← The Daily Stoic
Loading summary
Transcript51 lines
- [00:00]
A
Welcome to the daily Stoic podcast, designed to help bring those four key Stoic virtues, courage, discipline, justice and wisdom, into the real world. Look, if you hated the taste, if it didn't feel good, if you weren't getting good results, if it didn't guarantee a dopamine hit, it wouldn't exactly require much discipline to steer clear. That's the whole point of the virtue though, isn't it? That it's requiring you to resist an impulse or forego a pleasure. As we said recently, discipline is about doing what's hard. Just as courage is the triumph over fear, discipline is the triumph over another lower part of our nature. A lot of times discipline is pushing yourself to do something you don't want to do. But the other part of it, which we might call the temperance part, is not doing the stuff you want to do. And in some ways, this takes the most strength. Seneca said that we're all slaves to one thing or another, sex or ambition or attention or chaos. And by indulging in these passions often enough, eventually we lose the freedom to abstain from them. We just can't not. We need discipline to push through that, to resist the urge to keep going when it's not serving us, to step back when our ambition tempts us to overreach, to recognize that rest and recovery and restraint are not signs of weakness, but also of strength and wisdom. True discipline means knowing when to stop. It's having the courage to say no to the extra hour, the extra project, the unsustainable pace so that we can sustain ourselves for the long haul. Discipline in this way isn't just about action. It's also about control. Control over our impulses, our desires, even our most deeply ingrained habits. Without this, we are slaves. Only through temperance can we ever be free. And I talk a lot about that. And discipline is destiny. I didn't just want it to be a book about do this, do this. There's a bunch of things we need to stop doing. And discipline is understanding that. You can grab a signed copy@store.dailystoic.com but if you are looking for some better overall habits, you might like our daily Stoic course, Habits for Success, Habits for Happiness, which is a six week deep dive into better stoic habits, which you can sign up for right now or you can join us in Daily Stoic Life and get it and all our other courses for free. I'll link to that in today's show Notes. They're saying that every day your business is late to AI, you fall two days behind. So how do you keep up when the competition is moving faster? Well, there's NetSuite next. You probably know NetSuite. It's the AI powered business management suite that securely connects all your data. And it's a unified suite that brings your financials, inventory, commerce, HR and CRM into a single source of trust trusted by over 43,000 customers. And Netsuitenext is the next huge leap in how business gets done. Because AI is built into everything you do, it automatically surfaces custom insights throughout your day. AI agents work alongside to help you solve problems and handle routine work. And anytime you have a question, you just ask, just like having a conversation with a colleague. NetSuite is customized for a wide variety of industries, so it supports the way your business truly works, whether that company earns millions or even hundreds of millions of dollars. It's time for Netsu Next. For the first time ever, you can try netsuite Next for free. If your revenues are at least in the seven figures, go to NetSuite AI/STOIC. It's built for every industry, ready for every boardroom. Netsuite AI slash Stoic. We're traveling internationally this summer and some of those places, like in Europe, you know, you can't just go dress like your slobby American. You got to dress up a little bit. Can't show their shoulders, can't wear shorts. So I'm going to take some of the stuff I bought from Quince. It holds up well, looks good, it's fancy without being uncomfortable. I'm going to try to pack clothes that are light, airy and comfortable. I can use them in multiple settings. You've certainly seen me in some of these things. If you've ever watched a Daily Stoke video or seen me talk live, I want something that looks good on stage that I'm not going to sweat through. It's not going to get super wrinkled. Quint has got great T shirts, they've got great light sweaters. And everything at Quint is priced 50 to 80% less than similar brands. They work directly with ethical factories and cut out the middlemen. So you're paying for quality, not brand markup. And Quince goes way beyond clothing. They've got sofas and ceramic cookware, premium bedding. It's the kind of brand you end up recommending to everyone for everything. Elevate your summer wardrobe. Go to quince.com stoic for free shipping on your order and 365 day returns. Now available in Canada too. That's Q u I n c e.com stoic for free shipping and 365 day returns. Quince.com sto. So what does AI actually do? Like, what is it for? Does it make us better? Does it make us faster? Does it make life easier? Or is it here to replace us? Is it making us stupider? Is it ruining the world? Is it undermining the sort of basic skills that we have to develop as people and then we have to continue to develop as people? You know, sort of being thoughtful, being reflective, thinking, researching, practicing our skills? Right, these are the questions, right? These are the questions that if the people in the AI companies are not asking or if the politicians are not asking, then, then we ourselves have to ask personally and professionally. And it's actually one of the things I asked today's guest, one of my friends, David Epstein, about. I'm raved about David's work for years. He's the bestselling author of Range and the Sports Gene. He has this awesome new book called Inside the How Constraints Make Us Better, which came out on May 5th. Loved it. I actually read it back in January, so I had to sit. Speak of constraints. I couldn't tell you how awesome it was until closer to it coming out. But in today's episode, we riffed a little bit on AI, how we use these tools without letting them use us. How you have to use your brain first. Remember that story I tell on Wisdom Takes Work from Seneca about where he says, like, nobody becomes wise by chance. You don't become good at anything by accident. If you don't develop these skills, they atrophy. That doesn't mean you can't use tools. That doesn't mean there aren't some ways that it can make you better. I think that's in fact, quite clearly the case. But anyways, David and I are talking about using the brain first, then AI, how David uses it in his work, and how we need these constraints if we want technology to help us without becoming a crutch. We shape our tools and our tools shape us. And that's what this conversation is about. Like I said, if you haven't read David's work, you absolutely should. He's been on the podcast a bunch and this is just a short little episode with him and I talking about a very important topic in the news. Talk to me about your thoughts on AI, because what it does is it makes a lot of really hard time consuming things easy.
- [07:21]
B
Yeah.
- [07:22]
A
And so is that a good thing or a bad thing?
- [07:24]
B
Yeah, a few things. I mean, we were talking about desirable Difficulties, things that make learning more difficult. And so far this, this kind of first, let's say like, first run of research on AI looks terrible from that respect. There was an MIT study that got a lot of headlines talking about brain rot, how using ChatGPT will rot your brain. But there was actually more nuance in the study. So, so these MIT students had to write essays on different topics and, and they were broken up into groups. One group had to just write using their brains and one group had write using ChatGPT. They could use ChatGPT, they didn't have to. And they write on these topics and then they're brought, I think they wrote three essays and then they're brought back sometime later and they have to write on the same topics again and they're asked about the things they wrote and this time they're switched. So the Brain first group can now use ChatGPT and the ChatGPT can now go to brain. And basically everything for the ChatGPT group looked worse. They didn't remember anything they'd written. Their essays were more uniform and homogeneous and less interesting. Whereas the group that used ChatGPT second did, did quite well. You know, there, there were stuff about their brain activation, but I don't like to lean on FMRI data. So I think the rule that comes out of it as suggestion would be brain first, tool second. And there's just this mounting evidence that's seeing that people are using it to kind of replace their thinking instead of to supplement it. And that that's bad because then they do not learn. Like if, if they have it right, their first draft, they have real trouble recalling any of the content, it's actually in it. Even if they're like. And then I go over it, right?
- [08:54]
A
The point of writing the essay is, is to write the essay, not to have the essay. It's to force yourself to do a hard thing, even if it's somewhat arbitrary thing, right? It's to force you into this structure that requires you to think and comprehend and communicate in a certain way. And then you, on the other side of it, yes, you do have the essay, but more importantly, you have the skills of having done the thing.
- [09:21]
B
A huge portion of the writing that everyone does, especially when they're younger, is not because you care about getting the best essay. Yes, it's because you're learning how to make an argument, how to think. You're learning the material that's in there. You're learning how to organize things coherently. It's not because you need the most Polished essay in the world. And so I think this. So you know, we're still early days for this research, but so far it's looking pretty bad for how it affects people's thinking. I think there are other things I find it fascinating, like there's a research use I have for it that turns something that would take me 10 hours into one. So I use this site where I can upload a paper, scientific paper, and it will immediately create a map of other papers that cite my paper with bubbles of different size showing how cited those papers are. So I can see which are the real big deals in this field. To cite my paper, it'll try to automatically organize them into supporting or refuting. I'd say it's right, like say 75% of the time. And then it will make a list that excerpts a few sentences around the mention of the paper that I'm interested in. That's something I used to control f my way through PDFs for, for hours and hours and hours. So that's amazing. And I mean, I find it interesting in all sorts of ways. At the same time, I think there's that issue with the undesirable ease that it does at thinking where it's, you know, it depends how you use it.
- [10:33]
A
The task is supposed to be hard.
- [10:34]
B
The task is supposed to be hard. That's the nuance that was lost in the brainrot headlines, which is that it did actually show that if you used brain first, it was good. The other thing I think is it just gives people an infinite capacity to just start a bajillion different things and to create, you know, now you're hearing this term work slop to just create a ton of mediocre stuff that buries the people who, hey, can you research this for me?
- [10:54]
A
And then they give you 80 pages of shit they printed out as opposed to thinking about.
- [10:58]
B
So it's like creating more work in the system. It might offload from one person, but if there's this cartoon where someone is like, AI is great. I gave it these four bullet points. It turned it into a six page email. And then the other end, the person gets the email, is like, AI is great. It took this six page email and turned it into four bullet points, bullet points. So I think there's a lot of that going on. And I think, you know, through the lens that I'm thinking of the world right now, I think there's going to be even more need for proactive constraints because you can start so much stuff and do so much stuff there's actually one AI company that I've spent a bunch of time with because I know the founder and I wanted to understand how these things work, how they train. So I sit in on their demos and all this stuff. I've talked to other companies that use them and one of the things I've noticed is most of these companies are saying, I need AI. I know I need AI.
- [11:42]
A
They.
- [11:43]
B
But they haven't. So they implement it, but they didn't identify a problem.
- [11:47]
A
Yes.
- [11:48]
B
And define it well, that the tool matches.
- [11:50]
A
Right.
- [11:51]
B
And so they'll end up implementing it
- [11:52]
A
and it starts generally instead of specifically.
- [11:55]
B
Yeah. And they're, they're, you know, generating more of this work slop and all this stuff. And so I think we're often skipping over the. Define the problem and then let's see how the tool fits with this. Because there's this race, you know, you feel so behind if you haven't implemented it everywhere.
- [12:08]
A
Yeah. I was thinking about AI through the lens of your last book about range. Right. Because obviously if you only do a very specific thing, the chances of it replacing you, I guess would be higher in some ways. Right. Like we spent the last 15 years telling everyone to become coders and now it's really good at coding.
- [12:27]
B
That was like supposed to be the safest job five years ago.
- [12:29]
A
That sounds great. But I've also found it in my use of it, which is that if I have a decently broad base of knowledge about an area, actually I find the AI is not that great because I'm able to spot all the ways that it's wrong. And it's in fact where if you're a specialist, you're actually more vulnerable to this because you don't know what you don't know. And then you're more reliant on it. So there's some stuff it does really good. But I'm finding to be good at it, what you have to have is kind of a broad liberal arts knowledge base. Recognizing slop requires a lot of knowledge.
- [13:06]
B
Yeah. I think it requires actually knowing a ton of stuff.
- [13:08]
A
First of all, you're like, this is dogmatic.
- [13:10]
B
Knowing information.
- [13:10]
A
Yeah, yeah.
- [13:11]
B
Which people think is just obviated now. And just having a good sense of kind of information hygiene. Like where to say this doesn't sound right. Two things to your point. First of all, I had this experience recently of conversing with AI, I guess, and it brought up a fact that I was like, how did you know that it was about the number of Charles Darwin's pen Pals that he kept, and he would, like, hit them up for information on. And I was like, man, that is a very obscure fact that I happen to know because I read this specific paper that went into range for it and then it said. And then it talked about the researcher that studied his pen pal correspondence. I'm like, that is so weird. I know both of these things and that's the right fact, but the wrong researcher. And it's creating this whole narrative about it. And so I started asking, what's your source? It turns out it's me. It was excerpted, somebody. Somebody put up on Goodreads or whatever, some part of my book. And it had taken two things that were close in physical proximity, mashed them together, and then like, filled it out with a narrative, basically slop on top of it. Yeah. I'm like, you would never, like, I might be the one person that would have caught this particular thing, right? Yes. So there's that aspect of it. And also, I mean, one thing that's happened range is kind of having a little bit of a second life. Because the. The beginning of the book is framed early in this, like, looking at kind and wicked learning environments in the world. Like these situations that are sort of more stable where you solve the same problem over and over versus those where you have to kind of reinvent and learn new things all the time. And increasingly we're all in that wicked world where, like, work next year won't look like work last year. And the argument it makes is that people with breath are particularly prepared for that kind of world and even gets into, you know, using human computer chess partnerships as a model, like, where things are going. And I had no idea where we were going to be with AI now. But people have started to view that as prescient as this argument for, like, these synthesizers being more powerful and like, anthropic just said they only want programmers with side quests. And the Financial Times just did this big analysis that showed that the returns to, like, collaborative and coordination skills are accelerating and how fast they're outpacing, like, just technical skills. So I think we're gonna need people building the LLMs, of course, are specialists, but. But other people are going to have to learn broadly and reinvent if they want to not just be producing a
- [15:32]
A
ton of slop when you think even, like, what they call sort of query management or prompt engineering, but, like, so if you're like, hey, draw me a picture of this. Is going to give you some random slot, basically. But if you're like, if you're able to go, hey, I want a picture of this in the style of this artist from this era, by the way. I don't want it to look like X, Y or Z. Like, the more precise you're able to be. And the irony of that preciseness coming from your broad sense of all the different avenues that it could go, it's really important. Like, you have to have this sense of knowledge. I was having it do something. In this book I'm researching now I'm exploring, like this book, brief period in somebody's life when he was on a college football team. And I have this letter where he's listing some of the people he remembered playing with. And so I upload this chunk of the letter and I'm asking, okay, what did these people go on to do? Right. And so this is 70 years ago or something. You know, some of the names are common, some of them are not. And it's telling me, oh, this person went on to do this, and this person went on to do this, this person went on to do this. And like, it makes this one confusion. You know, Stan Turner goes on to play, you know, has a moderate professional football career. And it's like, okay, no, no, this is clearly Stan's field. Turner, who goes on to be the director of the CIA and the Carter administration. Like, but I knew enough to go, hey, you've confused this person, this person. It is never going to be like, it could be this person or it could be this person. I don't know. It's going to confidently give me the answer. And I have to. Not confidently, but. But be just unsure enough that I go, are you sure you don't mean this? And then we're able, like, so you have to have this basis of knowledge that allows you to spot bullshit.
- [17:25]
B
Yeah.
- [17:26]
A
And then just like anyone that's ever used a research assistant, also, you have to have this meta understanding that you can never trust anything anyone gives you. Yeah, you have to. If someone is prepping stuff for you, you have to then go independently verify all this information or there's a. Of vulnerability in your system. And so like the ability to know, have that sort of just intuitive bullshit detector. And then. And then the knowledge base to. To know what's not working and not working. And then also the standards, which is like, I will never turn in or publish or show anyone anything that I haven't personally vetted. These are all skills that you're going to need to function in this.
- [18:08]
B
I think those take a while to develop though, too. When I think about some of the mistakes I made as an early journalist, I didn't recognize certain potholes that are obvious in retrospect until I like, got something wrong.
- [18:17]
A
Yeah.
- [18:17]
B
You know, or had to trust a source.
- [18:19]
A
And then your editor was like, well, what do you mean? You just printed what they said or whatever you learn and you don't even
- [18:26]
B
know some of the assumptions you make. Right. So in this book, I ran a lot of stuff I was writing through AI you know, for fact check. But I had a hired a human fact checker also.
- [18:32]
A
Yes.
- [18:33]
B
And I was a fact checker in the past, and I also went through it myself, like line by line, crossing out one line at a time. And when I finished the book, because I kept really good. In my first book, the Sports Gene, I was a little sloppy with citation. I was like, this is so interesting. I'll remember where this came from. And then it took me six weeks to recreate my sources and do bibliography with range. I kept better track, but still sometimes was trying to recreate something. This time it was like every line I annotated so I wouldn't. So this time when I finished with Inside the Box, I was like, do I need to spend ten grand on a fact checker? I think I was really careful this time. I probably nailed it. And then I fact checked myself. I probably made a change on every page going back through. Like, you just are not attuned to all the assumptions you make. And if you're slowing down to catch them all, you never get the thing done. Also.
- [19:20]
A
Well, isn't that one of the benefits though, to flip your book's premise on its head? Right. Like there are things that constraints would prevent you from doing. Right. So it's like, I would say AI is not great at fact checking.
- [19:32]
B
Yeah.
- [19:33]
A
But it's also free.
- [19:34]
B
Yes.
- [19:35]
A
So. So like I, Yeah, my last book, I ran it through or even stuff as I'm writing now, I'll run it through and I'll go, hey, what am I missing? Where could this be better? What's wrong? You know, does anything jump out at you as being incorrect? And I would say 85% of what it brings up is totally wrong. Like, like it, it, it caught this thing where it was like, hey, you know, it says this. That can't possibly be true. It's like, okay, well, I have a, a letter from 1962. Yeah, yeah. Like, where this is in fact exactly what happened. And its false positives are very high. But if it gives you one or two or three or ten Things that you wouldn't have caught otherwise. And all you're paying is this monthly fee. That's great. Right. You probably wouldn't have hired two fact checkers or three fact checkers. So you still do it the old way and then you get this other way on top. That's only upside.
- [20:25]
B
And in this case. Right, only upside because you're just saying, oh, I better take a closer look at that. Essentially you're not saying this is going in. Right. So it's just attuning you to things you may have overlooked, which I think is great. And I would say for people, for people that are listening to us and thinking about using it in this way, I would say use very, very small chunks when you put that in, because I think that people have a tendency, if it spits out a long answer.
- [20:46]
A
Yeah.
- [20:47]
B
Like, the more text you're looking at at once, the less meticulous I think you're going to be about, like questioning the assumption. If it gives you a few sentences, I think you're going to focus in and question the things that it said. If you're like, fact check these 20 pages and it spits out 20 pages, I just don't think you're going to be as careful. So I would do it in very one chunk, two chunk, which again, is hard because you could throw in hundreds of pages and say, fact check this thing. Right.
- [21:10]
A
Well, and then the other interesting thing about constraints, right, Is like, so you can have read this section, give me your notes, and then it does it. And then you can do as you wouldn't do with a person because you're not a monster. You'd be like, do it again, do it again, do it again. And what's remarkable is how you'll get different answers each time. And so again, the lack of constraints, the reduction in cost. Let's put aside the environmental impact, which I don't know enough about, to speak authoritatively, but you can, you can have it do the task many, many times. Like, so if you're like, hey, I want a picture for a presentation. I want it to look like this, I want it to look like this. And then it does it and you go, okay, try it again. Yeah, actually, give me three options.
- [21:47]
B
Make it better.
- [21:48]
A
Give me. Right, you just say, make it better. And you get. And so the reduction in constraints allows you to get things that you wouldn't get out of it. A human, which is five options. And then you can mix and match them together. And. And there's benefits there for sure.
- [22:02]
B
And I mean, that's like the march of a lot of efficiency in a lot of our tools. Right. Or what Clayton Christensen called empowering innovations. They bring something that, you know, maybe someone with a staff could have done, but now someone without a staff can do it much more cheaply. It's like where, you know, you used to have to have, like a desktop computer, and then it gets cheaper and everybody can carry one around with them. But I do still think it's incumbent upon people to put a lot of constraints into that process, or else you just end up with the work slop stuff and try and outsourcing your thinking completely, which is a whole other problem. Like, I wonder for some jobs now, you know, so five years ago, coding is, you learn to code, you'll never need another job. Now it's like the least safe thing, maybe, other than some of what we do, I don't know. But I wonder how do you get kind of senior coders if you don't have junior? Like, where are the minor leagues for a lot of these things? Or when I didn't end up working as a. I mean, I did, when I was a grad student, work as a scientist.
- [22:55]
A
But how do you get your hours to be the senior person?
- [22:57]
B
Like, I developed in. So much of my work has been based on research, peer reviewed research that I thought was wrong, that I picked out because I had a strong quantitative background and I got that intuition from being in grad school for geology. Right, right. And so it's like, where do you. If you just skip over that, you don't really get that intuition that allows
- [23:15]
A
you to pick out stuff that's wrong.