5 Minutes a Day For Peak Productivity? - This Simple Hack Might Change Your Life

00:00:00.000 | So I've been writing about time management for a long time now. When I was 22 years old,

00:00:06.720 | I wrote a book called How to Become a Straight-A Student. Here it is for people who are watching

00:00:11.780 | instead of just listening. It was supposed to be a book of no-nonsense advice about how to get good

00:00:18.100 | grades in college. Had a quiet start. This book's actually quietly done well in the background. I

00:00:23.380 | think it sold like over 300,000 copies quietly since it came out in 2006. Anyways, this book

00:00:28.740 | includes a chapter. I'd forgotten this, but I saw the other day, this book includes a chapter

00:00:33.180 | where I describe a time management system for students. And the title of that chapter

00:00:38.680 | is How to Manage Your Time in Five Minutes a Day. This was written back 20 years ago.

00:00:45.700 | So I thought what would be interesting would be for me to go back and reread that chapter.

00:00:49.580 | Let's go back and revisit my ideas from 20 years ago about how to manage your time. Because what

00:00:54.600 | I'm really curious about is to see where was I on the something that remains true today? And where

00:00:59.720 | are there, where is there advice from 20 years ago that no longer holds to today? So what works from

00:01:04.120 | before and what doesn't? I decided not to really revisit this until we were live here on air. So

00:01:09.080 | this should be, should be interesting. God knows what's in here, Jesse.

00:01:13.040 | I'm looking forward to it. Yeah. It's going to be a lot of, let me tell you this by 2025,

00:01:18.240 | our biggest problem is going to be finding air parking for our flying cars. And let me tell you

00:01:24.820 | the, I guess I was going to do a whole like Donald Trump thing, but I only know, I guess he was on TV

00:01:28.940 | then. I don't know. All right, let's get into it. I'm going to open my book here. Page 19.

00:01:37.900 | Boom. All right. Manage your times in five minutes a day. I'm nervous about this. I haven't read this

00:01:43.560 | in a while. Here's the intro to this, this chapter. Real straight A students, like most reasonable

00:01:48.000 | students, hate time management. After all, college is supposed to be about intellectual curiosity,

00:01:52.980 | making new friends, becoming obsessed with needlessly complicated drinking games. See,

00:01:56.580 | that was social proof, Jesse. I was proving to the reader. See, I'm young too. I'm on your page

00:02:01.260 | drinking. An overwhelming interest in time management is best left to harried business executives.

00:02:06.080 | Oh man, that's me now. Or perhaps pre-meds at the same time. However, you can't abandon

00:02:10.600 | all attempts to keep tabs on your schedule. Uh, and so I'm going to give you some time management,

00:02:15.820 | blah, blah, blah. All right. So the setup is like, we're reluctantly going to do time management. We

00:02:21.220 | don't want to spend too much time thinking about this. So I guess I was really worried about students

00:02:24.540 | saying like, look, man, this is square. Cause that's the way we talked back then, Jesse.

00:02:27.820 | This is square, man. I'm looking for tubular advice. I don't want to hear about time management.

00:02:33.820 | So I was being careful about it. All right. I lay out now in the chapter,

00:02:36.920 | here are my criteria for a time management system. Four things. One requires no more than five to 10

00:02:43.920 | minutes of effort in a single 24 hour period. Two doesn't force an unchangeable minute by minute

00:02:48.960 | schedule on your day. That's interesting. That changed. Three helps you remember, plan and complete

00:02:53.900 | important tasks before the very last moment. And four can quickly be restarted after periods of

00:02:59.380 | neglect. Interesting thing I'm noticing already, Jesse, is like this tone reminds me of early Tim

00:03:05.480 | Ferris. But this is before, for our work week is before I knew Tim or I'd read his book. So there's,

00:03:11.700 | it must've been in the air back then. That's sort of really declarative, you know, like ambitious,

00:03:17.160 | like we're going to do this. We're going to do that. It was a tone that was in the water back then.

00:03:21.040 | All right. What you need. So my system requires a calendar and a list. And I say,

00:03:27.520 | the list is some piece of writing material that you can update throughout the day.

00:03:30.200 | You do have to carry this with you. So make it something simple, like a sheet of paper ripped

00:03:34.040 | out of a notebook each morning. So one sheet of paper each day and a more permanent calendar.

00:03:39.000 | I said the permanent calendar could be physical or it could be digital. All right, let's get into

00:03:43.940 | the details of this system from 20 years ago. Record all of your to-dos and deadlines on your calendar.

00:03:50.200 | This becomes your master schedule. The one place that stores everything you need to do.

00:03:55.940 | The key to our system, however, is that you need to deal with your calendar once every 24 hours. Each

00:04:00.980 | morning, you look at it to figure out what you should try to finish that day. Then throughout the

00:04:04.560 | day, whenever you encounter a new to-do or deadline, simply jot it down on your list. The next morning,

00:04:09.420 | you can transfer this new stuff from your list onto your calendar where it's safe.

00:04:14.240 | And we're back where we started. The whole system can be summarized in three easy steps.

00:04:19.160 | Jot down new tasks and assignments on your list during the day. Next morning, transfer these new items from

00:04:24.600 | your list onto your calendar. And three, take a couple of minutes to plan your day.

00:04:29.360 | All right. I then give an example schedule for a day. So I guess you look at your calendar each

00:04:37.400 | morning. You write down on your sheet of paper your schedule for the day. And so the example I had in

00:04:42.380 | here was on one side, a list of appointments. So it was like 10 to 12 econ. 12 to 1 lunch with Rob.

00:04:51.140 | 1 to 145 government reading. 2 to 4 government class. 4 to 530 finish government reading. 530

00:04:57.360 | to 630 start French essay. And then I had a list called things to remember. And I guess these are

00:05:03.020 | the things that you would jot down during the day. Econ study group moved. French quiz laundry. Start

00:05:09.900 | searching summer internship opportunities. So this is interesting, Jesse. It looks like the plan you write down

00:05:17.500 | includes basically rudimentary time blocking. Oh, it was. Okay, here we go. So let me look at this a little bit

00:05:26.600 | closer. So when you're making your plan for the day, I'm jumping ahead here. Look at your calendar entry for the

00:05:31.540 | current day. It will probably contain a handful of appointments and to-dos. Your goal is to figure out how much of this

00:05:35.860 | work you can realistically accomplish. You might be tempted to simply copy all of these tasks into your

00:05:40.520 | today's schedule column and then treat it as a simple to-do list for the day. Don't do this. If you want to

00:05:45.320 | avoid getting overwhelmed by your work, you need to be smarter about your time. Here's what you should do

00:05:49.580 | instead. I bolded this. Try to label each of your to-dos for the day with the specific time during which

00:05:55.020 | you're going to complete it. Be honest. Don't record that you're going to study for three hours starting at

00:05:59.840 | three if you know you have a meeting at five and be reasonable about how long things really take. I also

00:06:05.920 | say you can batch together little errands and you should try to end your day at an appropriate time.

00:06:11.600 | The only other thing that's important here is when you update your calendar

00:06:20.700 | year each day, you move the stuff that you didn't get done on to other days.

00:06:28.460 | So let me look at this example here. This was the kind of tricky part back then.

00:06:33.020 | Blah, blah, blah. Blah, blah, blah. Okay. So not everything gets done. The things that doesn't get done

00:06:41.160 | have to exist somewhere. And so you move them on to another day when you think you're going to get it done.

00:06:47.840 | Interesting. I'm interested by myself. You would appreciate this, Jesse. On one of my sample list

00:06:54.120 | here, I have one of the entries from five to seven, quote, get huge. I was being a little bit facetious

00:07:02.040 | there. I think that means go to the gym. All right. Interesting. So what were the features of this? And

00:07:07.820 | then what do I think works? And what do I think doesn't work? All right. So back then I was saying,

00:07:11.680 | I'm just going to summarize all of this, right? All of your tasks need to have a place to live. And back

00:07:16.580 | then I said, why don't you have them live on your calendar? So obviously like appointments and stuff

00:07:20.680 | exist on the specific days and specific times when it happens, but to-dos, you put them on the day you

00:07:25.120 | think you're going to do them. And then when you get to a current day, you have some stuff on your

00:07:30.660 | calendar that's timed, like classes and meetings, but you also are going to have this list of things,

00:07:34.940 | to-dos that happen to be assigned to that day. Then you make roughly a time block plan. That's what I was

00:07:40.900 | talking about way back then, 20 years ago. Make a plan for your hours of your day, like when you're going to do

00:07:45.100 | what, when your appointments are, see if you can fit the things, the to-dos you had on your list.

00:07:49.660 | Can you fit them in there? You might not fit everything, or you might try to fit everything

00:07:54.520 | and not everything gets done. Well, that's fine because what you do is the next day,

00:07:59.860 | when you create your new plan, you're going to take everything from your list from the day before

00:08:04.800 | that was added new, this capture stuff and put it places on your calendar. This is also when you'll

00:08:08.700 | take the stuff you didn't get done and make sure those show up for a new day. And maybe you move them to

00:08:12.320 | the current day, or you look into the future to find a day that's less crowded. And so in this plan,

00:08:17.080 | you make a plan, you do your best, you capture stuff on your sheet of paper throughout the day.

00:08:20.420 | The next day, the new stuff and the stuff you didn't get done goes back to your calendar,

00:08:24.580 | and then you make a new plan for the day ahead. So what I like about this plan,

00:08:29.760 | and then I'll get into what I think is missing. I like the full capture methodology. Now, I can tell

00:08:35.360 | you, I had just read Getting Things Done, David Allen's book, not long before this, that book came

00:08:40.420 | out, I think in 2004, maybe 2003. I wrote this book largely in 2005. So I had just read that. And so

00:08:47.660 | clearly that full capture idea of everything you need to do needs to be written down somewhere. There's

00:08:52.680 | nothing that you only are keeping track of in your head. That suffuses this system. And so I like that.

00:08:58.660 | I still think that's a very good idea. Things you're only keeping track of in your head either

00:09:02.400 | cause stress or will be forgotten. The other thing I like about this, and this was obviously a big

00:09:09.280 | emphasis in my student book, low friction. I was really careful that when you go through your day,

00:09:16.800 | you just have like a sheet of paper with you that you're glancing at to be like, what should I do next?

00:09:20.360 | And you can just jot things down on. It's a paper in your pocket, right? No systems, no logging into

00:09:25.140 | anything, no complicated planners back. It would just jot things down on this paper.

00:09:28.640 | And the only time you really have to grapple with tools beyond that sheet of paper is once every

00:09:34.080 | morning where you say, okay, let me move things back into my calendar. Let me update my calendar.

00:09:38.720 | Let me make a plan for the day ahead. And that was supposed to take only five minutes. So I like that

00:09:42.120 | simplicity. I like the planning, the rudimentary planning in this system too, of tasks go on a day.

00:09:50.360 | So you're kind of making a plan like this day looks light. Let me try to do things there.

00:09:54.480 | As opposed to just having a bunch of to-dos and saying, let me try to get some of these done

00:10:00.580 | today. So it was a much more of an intentional, that's a time block planning philosophy, being

00:10:06.560 | intentional about your time as opposed to just being reactive or saying, hey, what should I,

00:10:09.460 | what should I work on next? Hey, it's Cal. I wanted to interrupt briefly to say that if you're enjoying

00:10:14.540 | this video, then you need to check out my new book, Slow Productivity, The Lost Art of Accomplishment

00:10:21.520 | Without Burnout. This is like the Bible for most of the ideas we talk about here in these videos. You can

00:10:29.320 | get a free excerpt at calnewport.com slash slow. I know you're going to like it. Check it out. Now let's

00:10:37.400 | get back to the video. This plan works really well with something I talked about later in this book

00:10:42.400 | and I still talk about today, which was put your regular work on your calendar as well. So this

00:10:48.340 | type of planning for a student works really well when regular work like reading assignments, problem

00:10:53.340 | sets, labs, stuff you know is due on a regular basis. When you put the time when you're going to work on

00:10:58.640 | those things on your calendar in advance for the whole semester. And so then that just becomes part of

00:11:04.140 | your day. You're not even thinking about studying is just the same as like going to class. I have

00:11:08.240 | class at 11. I have studying from 1 to 2.30 on this type of reading. So making that more automatic

00:11:13.760 | worked really well with this system. Another thing I added later, I think after this book came out,

00:11:20.000 | I talked about this on my blog and it's a little known fact that, well, maybe somewhat known, Jesse,

00:11:24.200 | but my blog, Study Hacks, which now became my newsletter as well, originally started in the aftermath

00:11:28.660 | of this book as a way to discuss ideas that didn't make it into this book that were about this book.

00:11:33.140 | It was the missing chapters for the straight A student book was its original purpose.

00:11:36.840 | And I, there I introduced the idea of, you know what, at the beginning of the semester,

00:11:40.620 | do an extra planning. So this is the origin of multi-scale planning where you go and find all

00:11:44.840 | your major deadlines and start working backwards from those deadlines and putting markers on your

00:11:49.120 | calendar. Like three weeks before midterm, study plan for the midterm, you know, a month before the

00:11:55.140 | term paper, make your plan for when you're going to write this midterm and then get that onto your

00:11:58.860 | calendar. So it was all about like when you encounter your day, the stuff was there that you need.

00:12:03.120 | Okay. So what's missing here? Well, obviously the task volume is very low, which would make sense for

00:12:08.840 | a college student in the early two thousands. If you were, let's say a standard 2026, 2025 knowledge

00:12:16.680 | worker, your task volume is going to be much, much higher because of tools like email and Slack.

00:12:20.900 | That's not even dealt with in here. The task are things like do your laundry or pay your cell phone

00:12:27.380 | bill. You had a sort of very reasonable amount of tasks. So of course they could live on your calendar

00:12:30.860 | in the modern professional context. You might have many hundreds of different things to do. So to try

00:12:35.960 | to like assign everything to what day you think you're going to do it, that does not make sense with

00:12:39.980 | modern, uh, modern task volumes that you would see with sort of modern jobs. Um, I'm time blocking,

00:12:48.040 | but I'm doing it roughly. Whereas by the time I got 10 years after this book was deep work, I was like,

00:12:55.740 | no, no, just time block specifically, draw the blocks, block out all your working hours. Here,

00:12:59.500 | I was kind of roughly doing it. I was like, write down what to do as you want to do. And maybe like

00:13:03.660 | mark with the time you think you're going to do them that evolved into, you might as well just block

00:13:07.660 | your hours of your working day. Um, this is also missing shutdown rituals, which is very important,

00:13:12.480 | have a way of making a transition from I'm working to not working. That's very clear. I think that's

00:13:18.200 | particularly important in college context. You'll just work forever, but it's important in other

00:13:21.620 | work contexts as well. But all in all, I think this is actually, I am seeing seeds of lots of my ideas

00:13:28.740 | in this system and I like it simplicity. So where would this make sense? I guess if you don't have a large

00:13:33.120 | task volume, you're not plugged in the email, having to send them, receive 150 messages a day

00:13:38.660 | on 17 committees and working on 19 projects. If you have a pretty controlled task volume,

00:13:43.240 | a pretty autonomous schedule, something like this could work. Just make your calendar, your home.

00:13:48.060 | That's where your tasks live. You make a plan each day based on the calendar for the day,

00:13:51.760 | maybe throw in some higher scale planning, you know, once a month or once a semester, get some major

00:13:56.380 | deadlines on there. It works well. It really would take only five minutes a day. So there is a lesson

00:14:00.960 | here. It's like not every system needs to be the type of system that a super busy modern business

00:14:05.380 | person would need to tame their work. You know, some situations, your workflow is simpler, simpler

00:14:10.740 | systems work. As long as you keep the key ideas that show up here that have remained in my thinking

00:14:16.900 | about time management organization to this day, which is full capture, some intention over your time,

00:14:21.120 | as opposed to just working off of a list. And, um, basically that actually, those are the two

00:14:28.020 | things, full capture and intention about your time. And that's there. I guess maybe I'll throw

00:14:31.900 | a low friction in there and keeping the friction as low as possible. So I don't know. I think it

00:14:35.080 | holds up. Desi, what do you think? 20 years later? Yeah, I've read that book and I've gifted it to

00:14:41.260 | a bunch of kids that I coach that go off to school. Yeah. School is easier. School is easier. The main

00:14:46.900 | thing missing from this book is focus, training your mind. That's the missing chapter that today I would

00:14:52.320 | add if I was rewriting it. Um, you didn't start that till after your freshman year though, right?

00:14:56.980 | Yeah. Yeah. This was my, uh, God, what's the timeline? No, I graduated in 2004. This book,

00:15:04.300 | it came out in 2006. No, I'm talking about the process of you doing the, you doing the work.

00:15:09.680 | Oh. Like you studying that way. Yes, that's right. So that's right. So the, the, the, the backstory of

00:15:17.380 | this book, okay, how to become a straight A student, the, the original draft, and this is

00:15:22.300 | true of this, the final draft, like most of the ideas are mine, right? Because, uh, I did very well

00:15:27.440 | in school after my freshman year. So you're absolutely right about that. So my, my backstory was

00:15:31.800 | I went to Dartmouth college coming out of just like a public school. I wasn't coming out one of these

00:15:36.180 | elite private schools where they learn how to study. And so my first year was fine, but not great. I think

00:15:41.360 | I had to pass fail a course or two here to prevent like a bad, like a really bad grade. And okay.

00:15:45.440 | And I was rowing crew and I was kind of busy. And then I had this change where I said, I want to

00:15:50.100 | get serious about being a student. I got very systematic about my study habits. Why am I

00:15:55.540 | studying this way versus that? How am I managing my time and my schedule? And I experimented to figure

00:15:59.420 | out a set of tools that worked well. And that's when my grades got fantastic and got four O's every

00:16:04.300 | quarter, sophomore, junior, and senior, except for my senior spring in which I got one A minus.

00:16:08.400 | And it was in political philosophy, which is why I hate John Locke to this day.

00:16:13.820 | Damn you, Locke. I actually, it's true, Jesse, when we were doing the promotion for this book,

00:16:17.700 | I went back to Dartmouth and got my transcript. And so we would send it. We would send like copies of

00:16:24.160 | my transcript to prove that like I actually got all these A's or whatever. Yeah. So yeah. So then I

00:16:28.860 | figured out, so basically what I figured out is if you're systematic about being a student, by which

00:16:32.500 | like treat it like a business executive would treat their job, like really caring about how you do it.

00:16:37.640 | school wasn't that hard. Like you could do, you could do well without as much time as you thought.

00:16:44.320 | So basically what I did to pitch this book is I had gone to, I was an early inductee in the Phi Beta

00:16:53.400 | Kappa because I had such a high GPA. So like the top 1% of students, they would induct in the Phi Beta

00:16:59.080 | Kappa early. So I was, I was very, you know, I was, I think like top five in my class or something,

00:17:04.300 | right out of a thousand. I had a very high GPA. And that was back when it was harder to do that,

00:17:08.680 | especially in computer science. And so I went to this thing with like the top 30 students GPA wise

00:17:14.260 | in the class. And I was like, oh, wow, like I recognize some of these students. They're not

00:17:17.320 | grinds or whatever. And, and I interviewed them a little bit about like, how do you study? I think

00:17:21.120 | I interviewed everyone. I sent them an email interview and it was like based off my techniques

00:17:24.560 | and their techniques that I came up with most of the ideas in the book. But then after I handed in the

00:17:28.840 | first draft of the manuscript, my editor was like, uh, Ann Campbell is her name that she was like,

00:17:35.200 | I think, you know, we need to hear from other students, not just from Dartmouth and from your

00:17:38.620 | class at Dartmouth. So then I went out and surveyed, uh, straight A students from a bunch of schools.

00:17:44.600 | And then I integrated their quotes kind of into the book and I changed some of the, I learned some

00:17:49.340 | things from them and I used their ideas, but a lot of the ideas were in place, like before I interviewed

00:17:53.720 | them. So there's like, you know, there's a lot of quotes from students all thrown through.

00:17:56.400 | Those kind of were added later. So that's a little bit peek behind the curtain.

00:17:59.340 | How did you find the straight A students at other schools?

00:18:01.880 | I would look at things like, uh, back then my technique was typically look for

00:18:07.080 | press releases or articles from schools that would say things like

00:18:11.320 | Phi Beta Kappa class announced. And then what I would do, this was my technique for my first book as

00:18:17.320 | well, is all you have back then, all you had to do was find any student's email address and you

00:18:24.040 | could learn the naming convention. So you would find anyone's email address at Harvard and you'd be

00:18:29.460 | like, Oh, it's first name dot last name at FAS dot Harvard dot edu, right? Faculty of arts and

00:18:35.060 | sciences. And they're like, great. Then I'll see this other student's name on a press release about

00:18:39.980 | Phi Beta Kappa university or whatever at Harvard. So I'll just do their first name dot last name at

00:18:45.440 | FAS. I would learn to format and then I would cold email. And that used to work back then. Um,

00:18:50.580 | and that's how I would do it. And I got the interviews and I threw it in. Uh, so those are

00:18:53.800 | those, those are those ideas. But basically it was like, look, you treat, man, look at these old

00:18:58.160 | references, reality check bands, you treat school systematically. But you know what the thing is

00:19:04.440 | I've learned? I think this is true about a lot of jobs. Forget school. I think a lot of jobs,

00:19:09.960 | like you have some sort of reasonable system for organizing yourself, especially like in the

00:19:15.600 | first 10 years of a job, I think you're at our age jobs become more about like leadership and your

00:19:20.900 | ability to work with other people and other types of things. But like in the first 10 years of your

00:19:24.520 | job, 15 years of your job, you have some sort of system for organizing yourself, keeping track of

00:19:28.780 | what has to do some sort of way you make plans for what you're going to do. It really seems like a

00:19:32.540 | superpower. People are just like, you're amazing. Like it, people are so haphazard that having a plan

00:19:38.960 | can get you like the equivalent of straight A's. So I think that that's cool to see that,

00:19:42.980 | that system, I guess, more or less holds up. But that's the thing that's missing is back then,

00:19:47.180 | I never would have thought about focus. It didn't, we didn't have phones. When I started,

00:19:51.640 | most people didn't have laptops. There was no way to avoid, this goes back to our recent episodes,

00:19:55.840 | Jesse, there was no way to avoid getting lots of cognitive exercise. Like studying was boring.

00:19:59.760 | You had no distractions. You were at a desk at a carrel somewhere. There was nothing there to

00:20:04.000 | distract you. I mean, you could walk across the room to wait in line for a public computer terminal

00:20:08.520 | to check your email. Like that was it. So like people just got lots of, you just read and thought

00:20:14.060 | and studied and you never, you never thought about focus as mattering. 10 years after this book came

00:20:18.460 | out, especially 20 years after this book came out, this is like the number one problem faced by students

00:20:22.360 | is that they're so distracted that it's their brain is in a perpetual state of fragmented cognitive

00:20:28.000 | context, which makes it impossible to focus. So like that would have to be chapter. But I talk about

00:20:32.120 | focus in here as being important, but I don't see it as something like practice and protect. I just say

00:20:37.280 | if you, if you focus really hard, you get your work done faster, but I would have a much bigger

00:20:41.460 | chapter about now if I was ever to, was ever going to rewrite that. I don't really have time for that

00:20:46.420 | though. All right. So there we go. That's our deep dive. We got a lot of good questions. So let's do

00:20:52.520 | those. But first take a brief moment to hear from one of our sponsors. We'll talk about our friends

00:20:58.900 | at Cozy Earth. I love my Cozy Earth bamboo sheet set. So we have, my wife and I have two sets of

00:21:06.280 | these sheets. We've talked about this before. We like them so much. We want to make sure that when

00:21:09.320 | one set of sheet is, uh, is being cleaned, there's another one to be in our bed. Jesse, we just bought

00:21:13.940 | a third set of Cozy Earth bamboo sheets because here's what we were thinking about. We, uh, sometimes

00:21:18.820 | have visitors and like we were in, my wife and I went to Philadelphia and her parents watched the kids.

00:21:27.640 | so they stayed in our room. So we changed the sheets. My wife would like, we need a third set

00:21:31.040 | of sheets so that in that scenario, we can change back to a fresh pair of Cozy Earth sheets. So like

00:21:36.440 | our whole, when you get home. Yeah. When we get home. So our whole system is now like, there's no

00:21:40.620 | scenario in which we're not sleeping on the Cozy Earth bamboo sheets. That's really how comfortable

00:21:45.620 | they are. Softest, coolest, most breathable sheets you'll ever own. Designed to restore,

00:21:49.900 | refresh, and revitalize you every night. If I could just cut a hole in the top of the sheet and wear it

00:21:55.860 | like a robe, like a toga, I would. Jesse won't let me do it, but it's that comfortable.

00:22:00.160 | Um, they actually do have clothing as well. I have the sweatshirt. My wife has the pajamas.

00:22:04.920 | Uh, so they, you could get clothing that has the same comfortable fabric in it, but I'm telling you,

00:22:10.740 | when you sleep on these sheets, they're, they're so soft, they're cooling. We really have a hard time

00:22:17.340 | now sleeping without them. I a hundred percent recommend them. And then once you love those,

00:22:21.540 | get the pajamas so that you can kind of have the Cozy Earth feel, follow you everywhere else you go.

00:22:27.140 | Um, if you're on the fence, don't worry. You can try it risk-free for a hundred nights.

00:22:31.180 | Love it or send it back. No questions asked. There's a 10 year warranty on all their bedding

00:22:34.660 | product. You will love it. These sheets are very, very comfortable. Uh, I don't know. There's a few

00:22:38.500 | things I can endorse more strongly than I endorse these sheets. So make sleep a priority. Now visit

00:22:43.620 | cozyearth.com and use my exclusive code deep for up to 40% off Cozy Earth's best-selling sheets,

00:22:49.640 | towels, pajamas, and more. Oh, we have the towels as well. Hmm. Super soft. That's cozyearth.com

00:22:55.580 | code deep. And if you get a post-purchase survey, tell them you heard about Cozy Earth right here on

00:23:01.560 | the deep questions podcast. I also want to talk about our longtime sponsors at Grammarly from emails to

00:23:07.080 | reports and project proposals. It's more challenging than ever to meet the demands of today's competing

00:23:11.260 | priorities without some help. Grammarly is the essential AI communication assistant that boost

00:23:16.300 | productivity so you can get more of what you need done faster, no matter what and where you are

00:23:22.020 | writing. I really love Grammarly's product because it works in whatever tool you're using to write. It's

00:23:27.220 | going to help that writing be better. It's going to help you produce good writing faster. So Grammarly

00:23:33.900 | has been very smart, I think, about how they've been integrating generative AI into their already really

00:23:39.320 | good product. So now wherever you're writing, Grammarly can help you write a draft right away

00:23:45.780 | that you can start to work with. It can help you with things like tone detection or rewriting sections

00:23:51.800 | to be in a different tone or to be more professional, as well as like the standard Grammarly stuff of

00:23:56.960 | making sure that your writing is actually correct and clear and grammatically correct. It just makes your

00:24:02.020 | writing better. It allows you to communicate with more confidence. And now with the AI help,

00:24:07.780 | it speeds things up. There's a survey here that says 90% of professionals say Grammarly

00:24:11.460 | save them time writing and editing their work. 93% of professionals now also report that Grammarly

00:24:20.200 | helps them get more work done. So if you're in a knowledge of your writing, writing is important.

00:24:25.160 | Grammarly makes you a better writer. It's just that simple. So let Grammarly take the busy work

00:24:30.240 | you can focus on high impact work. Download Grammarly for free at grammarly.com

00:24:34.860 | slash podcast. That's grammarly.com slash podcast. All right, Jesse, let's do some questions.

00:24:42.340 | First questions from Alex. I'm starting a multi-scale plan. How specific should my quarterly plan be?

00:24:50.180 | For example, take long Saturday rides starting at 50 miles, increasing five miles each week,

00:24:55.520 | or bike regularly to prep for the 100 miler? That's a good question, Alex. I don't think

00:25:00.520 | we've had that one before, but it's a critical one. All right. So when you have a complicated long-term

00:25:05.620 | plan, the way I like to think about its interaction with multi-scale planning is that your quarterly plan

00:25:11.120 | gives you a pointer towards what you think is important on that time scale, but then you might

00:25:17.540 | have completely separate systems and rules and processes and tools that you use for that goal.

00:25:25.800 | Like for example, in this quarter, post my like surgery and other injuries, I've prioritized my

00:25:32.140 | quarterly plan. I call them semester plans because I'm an academic. It's pretty clear about

00:25:35.360 | getting back in shape in certain ways, right? That has with it. I mean, what I'm doing, I have a

00:25:42.420 | pretty complicated training plan and I have a trainer who's helping me and this and that.

00:25:46.160 | That's not all of my quarterly plan. My quarterly plan says we're focusing on getting back in shape,

00:25:50.720 | like generally aiming towards these goals. And then, you know, I have like my true coach app and

00:25:55.800 | working with my trainer and I have a particular way of a specific scheduling system. I'm using this,

00:26:00.540 | like these are the days I'm exercising and I kind of have a schedule worked out and this and that.

00:26:04.360 | So that's not all my quarterly plan. My quarterly plan is reminding me what's important.

00:26:07.880 | And then I have complicated other tools I'm using to track this. Like in your case, Alex,

00:26:11.860 | no, your quarterly plan would say prep for the hundred miler. And then however you track and

00:26:17.780 | train, you probably have a training notebook where you're keeping track of like what you're doing on

00:26:21.820 | each stage. You might have some training plan that you're following and you can keep track of that

00:26:25.260 | however you want to keep track of that. So yeah, you don't, your quarterly plan doesn't need to have

00:26:28.900 | the details of exactly how you're pursuing big goals. It's just a reminder

00:26:33.040 | that when you look at it every week, when you're making your weekly plan, oh yeah,

00:26:36.640 | I need to make sure that I'm making progress on these big things. And then whatever system you

00:26:40.840 | want to use to work on those other things, you can use those systems. All right, what do we got next?

00:26:46.820 | Next is Andy from Vienna. My son, 10 years old, reads mangas for hours. Does he had to reading

00:26:55.320 | mangas count as reading?

00:26:56.540 | I think it's manga.

00:26:57.220 | Mangas.

00:26:57.900 | Or is it manga? I think it's manga.

00:26:59.020 | Mangas.

00:26:59.900 | This is like Japanese novels.

00:27:02.040 | Yeah. I asked somebody about them, so they're like, yeah, I'm a little concerned on, but I love

00:27:06.680 | seeing him reading books, having books in his hands.

00:27:10.000 | This is actually like a longstanding debate. I was looking this up before.

00:27:13.140 | Does manga, I hope it is manga now that I'm saying it that way. You're going to get a lot

00:27:19.400 | of hate mail. Very popular. Does manga count as reading books in some sort of way? Here's the

00:27:24.640 | thing. It's good. Reading analog things is good, right? You're holding something in your hand that's

00:27:32.840 | not electronic. You're engaging in a story. There's a sort of visual literacy that you build up reading

00:27:38.080 | manga, trying to understand what's happening graphically. And there's a linguistic

00:27:41.780 | proficiency growing as well because the dialogue and the words, and you're seeing different people

00:27:48.680 | communicate. Some of the other things you get in novel reading is there to some degree.

00:27:53.320 | You're simulating the other minds of the characters in your own minds. There's some world building

00:27:57.200 | happening in your head. You have the visuals in front of you, but there's still a sort of like

00:28:00.060 | 3D version of that world that you're building in your head. There's complicated emotions that

00:28:04.880 | you're following. There's logic. Like it's good. I think reading analog things,

00:28:08.060 | is good. It brings you into a world that takes time. There's no algorithm flipping to random manga

00:28:14.280 | panels. You actually have to read it sequentially. So no, it's not, it's not a bad thing.

00:28:17.120 | Would you be like, should you be content with that being the only thing your son reads? I would say,

00:28:24.460 | no, I would also want to try to introduce more traditional text-only reading as well.

00:28:30.300 | They're both good. Text-only reading is going to do certain things though,

00:28:35.380 | that you're not going to get just in manga, right? I mean, first of all, the processing requires more

00:28:40.840 | intense. So it's more of a cognitive workout. You have to do a lot more in your head with text

00:28:45.440 | reading. So if you're reading a novel, you have to actually like construct a much more detailed model

00:28:49.960 | of what you're reading in your head. There's more, it tends to be a little bit more nuanced too,

00:28:54.280 | because you have descriptions of what's going on that are third person. So you get kind of like a more

00:29:00.200 | emotionally real world and you can do more sort of mental simulations and get more empathy

00:29:03.800 | and pure linguistic processing just requires more focus. So you're, you're practicing,

00:29:08.180 | focusing your brain more. So I don't know if we're going to draw a physical analogy. Manga is like,

00:29:13.180 | yeah, there's something that, that like my kid likes to do outside and it gets them moving and

00:29:16.540 | they're outside and it's better than just being inside. And reading might be like, they're actually,

00:29:20.620 | they're, you know, jogging or running or it's like, they're not just outside getting movement,

00:29:24.340 | but are like actually like training in a systematic way. It's, it's going to get you an even better

00:29:28.660 | shape than just being outside, moving around. I saw a suggestion somewhere. A librarian was saying

00:29:34.080 | there's certain like light Japanese novels that are a good stepping stone for manga. It's sort of

00:29:39.220 | similar themes, but in purely textual form. And that might be a good way. I mean, for kids, I'm all about

00:29:44.260 | matching books to interest. Like you, you, what you want is books, find the books that they're really

00:29:50.520 | excited to read and don't worry too much about what that genre is. Find what they're really excited to

00:29:56.900 | read because you want that exposure of not just reading, but being very excited about reading and

00:30:01.340 | getting pleasure out of it. You don't want it to be like math homework. You have to read your chapters.

00:30:06.020 | This is a good book and you're good because you got through these chapters. I've been a big believer

00:30:11.660 | in like, you really want to find, and it takes a lot of work. I mean, this is sort of my job in our

00:30:15.360 | house is because we have boys and I spend a lot of time because it really matters trying to find the

00:30:20.360 | book that's going to like spark each of the kids and they're going to be excited about reading it.

00:30:23.620 | And it's not always obvious. You'll try, this isn't working. This isn't working. And you have

00:30:27.680 | to find a thing that really works. And it's not always obvious. Like my, my 12 year old really likes

00:30:34.460 | not just Tom Clancy, but the, the really like Red Storm Rising. So the, the Tom Clancy books where

00:30:41.100 | it's not Jack Ryan going on adventures, but it's, we're going to just work through the scenario of like a

00:30:46.520 | world war, like all the military technology or this or that. And so like those sorts of books,

00:30:50.740 | he gets very excited about, right? It's a very specific genre, but fine. We'll like find those

00:30:55.760 | books. And, and, you know, that's, that's what you're going to read. My 10 year old likes a sci-fi

00:31:00.900 | more, but only specific. And so I'm constantly trying to search and try different books to try

00:31:06.160 | to see like what's actually going to work. Cause I just want them to be excited about reading

00:31:10.360 | and then just get used to it. And then you can read good stuff. I don't know. You can do that later.

00:31:14.380 | I read a lot of Crichton when I was growing up. Crichton, Grisham, all the sort of big nineties

00:31:21.060 | Clancy, all the big, there's like big nineties writers, genre writers. I was sort of raised on those

00:31:26.640 | and I came out. Okay. All right. Who do we got next? Next is Yolanda. How can I efficiently

00:31:33.540 | separate notes and reference information from actionable tasks in my productivity system without

00:31:38.620 | duplicating work? I use Trello for task management, but struggle with vague projects and scattered

00:31:43.880 | information across multiple tools. What strategies or rules can help streamline this process and keep

00:31:49.440 | everything organized? Well, Yolanda, my general approach is if what we're talking about is there's

00:31:56.440 | some information you need specifically to complete a task, like you're summarizing a report. And so like

00:32:03.260 | what mattered, the task is to summarize a report. So like the relevant information is that report,

00:32:07.060 | or you're trying to, uh, schedule something and the relevant information is like what people told you

00:32:13.480 | about their availability. I just moved that in the Trello onto the card corresponding to those tasks.

00:32:18.300 | If I could attach the relevant files to the card for the task that needs those files, I'll do it.

00:32:22.980 | If I can copy information from an email to a back of a card, and now I have the information there that I

00:32:27.900 | need, uh, I'll just do that. Sometimes if there's like a really detailed email thread, right? And so,

00:32:35.400 | and I have a task based on that, what I'll do is I'll copy, cause I use Gmail, the subject line of that

00:32:41.200 | thread into the task card. And I'll say, just search for this. So I'll know what to search for in the

00:32:46.080 | archive, the pull up exactly that thread. So I won't, you know, it's not being copied in, but I

00:32:49.600 | say, here's what to do. Um, I do that with book. Sometimes when I have a, like a book blurb, I need

00:32:54.320 | to get started on and they've sent me an electronic copy. Sometimes I'll attach the file to the card,

00:33:00.560 | or sometimes I'll just put the title of the email in which that file is attached. And I'll just search

00:33:05.340 | and find it and print it or load it onto a reader. You know, when the time comes for information,

00:33:10.320 | it's more complicated, like you're working on an article or something, and there's going to be a lot

00:33:14.180 | of different sources, right? Where you don't want three dozen sources attached to a Trello card.

00:33:19.780 | My general rule is when possible, store research information in the tool you're going to use when

00:33:25.680 | you need it. Right? So I write books and articles using Scrivener. I move relevant notes and links and

00:33:33.700 | resources and files that are relevant to projects, like to a Scrivener project for that article or for that

00:33:39.800 | book chapter. So I just move the information straight in the Scrivener. When I'm writing

00:33:43.160 | an academic article, I use a LaTeX markup language that mathematicians use, right? I'll just move

00:33:49.000 | stuff I need right into the LaTeX file in which I'm working on. I have an idea. I'll just start

00:33:53.780 | building a LaTeX file, basically like I'm writing a technical article and I'll just add citations and

00:33:57.800 | notes right into that document and it grows and grows. And then eventually what happens is the actual

00:34:02.500 | paper starts getting written at the front of this and the notes get pushed to the back. And then

00:34:06.040 | eventually they get moved out and you have the final paper. But I just put the information

00:34:09.280 | like proofs and links and citations and summaries, like right into the same tool I'm going to use

00:34:14.080 | to write the final paper. So for complicated information, if you can keep it close to the

00:34:18.820 | tool you're going to use, I tend to think that's, that's probably best as well.

00:34:22.100 | So with the email search, do you just keep the email on your inbox?

00:34:27.440 | Well, so in Gmail, it archives. So you archive it, it's out of your inbox.

00:34:32.260 | And then you, uh, it's searchable at that point, but Gmail search is not very good. So you really need

00:34:38.880 | the, uh, exact subject line. So I'll just copy, like I'll actually copy and paste the subject line

00:34:43.940 | and just put it right on the card.

00:34:44.920 | Do you ever go into your archive folder and take stuff out or is it just...

00:34:48.660 | It's not a folder. So the Gmail paradigm, it's based on search. So the Gmail paradigm

00:34:53.040 | is the archive. It's like the internet or something. You never navigate it. It's just something you search

00:34:58.120 | into. So you just have this, like that was their big idea. Like we're good at search. So when you're

00:35:02.480 | done with a thing, don't organize it to a file, a folder, don't have like a hierarchy of different

00:35:07.640 | folders, just archive it. And you can just use our smart search, define what you need later.

00:35:12.960 | The unstructured storage. The problem is their search is not that smart. Like the whole reason why

00:35:17.060 | Google searching works well on the web is that it's leveraging all this extra information in

00:35:21.620 | particular link structure. So it knows like this page is probably what you're looking for because it's

00:35:26.440 | being linked to by a lot of other important pages. You don't have any of that information

00:35:30.220 | when it's just your own emails. So it's really just doing like a much more standard search. It's

00:35:34.020 | not very good to be honest. So you really need like a subject line. Like I know exactly what I'm

00:35:40.960 | looking for. And if you want to pull up again, you always need to have the subject line.

00:35:44.320 | Yeah. Yeah. Yeah. That's the way that works. All right. Who do we got? Next question is from Diana.

00:35:49.380 | How can I improve my ability to communicate effectively in spontaneous workplace

00:35:54.180 | conversations, especially when I struggle with clarity, organization, and understanding

00:35:58.360 | others' needs? I speak well for planned presentations, but not as well in off the cuff scenarios.

00:36:04.320 | Well, Diana, I think a good rule of thumb, and Jesse will agree, is you want to break the ice.

00:36:11.700 | The thing is you want to find some sort of characteristic of the person you're talking to,

00:36:17.480 | preferably like an immutable characteristic. They have no control over it.

00:36:20.640 | make fun of them and then say, you're burned. That is like, it kind of breaks the ice. And then

00:36:26.960 | you're not going to be, you're not going to be, uh, you're not going to be so nervous. Right.

00:36:32.460 | Like you so bald. He's got to talk like that. You so bald. I'm blinded by light off your head.

00:36:40.280 | You're burned. Right. And then the boss is like, all right, well, I mean, you're clearly fired.

00:36:44.780 | Um, no, it's a good question. Speaking is a practiced art. I think people don't recognize this.

00:36:50.320 | Younger people supposedly are worse at this because there's actually less in-person spontaneous

00:36:55.680 | interaction than before, because more things are done on phones and more things are done digitally,

00:37:00.900 | like through text messaging. So there's this, this idea out there that I think Sherry Turkle wrote

00:37:06.640 | about this. Like there's this idea out here that younger people who are coming up with smartphones are

00:37:10.720 | particularly bad at in-person conversation. They don't want to sit down and talk to their bosses.

00:37:14.040 | They certainly don't want to call someone that's really fraught. When Jesse and I were growing up,

00:37:17.700 | that was a big part of our childhood, right? Was calling someone and you would say the person's

00:37:23.180 | name, Mr. and Mrs. You'd be like, you know, uh, hi, Mr. Miller. This is Cal Newport. It's like Jesse

00:37:27.960 | there. And you had to learn how to do that and like how to, how to interact. Interestingly, and maybe

00:37:33.000 | just because I'm in like elite institutions, I haven't seen this issue. Like my Georgetown students,

00:37:37.140 | they're super articulate, but that might be self-selecting. When I was up at Dartmouth teaching a couple of

00:37:41.780 | years ago, these kids are super articulate, but that might not be a representative sample.

00:37:46.320 | Um, that being said, speaking is hard. Speaking is hard. Like I make a living speaking. I've been

00:37:52.960 | working at it for a really long time. It's the degree to which right now, for example,

00:37:58.680 | speaking on a podcast that I'm having to rely on essentially cognitive muscle memory training,

00:38:04.860 | not to say, um, not to say like, to get the right pauses, to keep things moving. Like that's very

00:38:12.220 | practiced to have an idea that you're pursuing to pursue it clearly, not to get caught in a cul-de-sac,

00:38:17.480 | you know, that takes practice. So what I'm trying to tell you, Diana, is it's not a problem that you're

00:38:21.540 | struggling right now. It's, you haven't practiced it yet. It's no different than you say, look, I picked

00:38:25.180 | up a banjo the other day. It's not going well. It doesn't sound good. It's like, yeah, of course you

00:38:28.320 | got to practice. But if you practice, you'd get better. And that's the way you want to think about

00:38:32.260 | speaking off the cuff, writing and reading actually helps people who read a lot and write a lot.

00:38:37.660 | They're just being exposed so much to structured formal language that you build those structures,

00:38:42.900 | those predictive structures in your head. It's like you're a language model and you get better at

00:38:46.420 | reproducing it. So reading will help trying to write very clearly will also help. Practicing is

00:38:52.640 | ultimately a thing that's going to help best. You can practice in situations where it's lower stakes.

00:38:56.860 | So like with your friends or family, you're having conversations over dinner. Say, hey, I'm going to

00:39:01.060 | try a little harder in this interaction here. I'm going to be a little bit more formal. Just

00:39:05.100 | straight up practice is going to make you better. There are heuristics that can help as well.

00:39:10.340 | So for example, if you're talking to someone at a business meeting or at a conference, or you're

00:39:15.840 | standing up to make a point, sometimes it's useful to think about the following structure. Set up point

00:39:21.320 | contrast implication. Let me clearly set up what I'm going to say. So here's the context

00:39:26.740 | or the stakes. Hey, I know that we were, we've been talking, you know, recently about making our

00:39:32.800 | meetings more efficient. I have an idea there to share. That's set up. So make it very clear what

00:39:37.980 | you're going to talk about. Point. Deliver clearly. Here's my point. So here's the original information

00:39:44.160 | I'm offering. So in this sample, this is where you might say something like, we have too many meetings

00:39:50.760 | that could probably be consolidated into fewer. We should think about ways of combining meetings.

00:39:55.480 | You've set up what you're going to talk about. You've made your point. Then you contrast, right?

00:40:00.200 | This is the dialectical approach. As opposed to what we have been doing, which is if anyone has

00:40:07.380 | something they want to discuss in a group, they send out a separate invite for it. So you've made

00:40:11.100 | your point, what you want to deliver. You contrast it to existing points, and then you give details and

00:40:15.680 | implications. So the way we might do this, for example, would be instead to have a standing

00:40:20.620 | meeting block every afternoon that can be used for multiple purposes. The implications of that

00:40:26.500 | details would be some stuff might have to wait or that meeting might get rushed, but I think it's

00:40:31.200 | probably worth it. So if you hear someone follow that pattern, you don't recognize that they're

00:40:36.520 | following that pattern, but it seems very persuasive. Let me cue you in setup. Let me cue you into what I

00:40:41.520 | want to tell you about. Let me give you very clearly. Here's my point. Let me contrast it to

00:40:45.500 | other ways you might do this to clarify the contours, give you some details and the implications

00:40:50.840 | of it. Report and I'm out. You talk that way in a business setting and they're gonna say, Diana's

00:40:56.420 | sharp, really clear, interesting ideas. They're really on the ball. So there are heuristics like

00:41:01.100 | that you can do, but practice is the best. If you listen to, this is a canard, but it's true.

00:41:07.080 | This is like one of these cliches that true. Screenwriters will tell you, like Hollywood

00:41:10.920 | screenwriters. I know some Hollywood screenwriters. They'll tell you that you want your dialogue to

00:41:16.360 | sound quote unquote natural, right? You don't want people to sound like they're reading a script,

00:41:21.440 | but what'll happen is most screenwriters early on in their career will go through this point where

00:41:26.420 | they're like, let's just listen to real conversation. I can take notes on real conversation so I can make

00:41:30.300 | it very natural. And what they realize is real conversation is barely comprehensible. If you just sit and

00:41:36.980 | listen over, listen to people casually talking, and then you transcribe that and try to put it into a

00:41:41.940 | movie, people are going to think that the point was like that character had a traumatic brain injury

00:41:47.640 | recently. Like this, this is so incoherent. Like what's going on? Because the way most people talk

00:41:52.620 | when the stakes are low and it's casual and it's with a friend or whatever, it's weird and back and

00:41:57.100 | forth and interruptive, very digressive, all sorts of verbal tics and re returns and ums and likes.

00:42:03.640 | And yeah, I don't know, man, what's going on. And it's fine in that informal context, but you know,

00:42:08.320 | it's not at all the way people actually, it's not formal at all. And so in Hollywood, naturalistic

00:42:14.580 | dialogue writing is its own thing. It sounds naturalistic, but it's actually way more structured

00:42:20.060 | and clear and directed than the normal people actually really talk. And it's really in the

00:42:24.300 | delivery of the actor that makes it sound naturalistic, which is all to say to practice

00:42:27.640 | this. You have to think I'm having a conversation with someone where I could be super casual and

00:42:32.040 | doesn't matter. Let me try to be clear. The other thing people always suggest is Toastmasters. I

00:42:37.000 | don't know if these still exist, but yeah, you can also go to a speaking club like Toastmasters where

00:42:41.740 | you practice getting up and giving speeches, but practice is the key here and have some heuristics

00:42:45.940 | like setup, point, contrast, details, implication. It's a cool skill to master because the clearer you

00:42:52.680 | communicate, the smarter people think you are, which is going to be a big boost. All right. What

00:42:57.700 | do we got next? Next question is from Flying Professor. My life and work are divided between

00:43:02.700 | two different countries with a seven hour time difference. Google calendar just couldn't handle

00:43:07.240 | that. So I've gone back to paper and pencil to calendars. Is this okay? Or have I missed something?

00:43:12.380 | Yeah, that's a good question. Quick follow-up on Diana. I was just thinking about this, Jesse,

00:43:15.680 | though. Quick follow-up on Diana. I was a very good speaker as a kid and I got a lot of like unfair

00:43:20.540 | advantages out of that. I think because I read a lot. But I remember doing an oration at,

00:43:28.420 | my God, I guess I was at like a VA and I had to give a speech in front of everyone about flag burning

00:43:35.640 | and whether there should be a constitutional amendment. And I remember just ad-libbing it. I didn't realize

00:43:39.720 | it was supposed to prepare something. Anyways, they selected me. I got to go to Boy State. Have you heard

00:43:43.800 | about Boy State?

00:43:44.460 | I think you might've mentioned it before.

00:43:46.640 | Yeah. So like that was all oration. So I was considering, I was good at Model U N as well.

00:43:50.720 | And I would, I would do very little work, but I could speak well. And I think it was largely

00:43:55.860 | because I read a lot.

00:43:56.920 | Mm-hmm.

00:43:57.600 | There you go. That's my secret power. All right. Flying Professor, two time zones. Yeah. You need

00:44:01.280 | two calendars. You need two calendars. You can use electronic calendars. You just need two of them.

00:44:07.400 | So you could have an electronic calendar for one time zone and another electronic calendar for the

00:44:11.200 | other time zone. The problem is you can't use the same Google account. As far as I know, I think the

00:44:16.500 | time zone is going to be consistent across your various calendars. So you have to create, it's fine.

00:44:20.740 | Just create a different Google account just for this calendar. And then you can share, you know,

00:44:26.160 | whatever you actually don't want to share them. You don't want the one time zone calendar to share

00:44:30.700 | onto your current time zone calendar because it's going to be the wrong time zone. Let me think about

00:44:37.400 | this for a second. So I have one calendar for your time zone here. Have another calendar with another

00:44:41.980 | account for the time zone for the place you go. So those are both now local calendars. So when you're

00:44:47.180 | in the one place, you look at the local calendar for that place, correct times. When you're in this

00:44:51.560 | place, look at the calendar for that place. You're seeing the correct times for that place.

00:44:55.700 | You can then share between these different time zone calendars if you want to, and you'll see where

00:45:01.380 | these things fall on each other's time zone. So seven hours, I'm assuming you're talking like Europe

00:45:06.300 | and East Coast, let's just say. When you look at your East Coast calendar, you can share your Europe

00:45:11.560 | calendar onto your East Coast calendar, and it will show it like the East Coast times, but you probably

00:45:15.540 | don't care about that. But it'll show it like, yeah, this thing's happening at two in the morning or

00:45:18.520 | whatever. But I guess you wouldn't want to share them. Just have two separate calendars. So you can use

00:45:21.900 | paper calendars, fine. I like electronic calendars because I want to do email reminders. I want to

00:45:28.120 | add information, like here's the Zoom link for this particular Zoom meeting that's going on my

00:45:33.040 | calendar. I want to move things around easier. I want to do repeats. So I would create two Google

00:45:36.900 | calendars if I was you. Completely different accounts, completely different time zones, and you can just

00:45:41.180 | switch between them depending on where you are. And you're right, don't share between the two. That'll

00:45:44.580 | confuse you with the time zones. The time zone issue that's the worst, Jesse, speaking of Europe,

00:45:50.500 | is there's a month. Everyone who's worked with the UK knows this. And I work close with the UK. We do a

00:45:56.720 | lot of book sales there. There's a month in March where we get out of, it's like we, daylight savings

00:46:04.720 | times changes in the US and not there. It's different. And there's a month where, until recently,

00:46:12.800 | no one from the US meeting with someone in the UK would ever be on time. It was just a crapshoot.

00:46:16.940 | Everything was off by this hour and it was different than it was the month before and it changes to

00:46:20.520 | the month after. Now it's better because of calendar invites. So if like my UK publisher sends me a

00:46:29.520 | meeting invite for a phone call, Google fixes the time zone. And so now it's okay. But there's a long

00:46:37.500 | time where I just assumed I was going to get everything wrong. I think when I was doing UK publicity,

00:46:41.180 | even as late as like digital minimalism, it was like, this is, we're going to dismiss things.

00:46:45.100 | And so finally Google calendars has made that possible. Everyone gets that wrong. It becomes

00:46:50.000 | like a six hour difference or an eight hour difference temporarily for just one month.

00:46:53.440 | The worst thing.

00:46:55.140 | I have a couple of follow up questions. What do you do with email reminders?

00:47:00.820 | I use them for, uh, important appointments and stuff like that.

00:47:04.840 | So you, how does it work? You Google calendar, you can click on event and then you can go to

00:47:09.480 | notifications and you can choose email me about this. And here's how early before the event to email me

00:47:15.780 | about it. So I often do that for important things. So I'll see it first thing. So I don't always trust

00:47:20.480 | myself. I'm not like an executive where, you know, my day has been scheduled inside and out. And I,

00:47:27.720 | my whole day is run off my calendar. First thing I'm going to do is look at my calendar. I'm not like

00:47:30.900 | that. Like there might be days where I'm, I think, yeah, I know I have to go in at 11 to teach. I'm

00:47:35.880 | going to try to write this morning. Yeah. And so like, I might not look at my calendar until then,

00:47:40.000 | but maybe there was something important on there that I forgot about. Like, Oh, I'm supposed to go to

00:47:43.580 | some appointment. So I'll have an email me because I usually will, you know, see my usual email the

00:47:48.880 | day before. So when I check my, you know, shut down my email the day before I see this reminder of

00:47:53.160 | like, Oh yeah, yeah. Tomorrow morning I got an event. Um, mainly it's psychological. It just makes

00:47:58.140 | me worry less about missing something. If I know I'm also going to get an email about it.

00:48:03.040 | And then I have another follow-up question about the archives. So do you only archive items that you

00:48:09.000 | have those detailed notes of, or sometimes do you archive things and just knowing that you'll never be

00:48:13.200 | else find them because the search won't be good? That's the Gmail. The Gmail methodology is you never

00:48:17.620 | delete. You archive. So in theory, their whole thing is storage is cheap and just keep archiving all your

00:48:25.300 | emails. So you never delete in. So you archive all yours? Yeah. But you only search for like a small

00:48:30.560 | percentage. Yeah. Yeah. That's the Gmail data model is never delete, just archive. Interesting. Yeah. And

00:48:38.680 | you can't see the archive really. It's just, you can search into it. Um, and then they want you to do

00:48:43.860 | like to use the data. So it's sort of, they, they use the data from your email, but like train things

00:48:48.360 | and then target ads to you. It's, it's a whole, Oh, it's a whole world. Surveillance capitalism in its

00:48:54.820 | best. All right. We're going to do two calls today. Do I have that right? Yeah. I love that. Let's get

00:48:58.600 | some calls in. All right. Let's get our first call. Hi Cal. Uh, it's Jessica. Um, I have a question about

00:49:04.580 | coaching. So I was listening to one of your podcasts and you mentioned that you discussed

00:49:07.840 | something with a coach. Um, I'm a female academic in Europe and there's a lot of coaching programs

00:49:13.920 | that are free and offered by the sort of university development service, um, often to women, but also

00:49:19.620 | to sort of junior academics, um, at the university. Um, and I struggle a bit to see which ones are sort

00:49:25.220 | of worthwhile and useful to go to and which ones, um, are not so worthwhile. And I was wondering if you

00:49:31.160 | could expand a little bit on how you integrated, um, uh, coaching, I guess, uh, receiving coaching

00:49:36.660 | into sort of, um, building a deep life and, and setting up a schedule, uh, that, that has sort of

00:49:42.840 | appropriate, uh, blocks of, of deep work. Thank you. Well, that's a good question. I think coaching's

00:49:48.900 | underrated, you know, in general, I think there should be more coaching. It's part of it is informational.

00:49:57.760 | So you get information you wouldn't otherwise have. A big part of coaching is often backstopping,

00:50:01.900 | right? Having someone backstop a decision that you're making so that you have some confidence in

00:50:07.520 | it. So it's a whole bunch of confidence boosting, right? So, so coaching, I think it's really useful.

00:50:11.180 | It can mean a lot of different things. If you're an academic, there's a couple of types of coaching

00:50:14.920 | that's relevant. The first type is what I might more accurately call mentorship. And this is where

00:50:19.860 | you're receiving advice from someone who is in your field and academic, but it's just more senior.

00:50:24.160 | And they're really important. Actually, this might be one of the more important things you can do,

00:50:28.040 | especially as a junior academic is go to lunch once a month with a more senior academic in your field and

00:50:34.460 | just pick their brain about what about this? I'm, I'm worried about this. What do you recommend here?

00:50:40.540 | What do you think is going well or not? There's a lot of nuances in this world as in lots of different

00:50:44.560 | knowledge work worlds, lots of nuances that aren't given to you in a manual.

00:50:48.800 | And it's left to you to figure it out. And you could be going down a wrong end Georgetown,

00:50:53.340 | you know, where I've been my whole academic, my professor career, they formalize this. You would

00:50:58.200 | be assigned as a, as an assistant professor, junior faculty, you'd be assigned or you could choose,

00:51:03.560 | but you needed a mentor. And then the university or department would pay for you to go to lunch.

00:51:07.760 | And so I'd go to the tombs once a month for a while with, with my mentor. Good place, right?

00:51:13.020 | I was just there. I love the tombs.

00:51:14.320 | The tombs is awesome. And phones don't work well down there.

00:51:17.120 | Yeah.

00:51:17.680 | Yeah. And they have rowing stuff on the wall. You know, tombs is great. But anyways,

00:51:21.420 | we go to the tombs once a month and it's just like whatever was on your mind. And I'd learned a lot

00:51:26.240 | about navigating academia, but navigating the specific institution and like who's who and who's what and

00:51:31.280 | what matters and what doesn't and what you should be worried about. That's very useful. So I think

00:51:35.100 | mentorship is very useful and that's different than coaching because it's not someone coming in from the

00:51:40.000 | outside of your world. It's someone in your world who's ahead of you. So definitely try to find

00:51:44.720 | mentorship, make that happen. People, senior professors like to talk to their junior professors.

00:51:48.800 | That shouldn't be hard. Then coaching is more, I would say, tactical. It's okay. Help me. I have

00:51:54.500 | this, this problem I want to solve. I want to, here's my goal. I want to be doing more of X.

00:51:59.060 | Help me figure out how to do that. In academia, I think the most effective type of coaching of that

00:52:03.800 | type that I've seen really does focus on the issue that you mentioned there, which is making sure that

00:52:08.460 | you're finding enough time to get the work done, the work that matters, which is going to be your

00:52:12.360 | research production. So having someone in your life who's looking at your schedule and is helping you

00:52:16.000 | figure out smart ways of you need more time. You're not getting enough time writing. How can we help you

00:52:20.380 | get that time? They're a sounding board. They're a backstop. They're a confidence booster. They're a

00:52:24.520 | source of ideas. They'll say, you need to drop that committee. That's okay. You need to protect your

00:52:28.500 | mornings. Uh, they might see insights you might not have. I mean, I don't know if you remember this,

00:52:33.800 | Jesse, but when we had my friend, uh, Laura Vanderkam on the podcast, we were doing a case

00:52:39.000 | study. I think she helped me answer questions. If I remember properly. Um, and we had a question

00:52:44.580 | from a professor and she was having a hard time finding time for deep work. And Laura had like

00:52:48.900 | this really interesting scheduling idea that I hadn't thought of. And the caller hadn't thought

00:52:54.240 | of, but it made a really big difference. She's like, here's what you really need to do. And I don't

00:52:57.660 | remember the details other than that. This young professor was trying to fit in her deep work into

00:53:03.780 | all these little slivers. And Laura's like, no, no, here's a change you can make. If you have a

00:53:09.600 | babysitter here and you swap. So your husband takes these shifts and you take those shifts. Now you have

00:53:15.460 | two long count sessions. You can count on each week, four plus hours. And now you can actually make

00:53:21.940 | progress on papers. And they were relatively minor. They're not hard things to do, but she just wasn't

00:53:26.900 | thinking about it. But having someone who's done a ton of time management, like Laura was like, Hey,

00:53:30.600 | I think you're missing this opportunity, this option here. And it made a big difference.

00:53:33.660 | So I think that type of tactical coaching, and there is a lot of coaches that focus on this in

00:53:37.680 | academia. I help. They're often people who left the academic track. Like I help professors make

00:53:43.540 | sure they get the important things done. I think that is absolutely worth the time and worth the money.

00:53:48.160 | The coach I have, shout out to Cheryl. She focuses on something very specific, which is people who are

00:53:54.580 | doing creative work and have to deal with like the business and logistical challenges of making that

00:54:00.040 | happen. So like she also, we've mentioned this works with a lot of professional writers, screenwriters,

00:54:04.960 | movie directors, people where the core thing you do is creative. And as you do it better,

00:54:12.200 | it becomes harder and harder to actually do that thing that you do well, right? Like all these new

00:54:18.400 | responsibilities and opportunities and options come up. So it's really focused. She focuses on

00:54:23.120 | people. So it's really focused on my writing life, not my professor life. And how do I keep producing

00:54:27.540 | and enjoying what I'm doing? She's really focused on like enjoying what you're doing and producing

00:54:31.460 | creative output that I'm really proud of while managing the business aspects of this in a way that

00:54:35.460 | that doesn't just take over. And so it's a very specific type of coaching. So anyways, I'm a big fan of

00:54:41.100 | coaching. It's mentorship and coaching itself. Do it all. Do it all. It's only going to help and it

00:54:48.100 | makes things seem less scary and less lonely. Coaching is a good business. Our friend Brad

00:54:53.440 | coaches Stolberg. Yeah. That's like one of his, that's like his job outside of writing. He's in

00:54:58.620 | super demand too. I think you can't get on his, his, it's like a long wait list at this point.

00:55:03.420 | Hmm. All right. Who do we got next? Another call? Hi Cal. On a recent episode, you talked about

00:55:10.520 | scheduling a time in your day to embrace boredom so that the rest of your day, you don't have to worry

00:55:17.820 | about feeling bad about listening to a podcast, et cetera, et cetera. If I were to schedule boredom,

00:55:27.740 | how would that look? Could you elaborate? Do you just sit in a room for 20 minutes or the next time

00:55:34.580 | you're in traffic, you just deal with it? Am I allowed to just let my mind wander? Please explain

00:55:43.040 | how this works for you. Thank you. Okay. So embrace boredom. That phrase comes from deep work.

00:55:49.560 | So it's a chapter in deep work. And I think what's important about it is boredom is not something that I

00:55:55.460 | think has a moral valence. So unlike some commentators, I don't think boredom is somehow

00:56:01.960 | good and non-boredom is bad. And therefore by having more boredom, that somehow like that act

00:56:07.580 | itself of being bored is virtuous. I see this way more technically. So when I say embrace boredom,

00:56:14.480 | what I'm hoping that you accomplish is that your brain gets used to craving novel stimuli and not

00:56:21.880 | getting novel stimuli in response. So it's really just working on the dopamine mediated short-term reward

00:56:27.320 | circuits. So if your brain learns every time it feels boredom, it gets a treat from a phone, right?

00:56:37.140 | Something comes out and it gets a novel stimuli. You build up these reward circuits that says,

00:56:42.340 | ooh, as soon as I feel that sort of discomfort of lack of novel stimuli, which is natural to people.

00:56:47.080 | And we've felt it throughout all of our species history. As soon as I feel that there's this shiny

00:56:51.780 | treat reward I'm going to get, it really floods the zone with dopamine and makes it really hard not to

00:56:56.680 | look at your phone. Why do I worry about that? It's because sometimes you don't want to look at your

00:57:00.740 | phone when it becomes important that you actually do focus on something like you're working on a really

00:57:05.120 | important paper or a breakthrough or a meaningful conversation or trying to figure out something

00:57:08.680 | hard. There's no novel stimuli when you're doing that. But if your brain has learned,

00:57:12.260 | I always get a shiny treat when I feel that lack of stimuli that we commonly describe as boredom,

00:57:16.280 | you're not gonna be able to focus and you're going to have to look at your phone.

00:57:19.220 | So the way you break that is you make sure that you get consistent practice, feeling boredom and not

00:57:26.560 | getting the reward and things being okay. And then your brain's, the strength of that Pavlovian response

00:57:31.900 | weakens. And now when it comes time to do something like deep work on something important, you'll do

00:57:35.220 | better at it. So I see embracing boredom. That's why it came in my book. Deep work is just training your

00:57:39.360 | brain to be better at deep work down the line. All right, how do you actually do this? There's two

00:57:44.180 | scales at which you need to embrace boredom. Every day, do something short where you're free from input

00:57:50.200 | from other minds. So you're welcome to think as much as you want inside your own head. You're welcome to

00:57:55.080 | observe the world around you as much as you want and make observations and be interested by it.

00:57:58.920 | But you're not inputting input that was generated by another mind. You're not reading something or

00:58:03.380 | listening to something. Every day, do some sort of short exposure to that like once or twice. And by short,

00:58:08.480 | I mean, you know, you do an errand, you don't take out your phone. You take the AirPods out. Like I'm going in the

00:58:13.060 | pharmacy to grab something. I'll just go and do that and come back. It takes 10 minutes, but I'm going to do it

00:58:16.720 | without my phone. Or I'm doing a short drive, you know, to a friend's house. It's 10 minutes away. I'll just do that

00:58:22.280 | drive. Nothing on the radio, right? So that's like a short one or two exposures, you know, once a day.

00:58:28.340 | Then once a week, if possible, try to have a longer exposure, which really I typically think of as being a long

00:58:34.420 | walk is probably the best and go for a long walk or a hike or something like this without anything in

00:58:38.860 | your ear or anything in your hand and just be alone with your thoughts. And this gives you time to do

00:58:42.080 | some more structured thinking and introspection. And this just gets your mind used to this idea of

00:58:46.180 | our brain can function just fine without shiny treat stimuli. I've been adding, however, if I was to

00:58:51.340 | rewrite that chapter, I would add a third type of boredom training that wasn't as relevant back when I

00:58:57.740 | wrote deep work, but I'm caring more about now. I would say, be very wary about dopamine stacking.

00:59:03.780 | So if you're watching something, don't look at your phone, have your phone plugged in somewhere else,

00:59:08.060 | right? Don't get in this idea of, because that really gets those reward circuits going that even

00:59:13.940 | if I'm watching something that's pretty interesting, I might want to stack on something that's even more

00:59:18.720 | interesting. That really is going to short circuit those, those reward circuits and make it really hard.

00:59:23.280 | So be very wary about dopamine stacking. If you're going to do something, look at a screen,

00:59:29.300 | only look at one screen at a time. So I would add that in as well. You do those things. You're not

00:59:34.520 | bored all the time. We're not lauding boredom, but you'll get more attraction to thinking and being

00:59:39.720 | with your own thoughts. That'll become more appealing as you get practice. But it makes sure your brain does

00:59:43.640 | not develop this knee-jerk reaction. When I feel boredom, I have to get stimuli. So that's what I would

00:59:48.820 | recommend. What if you have like a laptop that you keep on your coffee table and you want to,

00:59:53.260 | like look something up when you're watching something? I would keep it somewhere where you

00:59:58.100 | have to move. That's what I would say. Like I, my, my couch face on my TV, we have like have an open

01:00:05.020 | plan thing. So the kitchen's in that same bigger space. And I like to plug my phone in, in the

01:00:11.020 | kitchen, we have like a charging thing. Uh, so I can see over there from the TV, but if I want to look

01:00:15.740 | something up, I walk over there and look it up on my phone and then I can come back to the couch.

01:00:20.040 | So that's fine, but you can't dopamine stack, right? It's a little bit different. If it's right

01:00:23.940 | there, you're going to start rabbit holing and I'll find myself, if I start rabbit holing, I'll

01:00:28.740 | like to turn off the thing I'm doing and then do that. And then when I'm done with that, go back and

01:00:32.640 | do the thing, the thing that I'm doing. Oh, I like that tip. Yeah. Or press pause or something.

01:00:37.520 | Yeah. Yeah. Yeah. Yeah. Or muted or something like that. All right. Uh, we got a, speaking of,

01:00:42.720 | it's not really speaking of, but we got a geeky final segment coming up, but really what we're

01:00:47.100 | saying there had nothing to do with reinforcement learning is what I'm going to talk about. But

01:00:50.940 | before we get to our final tech corner, let's talk about another sponsor. Talk about our friends

01:00:56.700 | at express VPN. Going online without express VPN is like leaving your blinds open at night. Anyone

01:01:03.940 | can watch what you're doing and you have no idea who might be out there lurking in the dark.

01:01:09.960 | If you let me put my CS hat on, this is why when you communicate over the internet, you're sending

01:01:16.040 | what are known as packets. So the packets have an address of who you want to talk to. And then the

01:01:20.300 | contents like what you want to say. Now those contents are often encrypted, right? If you're using a

01:01:25.340 | protocol like HTTPS, those are encrypted, but not the address. So what you're saying is encrypted,

01:01:31.800 | but not the address of who you're talking to, which means if you're on a wireless access point,

01:01:35.940 | anyone with a radio antenna in the right software can read those addresses from your packets and say,

01:01:41.840 | oh, I know what websites or services you're talking to. Or if you're at home and you're like, look,

01:01:46.960 | I live on a farm and I'm isolated. No one's near me. Your internet service provider

01:01:50.780 | who is communicating to the internet, your portal to the internet, they see who you're talking to.

01:01:55.600 | They see the addresses and they could sell that information. And you know what? A lot of them

01:01:59.920 | actually do. This is where a VPN like ExpressVPN enters the scene because what it does is instead

01:02:06.060 | of just talking directly to the website you want to talk to, you take that message and you encrypt

01:02:10.020 | the whole thing and you send it to the VPN server. And the VPN server opens up that envelope of

01:02:15.620 | encryption, takes out the original message and talks to the site and server on your behalf,

01:02:19.440 | encrypts the response and sends it back to you. So the people who are just watching you and reading

01:02:24.500 | the addresses of your packets, all they learn is that you're talking to a VPN server. They have no

01:02:28.500 | idea what site and service you're actually talking to. And therefore you get some privacy back. Your

01:02:32.960 | internet service provider doesn't know what sites and services you're using. That creepy guy sitting

01:02:37.700 | next to you in the coffee shop with like the satellite dish hat who keeps like staring at you oddly. He

01:02:42.340 | doesn't know who you're talking to. All you learn is that you're talking to a VPN server. So you should

01:02:47.740 | use a VPN, especially, you know, when you're traveling or around other people. If you're going to use a VPN,

01:02:53.340 | I suggest that you use express VPN. Express VPN has servers all around the world. They have a lot of

01:03:00.280 | bandwidth. So it's fast connections. There's going to be a server nearby. It's easy to use. You turn it

01:03:06.080 | on and now all of your internet traffic automatically goes through the server and you use your sites and

01:03:10.780 | services on your devices, just like normal. Um, it works well and it gives you that privacy back,

01:03:16.940 | uh, not just on your computer. You can use this on your phones. You can use this on your tablets, like whatever

01:03:23.400 | devices you use. They even now have an optional dedicated IP service that's engineered with innovative zero

01:03:29.600 | knowledge design. So now not even express VPN can trace an IP address back to the user. Zero knowledge. I could

01:03:37.320 | go on about that for a while. It's a complicated cryptography idea, but let's just put it this way. It gives you an

01:03:42.940 | even extra bigger layer of privacy. All right. There's a reason why express VPN has been rated

01:03:47.620 | number one by the top tech reviewers like CNET and the verge because it works well. So protect your

01:03:51.860 | online privacy today by visiting expressvpn.com slash deep. That's E X P R E S S vpn.com slash deep to find

01:04:00.920 | out how you can get up to four extra months free. That's expressvpn.com slash deep. Also want to talk

01:04:08.720 | about our friends at Oracle cloud infrastructure. There is a growing expense eating into your company's

01:04:13.700 | profits. It's your cloud computing bill. You may have gotten a deal to start, but now the spend is sky

01:04:19.520 | high and increasing every year. What if you could cut your cloud bill in half and improve performance at

01:04:25.060 | the same time? Well, if you act by May 31st, Oracle cloud infrastructure can help you do just that.

01:04:32.660 | OCI is the next generation cloud designed for every workload. You can run any application,

01:04:37.640 | including any AI projects faster and more securely for less. In fact, Oracle has a special promotion

01:04:43.420 | where you can cut your cloud bill in half when you switch to OCI. The savings are real. On average,

01:04:49.420 | OCI costs 50% less for compute, 70% less for storage, and 80% less for networking. Join modal Skydance

01:04:57.440 | animation and today's innovative AI tech companies who upgraded to OCI and saved. This offer is only for

01:05:03.780 | new U.S. customers with a minimal financial commitment. See if you qualify for half off at

01:05:07.560 | oracle.com slash deep questions. That's oracle.com slash deep questions. All right, Jesse, with that

01:05:15.920 | technical ad read in mind, let's move on to our final segment. All right, well, because I can't help myself

01:05:22.400 | for our final segment, I want to do another tech corner where we can dive into some needlessly

01:05:28.680 | technical details about the world of AI. All right, so what I want to talk about today, Jesse, is two

01:05:36.180 | different types of AI technologies that are often getting mixed up. And I want to talk about what they

01:05:41.320 | have in common and what makes them different and what we have to worry or not worry about each of them

01:05:46.360 | so we can kind of have a more nuanced understanding of what's going on out there. So let me start with

01:05:51.020 | two examples. Here's two things AI has done in recent years that has been deemed impressive. One, AI beat a

01:05:59.520 | grand champion for the first time at a board game called Go, which is a, it's a simple to learn game

01:06:06.680 | where you take turns placing black or white rocks on a grid pattern. And if you surround, one color

01:06:14.440 | surrounds all of the rocks of another color, you flip them over to your color and you're trying to win on

01:06:20.160 | the board. And this was considered a hard game for a computer to win because there's this sort of

01:06:24.320 | astronomical number of possible boards. And so you can't just brute force search. Like if I do this and

01:06:29.360 | they do this, and then I do this, you can't brute force search as easily as you can for chess. So it was

01:06:34.580 | considered a hard game. And DeepMind, which is now owned by Google, created a system called AlphaGo that

01:06:40.240 | beat for the first time a grand champion, Lisa Dell, and it was a big deal.

01:06:43.160 | Another recent accomplishment with AI was the latest models from OpenAI, their chat GPT reasoning models

01:06:50.320 | are doing really well on math tests that are international caliber and something that AI had

01:06:56.620 | stumbled on before. Like, look, this is indication that the reasonability of these models is really

01:07:01.280 | good. I think a lot of people just see this all as like generally AI is getting smarter, but those two

01:07:07.080 | accomplishments have two very different types of AI systems underlying them. And it's useful to know

01:07:12.100 | the distinctions. The system that's winning at the game AlphaGo, like most other game-playing AIs that

01:07:18.160 | we see, is using an AI technology called reinforcement learning, whereas the system that is doing really

01:07:24.420 | well on that math test, so these like chat GPT systems, for example, these are large language models that

01:07:30.560 | are more data-trained generative models. These are different technologies. So let me start quickly

01:07:38.660 | with the common history of these two technologies, and then I'll talk about where they change.

01:07:42.640 | So there is a common revolution that occurred in the early 2010s that paved the way for lots of these

01:07:50.160 | different AI innovations that are happening today, even if the technologies in these innovations aren't the

01:07:54.720 | same. That revolution was the introduction of what are called deep networks, which you can just

01:07:59.920 | think of as a neural network that has a lot of layers, really big neural networks. Up to that point,

01:08:05.540 | AI researchers had not really tried to train big neural networks to do things. Part of the problem

01:08:12.660 | was figuring out, okay, how do we train neural networks that have lots of layers, right? You have

01:08:17.360 | to kind of propagate through signals back through all the layers to adjust the weights when you're

01:08:21.320 | training it. We saw that in the 80s. This is when Jeff Hinton and others figured out back

01:08:25.360 | propagation, a particular algorithm that was going to, could in theory train these networks.

01:08:29.780 | Okay, fine. But there's another problem there. They're so big that you would need a lot of

01:08:34.220 | compute power to train them and you would need a lot of data. It would require a lot of iterations

01:08:37.300 | to train them. Those obstacles fell as we got more compute and we got more data because of the

01:08:44.500 | internet. We got more compute in part because of the rise of graphic processing units, which the

01:08:49.380 | video game industry pioneered to make 3D graphics faster on PlayStations. Turns out those,

01:08:54.520 | those same cards, GPUs are great for training neural networks. So suddenly we got a lot more compute

01:08:59.920 | and the internet gave us a lot more data. And so now these ops, we knew how to train these things. We had

01:09:06.400 | enough compute power to train these things and we had enough data to train these things. The final obstacle

01:09:12.880 | was just philosophical. Machine learning experts up to that point said, no, no, no, you can't. If you train a really

01:09:18.480 | large network, they have so much potential memory, right? Because they're so big. If you train them on

01:09:24.360 | some data set, like I'm going to show you a bunch of pictures, some have cats, some don't. And I want

01:09:28.960 | you to learn what a cat is. I want to train you to recognize pictures with cats. They're like, the problem

01:09:34.500 | is that these networks are so big, they could just memorize a feature of every single picture with a

01:09:40.380 | cat from your training set. And they haven't really learned anything about cats. They've just learned

01:09:44.480 | about these pictures you've shown them. And then when you give it a novel picture that's never seen

01:09:48.280 | before, it'll do terribly because it has quote unquote overfit. All right. We finally got over

01:09:53.100 | that obstacle and said, well, let's just try it. This was basically the revolution of modern AI was like,

01:09:57.140 | let's just try it. And the original things that were tried were on image recognition and they trained

01:10:02.160 | these big networks and they didn't overfit. Instead, they got really good at recognizing novel

01:10:08.700 | images and started far outperforming other AI systems. Because it turns out what happened is

01:10:13.160 | if you have enough compute and time to train these things, instead of just memorizing your

01:10:17.460 | training set, they learned and generalized a lot of interesting information about the context in which

01:10:23.140 | they were operating. So suddenly we realized really big neural networks could do stuff that we never

01:10:27.800 | thought was possible. That was the origin of sort of all of the AI revolutions. But if we look at this

01:10:33.940 | technical sort of phylogenic tree, things began to split. So if we look to the language model split of this

01:10:42.760 | tree that led to things like chat GPT, the way you're training these very large neural networks

01:10:48.880 | for something like a language model is with a lot of data, a lot of data from the real world. So the

01:10:55.380 | original models is a lot of text and you're giving it real text from the real world. And you're knocking

01:11:02.080 | out a word from that text and saying, do your best to replace that word with what makes sense.

01:11:07.920 | And if it does a pretty good job, like produces a word that's pretty close to the actual word that

01:11:12.300 | was there, you sort of reinforce those weights in the network. And if it does a bad job, like now

01:11:16.760 | that's not working too well. And if you do this over enough data, what it does is it learns, it basically

01:11:21.360 | estimates the underlying processes that produce those texts. So the underlying grammatical subject

01:11:27.720 | matter processes we use when we produce text, it began to estimate those in these really large neural

01:11:32.500 | networks. And the larger we made the networks, the more complicated processes it could estimate.

01:11:37.040 | And now when you ask it to produce text from scratch, it can use, it has an estimate of the

01:11:41.060 | process we use in our head and it can produce really good text. It's a data-driven data trained.

01:11:46.720 | So there's a lot of understanding baked into those rules, but really the understanding is,

01:11:52.200 | it's trying to estimate an existing process for which it was seen a lot of products.

01:11:55.080 | The machines that play games, like the machine that played Go or the breakthrough that happened

01:12:02.480 | around 2014 when DeepMind said, look, we can win at Atari games. This was sort of like the big

01:12:07.760 | breakthrough. Those are using reinforcement learning. And the way they work is different. So again, they

01:12:13.660 | have a big neural network underneath it. They're using big neural networks. But now what they're doing

01:12:18.580 | is they're having that neural network interact with a world. It's not just being given a lot of data.

01:12:24.880 | Typically, it's going to be an online training. So like if you're training a game to play Atari games,

01:12:29.660 | they were actually giving as input to the neural network, every pixel on the screen of the Atari

01:12:35.500 | game. And then the network was outputting an action. Move the controller left, move the controller

01:12:42.160 | right, press A, press B, whatever. It was outputting an action. And then the simulation system would

01:12:47.880 | update it. Oh, great. Let's make that move. And then it would give feedback. Like, did this make our

01:12:51.720 | situation better or worse? And if it was worse, it was sort of say, don't let's change those weights

01:12:56.240 | away from this. And if it was good, it was like, let's change our weights towards this. And through

01:12:59.980 | these like interactions with a real world, with these reinforcement signals, what the model ends up

01:13:07.120 | learning is a policy. It learns a policy that is good at doing well by whatever definition of well you

01:13:17.100 | had, whatever reward function you're using in the training. It comes up with a policy for how do I react

01:13:23.780 | to different situations in a way that's going to help me maximize this, whatever reward I care about,

01:13:28.520 | like a point or whatever it is, the reward I care about. So these are two very different ways of

01:13:33.680 | training things. And this is how they trade AlphaGo. AlphaGo, they trained it at first with like

01:13:40.080 | actual Go games. And then they created two models to play Go against each other. And it was, you know,

01:13:47.100 | if it would do well against the other computer model, it would sort of reinforce those. And if it did worse,

01:13:51.300 | it would unreinforced and they played millions of games with each other and it created policies for

01:13:55.160 | how to play the game. These are two different things. So something you're going to get with a data

01:14:00.040 | trained language model, it's estimating real processes in the real world, like the way generally

01:14:05.980 | that we produce text. And the bigger the model, the more sophisticated that, that estimation is going

01:14:10.140 | to be, that's not going to, first of all, it's just, it's, it's, it's producing text. And it's trying to

01:14:18.820 | estimate what it's seen. So it's, in some sense, it's not, it can do kind of novelty things, but really

01:14:24.520 | it's trying to, its goal is to be as close as possible to like the things that produced the text

01:14:29.380 | it trained on. In reinforcement learning, all it cares about is having a policy that does well with the

01:14:35.960 | reward. It's not trying to estimate an existing way that people in a game play a game. It's not

01:14:41.760 | trying to look at a bunch of examples of a game and figure out how do people play games. It's just

01:14:46.300 | coming up with something that does really well in the reward function, which allows for a lot more

01:14:50.960 | originality. So like the classic case is when DeepMind was training one of these big neural networks to

01:14:57.000 | play the Atari game breakout, where you move the paddle and you knock the bricks. It devised a strategy

01:15:02.740 | that no human player had known about. Like it figured out that there was space on top of all

01:15:08.100 | the bricks and you could tunnel, if you're very precise, you could knock out the bricks out of

01:15:12.760 | diagonal, open up a channel and then send the ball up that channel. It would get on top of the bricks and

01:15:19.360 | just bounce down and take all the bricks out, just bouncing around from top. And human, because it took

01:15:24.800 | super precise moves, but it just learned like, oh, that's the right thing to do. There's another example of

01:15:30.160 | this. It's often given by reinforcement learning experts where they're trying to train it to do

01:15:34.320 | well with a boat racing video game where you collect points for doing, getting various rewards as you

01:15:40.800 | make your way through the track or whatever. And the policy that came up with, because all it's trying

01:15:44.720 | to do is maximize its points. Is it figured out like, forget the racing. There's a place where the

01:15:50.280 | rewards recharge pretty quickly. I'm just going to endlessly go in this loop, kind of crashing into

01:15:56.580 | things. And I'm just going to keep getting these, these coins that appear again and again and again

01:16:00.060 | in this like spiral or whatever, because it's a policy that did really well at maximizing reward

01:16:04.800 | function. Whereas if you went to chat GPT and said, let me explain to you, like I'm a situation,

01:16:11.620 | it's a boat and I'm trying to drive the boat and here's what's around me. Where should I go?

01:16:15.920 | It'll give you like pretty reasonable, like, this is like the type of thing someone would say here,

01:16:19.620 | like avoid the obstacles, you know, go, go this direction. So there are two different things that

01:16:25.640 | are going on. I think the place, if you want to be worried about things, well, first of all,

01:16:30.600 | keep this in mind. These are two different technologies. Let's just start there, right?

01:16:33.480 | All they have in common is making big networks bigger tends to lead to more intelligence,

01:16:39.200 | but the way they're trained are completely different. A breakthrough in one does not mean a

01:16:42.960 | breakthrough in the other. Either of these could get stuck in different places and the concerns are

01:16:48.060 | completely different. So they, you can't just assume they're both innovating at the same level,

01:16:52.160 | but I'm saying, if you want to get a little bit more worried about something, the sci-fi horror

01:16:56.500 | stories, the sort of like cautionary tales, you would worry about that reinforcement learning world

01:17:01.040 | because it's not trying to estimate how people would talk. It will learn whatever it can to try to get

01:17:07.220 | that reward function maximized. And it could do things you don't expect. And if you give a reinforcement

01:17:12.520 | trained model actuation, like it can control the real world and just say it. I just

01:17:18.040 | trust its policy will work well. It can go to really weird places because you don't know what

01:17:23.900 | that policy is. It's very different than a data-driven model. You can have a data-driven model

01:17:27.620 | is like, it's trying to be as close to a person as possible. It's close to whatever process was

01:17:32.460 | producing the text that trained on. It wants to be as close to that as possible. So there's maybe more

01:17:37.020 | to be worried about in that reinforcement learning world than there is in the chat GPT world. But keep

01:17:44.040 | in mind, these are different technologies. And when you think about things that are learning how to

01:17:47.780 | work in the world and walk and locomode and play games, that's not the same thing that's happening

01:17:52.460 | with chat GPT. That's a completely different technology and one that we should keep an eye on.

01:17:56.740 | Anyways, it's an important distinction. These worlds are kind of coming together. There's people who are

01:18:00.320 | using language models to help do reinforcement learning. Jan LeCun doesn't like reinforcement models

01:18:05.100 | at all. He thinks we should just build up understanding of the world and have sort of more intentional

01:18:10.000 | simulation of the world. I agree with that because I think it's a lot safer. That's where my

01:18:13.480 | intentional AI model comes in. A lot to talk about here, but at least we have this distinction

01:18:18.020 | to start with. Reinforcement learning is different than data-driven learning,

01:18:22.440 | semi-supervised learning, unsupervised learning rather. These are different models, different

01:18:27.660 | technologies, different capabilities, different concerns. All right. Well, there we go, Jesse.

01:18:33.460 | That was my lecture for the day.

01:18:34.320 | I like it.

01:18:35.300 | Let's leave it there, but we'll be back next week with another episode. And until then, as always,

01:18:39.800 | stay deep. Hey, if you liked today's conversation about managing your time in five minutes or less,

01:18:45.140 | I think you'll also like episode 261, which is about controlling your time. Check it out. I think

01:18:51.760 | you'll like it. Actually, I want to shift towards the more practical world of controlling your time.

5 Minutes a Day For Peak Productivity? - This Simple Hack Might Change Your Life | Cal Newport

Chapters