinkedIn, YouTube, all the places. All the places. All right. Thanks so much.

47:32Jeremy Utley: Thank you.

47:36Jeremy Utley: This podcast is based on the book, Radical Candor, Be a Kick-Ass Boss Without Losing Your Humanity by Kim Scott. The Radical Candor podcast theme music was composed by Cliff Goldmarker. Follow us on LinkedIn, Radical Candor the Company, and visit us on RadicalCandor.com.