The collapse of modern attention (and how to get it back) - Cal Newport

English
CChris Williamson
ManagementBooks & LiteratureTelecommutingMental HealthComputing/Software

Transcript

00:00:00- Dude, you must be feeling like Cassandra at the moment.
00:00:04So prescient, the distraction, the necessity of deep work,
00:00:09the inherent bombardment of our attention.
00:00:11Do you feel like you saw the future earlier
00:00:14than what even at the time maybe felt late with deep work
00:00:17and focusing on quality over quantity and stuff?
00:00:20- I mean, I think part of what I noticed
00:00:21was the present was crazy to me
00:00:24and no one else recognized it.
00:00:25So it was less even predicting the future.
00:00:27I feel like there was a time, God, it's like 10 years ago now
00:00:31where I was looking around and yeah, saying two things.
00:00:34One, the social media doesn't make sense.
00:00:36Why are we all pretending like this is at the center
00:00:39of democracy and civic life and all business?
00:00:41We all have to be on here all the time.
00:00:43And two, email doesn't make sense.
00:00:45Not what was gonna happen in the future.
00:00:47I'm just like looking at the way we're working today
00:00:50with email and Slack and Teams was coming.
00:00:51Like this completely does not make sense.
00:00:53You're switching your context once every two or three minutes.
00:00:55This is a terrible way to actually use your brain.
00:00:58So I never thought of myself as predicting the future
00:01:00as much as just telling people what was going on then
00:01:01didn't make sense.
00:01:03And everyone thought I was crazy.
00:01:04And 10 years later, it just kind of jumped from
00:01:06I was crazy to it's common sense.
00:01:08So it's not even that interesting that I'm saying it anymore.
00:01:11So I kind of skipped the part where it sounded prescient.
00:01:15- Do you feel vindicated?
00:01:16- I think certainly on a couple issues.
00:01:20The social media issue was a big one
00:01:22because I used to get a lot of flack for that.
00:01:25For going out and I wasn't even saying
00:01:27the social media was bad
00:01:29or that no one should use it.
00:01:31Really what I was pushing back on
00:01:33was just the idea of ubiquity.
00:01:35The idea that everyone had to use it.
00:01:37I said, this doesn't make sense.
00:01:37I get there's some people this makes sense for.
00:01:39There's a lot of technologies that have markets
00:01:41that make sense for it.
00:01:42But why is there this pressure
00:01:43for everyone to be on these services?
00:01:45This is not going to a good place.
00:01:47They're spending a lot of money to mine attention
00:01:50and they're gonna get better at it, right?
00:01:52And at the time this was considered crazy.
00:01:55What do you mean like you wouldn't use social media?
00:01:58I wrote a New York Times op-ed back.
00:01:59I looked this up the other day.
00:02:01It was 2016.
00:02:03And it argued maybe social media is not the biggest thing
00:02:07for a young person to focus on
00:02:09if they're thinking about their career.
00:02:10That's what it was.
00:02:11It was like focus on your career instead of social media.
00:02:13Actually doing things well is what really matters.
00:02:15And you would think that I had just come on
00:02:19and said like America has an idea is done
00:02:21and grandmothers should be kicked.
00:02:22Like people were upset about this.
00:02:24The New York Times commissioned a response op-ed
00:02:26two weeks later that went through mine
00:02:30and said this is what is wrong
00:02:32about Cal Newport's op-ed or whatever.
00:02:34'Cause it made such a fear to suggest it.
00:02:36And today it's boring to suggest like,
00:02:38ah, you know, social media has problems
00:02:39and most people probably shouldn't use it.
00:02:41People agree with that.
00:02:42The one that upsets me though,
00:02:44that one I feel like people have come along too
00:02:47and more and more people are being much more selective
00:02:49and minimalist about their social media.
00:02:51The distraction, email, Slack,
00:02:53constantly jumping back and forth between different things.
00:02:56That just got worse.
00:02:57I mean, I think people recognize it now.
00:02:59This is probably not a good way to work.
00:03:01But I thought because there was dollars and cents here,
00:03:04this is less productive
00:03:05from an economic productivity standpoint.
00:03:07To have all of your workers
00:03:09changing their attention all the time.
00:03:11You're just getting a really low return
00:03:12on all the money you're investing in these human brains.
00:03:14So I thought, oh, this is dollars and cents.
00:03:16This is the one that's gonna change.
00:03:18Social media is fun.
00:03:19Like that's gonna be hard to change people's behavior.
00:03:21But certainly this hyper distraction thing
00:03:23and knowledge work, that'll change
00:03:24because we're leaving money on the table.
00:03:26It hasn't changed at all.
00:03:27It's gotten worse.
00:03:27It's worse than it was.
00:03:29I'm at the 10 year anniversary now of the book, "Deep Works."
00:03:33So this month is a 10 year anniversary.
00:03:35- Congratulations, dude.
00:03:36That's fucking seminal.
00:03:37Like that has become a part of the lexicon.
00:03:39That's really, really cool.
00:03:41- Yeah, but it's got me a little bit depressed
00:03:43because I've been doing this 10 year reflection.
00:03:46Like, okay, it's been 10 years and the book was a hit
00:03:48and it's millions of copies, et cetera.
00:03:50And that is the issues I talked about are worse.
00:03:52They're like really worse than they were 10 years ago.
00:03:55So people know the problem, nothing has changed.
00:03:57- What does the data suggest around the worstness of it now?
00:04:02- The one I've been following,
00:04:06the study that I think is useful as a trend line
00:04:08is Microsoft actually does this annual report
00:04:11where they gather data from Microsoft 365.
00:04:14So it's like Office and Word and PowerPoint and Excel.
00:04:17Nowadays you use this sort of the web based version of these.
00:04:19It's very common.
00:04:20So they can gather data from just tens of thousands
00:04:23of knowledge workers actually using
00:04:25all these different tools.
00:04:27And the latest report they put out in 2025
00:04:29now has the interruptions on average once every two minutes.
00:04:32So it's just gotten out of control.
00:04:36So switching to a communication tool once every two minutes.
00:04:38They also found the latest report
00:04:40and this is depressing to me as well.
00:04:42There's one time in the week where they see a notable rise
00:04:47in the use of the non-communication.
00:04:49So actually using the core productivity tools like Word
00:04:52or PowerPoint and it's Saturday and Sunday morning.
00:04:55So we've just put the work off until the weekend
00:04:58when there's no expectations of responses
00:05:00and spend the actual weekdays talking about work.
00:05:04Which I just don't get.
00:05:06Like that is not economically productive.
00:05:08Like companies are leaving money on the table
00:05:10but it's just where we are.
00:05:11We really can't quit this behavior.
00:05:13- Isn't it interesting that you had to try and appeal
00:05:15to a very utilitarian approach for this?
00:05:17But you didn't say this is probably making staff miserable.
00:05:22It's not a good use of time.
00:05:24We've got some really strong evidence that suggests
00:05:26that doing one thing and getting better at it
00:05:29over a protracted period of time
00:05:31actually makes you feel more satisfied.
00:05:32You get into a flow state, et cetera, et cetera.
00:05:35You look back on your day
00:05:36and you can look at the things that you did.
00:05:39None of that, which is the much more immediate experiential
00:05:44way that people interface with distraction.
00:05:47You tried to appeal to the bottom line,
00:05:49which you thought, incentives, incentives,
00:05:51align the fucking incentives.
00:05:54And that didn't work, which obviously means also
00:05:57that people's level of administrative burden misery
00:06:02is also coming along for the ride at the same time.
00:06:04Yeah, it's a fucking mess, dude.
00:06:06And I think even with what I do, it's not a very big team,
00:06:09but Slack is like, it's so useful and invites so much chaos
00:06:17at the same time.
00:06:18It is, and was Slack, Slack wouldn't have been that big
00:06:21during Deep Work, I'm gonna guess.
00:06:23It wasn't big, it wasn't out yet.
00:06:25I talk in Deep Work about these very early
00:06:27instant messenger tools that no longer exist,
00:06:30like HipChat, it was just emerging among the programmer class.
00:06:33I was basically saying there be dragons,
00:06:35like let's be careful about that.
00:06:37But I wrote an article about Slack years later
00:06:41when Slack was bought.
00:06:42So I think Salesforce bought Slack.
00:06:44I wrote an article about it for The New Yorker.
00:06:46And I think the title of that article gets to the core
00:06:48of the issue you're talking about.
00:06:50The title was Slack is the Right Tool
00:06:53for the Wrong Way to Work.
00:06:54And I think what happened, here's my whole theory on Slack,
00:06:58is that when email arrived, it moved us to this new style
00:07:01of collaboration that I call the hyperactive hive mind,
00:07:04where we'll just figure things out on the go
00:07:06with ad hoc back and forth, unscheduled messaging,
00:07:08just sort of like shooting messages back and forth.
00:07:10We'll figure things out,
00:07:11like we're all just kind of connected all the time.
00:07:14That's a terrible way to work for all the reasons
00:07:16I talked about.
00:07:16It's distracting, it's context switching,
00:07:18you can't do anything deep, it's hard to produce value.
00:07:20But if that's the way you're gonna work,
00:07:22email clients are not a very good tool for that.
00:07:25You have threads and it's clunky
00:07:27and it's hard to search through your email
00:07:29and find what you did before.
00:07:30So Slack came along and said,
00:07:31look, if this is the way you're gonna work,
00:07:33hyperactive hive mind, constant back and forth,
00:07:35ad hoc coordination, we'll build you a better tool for that.
00:07:38So that's why people both love and hate Slack.
00:07:41It's a really good tool for that style of collaboration.
00:07:44It works really well.
00:07:45But that style of collaboration makes us miserable.
00:07:48So it's this weird love-hate relationship we have,
00:07:50like this works great.
00:07:51I hate the thing that is making it easier.
00:07:53- Why does it make us miserable, that style of collaboration?
00:07:57- Because our brain isn't meant to switch
00:07:59our target of attention that quickly.
00:08:00It just takes us a long time
00:08:02if we're talking about targets
00:08:04that are abstract and symbolic.
00:08:06It takes us a long time to switch from one to another.
00:08:09Physical world targets, we can switch quickly, right?
00:08:11We're wired for that.
00:08:12If there's a tiger's roar, I can boom.
00:08:15A hundred percent attention, what's going on over there.
00:08:17But when we're thinking about abstract things,
00:08:20information, ideas, things that are symbolic
00:08:22and in our head, that's us.
00:08:24We're basically reappropriating our brain hardware
00:08:27to do something we're not evolved to do.
00:08:29It takes a lot of effort to do symbolic thinking,
00:08:31to think about abstract concepts.
00:08:34And we know it takes 10 to 20 minutes
00:08:36to fully change our attention context
00:08:39from one abstract target to another.
00:08:40It takes a long time.
00:08:41That's why if you sit down to write something,
00:08:43everyone has this experience.
00:08:45The first five or 10 minutes, you're like,
00:08:46"Man, this is terrible."
00:08:47Like, "I'm making no progress," or whatever.
00:08:49And then after a while, you're like,
00:08:50"Oh, this is starting to flow."
00:08:52Like, "It's going better."
00:08:53That's because it took that much time
00:08:55for your brain to load up all of the relevant information
00:08:59and to inhibit all the unrelated circuits
00:09:01and get your brain really ready to do that activity.
00:09:03So if you now interrupt that brain once every two minutes,
00:09:06it never can lock in on anything.
00:09:09And what you feel then
00:09:09is this sort of diffuse cognitive friction
00:09:12that we begin to experience as fatigue, cognitive fatigue.
00:09:15And it's a really frustrating experience.
00:09:17It's why if you go to an email inbox,
00:09:20you're like, "I have time.
00:09:21"I'm gonna empty this inbox.
00:09:22"I'm gonna go message by message.
00:09:23"Here's the best way to do it, write on paper.
00:09:25"I'm gonna go message by message,
00:09:27"and I'm gonna answer these messages."
00:09:28Why does that get so hard?
00:09:30Why do you find yourself like jumping around
00:09:32and looking for easier messages?
00:09:33Because each message is a different context than the other,
00:09:35and that's torture for the brain.
00:09:37It's really, really hard to go from, all right,
00:09:40this is a complicated question
00:09:41one of my employees is asking me,
00:09:43and now this is a completely different issue,
00:09:45completely unrelated to that,
00:09:47where I have to think up a good title for something,
00:09:48and now here's a completely different issue,
00:09:50and you're trying to switch one after another.
00:09:52Our brains aren't wired for that.
00:09:54It really makes us unhappy.
00:09:55- What would you say to someone
00:09:58who wants to try and retrain that attention?
00:10:00Maybe they're gonna try and make some sort of a stand
00:10:07inside of Slack and say,
00:10:09"I will only be available at certain times of the day,"
00:10:11but regardless of the inbound,
00:10:14let's say that they fixed the inbound,
00:10:15'cause that's a totally separate problem.
00:10:16That's much more sort of structural,
00:10:18unless you've got any advice for that as well.
00:10:21But how does someone go about re-appraising,
00:10:26retraining their mind away from that?
00:10:29Because we do become, we get like Stockholm syndrome,
00:10:34the Slack Stockholm syndrome, where our captor tormentor
00:10:38becomes the way that we operate.
00:10:40We've got our favorite little ways of working,
00:10:42and it feels like we've done,
00:10:43but then at the end of the day,
00:10:44we look back and have this sort of odd malaise thing
00:10:48about, well, what did I actually do today?
00:10:53What got done?
00:10:54Well, not much, not much got done.
00:10:57- Yeah, well, it's hard unilaterally.
00:11:00If you've changed nothing else about your workload
00:11:02or your communication protocols,
00:11:03if you just say, I'm not gonna be on Slack
00:11:07from this hour to this hour,
00:11:08I only check my email twice a day
00:11:09or whatever that standard advice was from 15 years ago,
00:11:12it doesn't work well.
00:11:13Because if you're involved in a large number of projects
00:11:16that are timely, and the way progress is gonna be made
00:11:19is with ad hoc back and forth messaging,
00:11:21you have to be in there checking.
00:11:23That's the brutal part of the hyperactive hive mind,
00:11:26is that it has defenses to its elimination
00:11:28built into its very nature.
00:11:30Because if this is how we're gonna figure this out,
00:11:32like we have to have five or six back and forth messages
00:11:35to figure out what we're gonna do
00:11:36about this client coming tomorrow,
00:11:37we have to get this done today.
00:11:39That means you have to see my next message right away
00:11:42so that we have time for me to answer you
00:11:43and you to answer me and for that ping pong match to happen.
00:11:46That means you have to be checking your inbox
00:11:48or Slack constantly.
00:11:49Otherwise you're not gonna see my next message in time
00:11:51for this whole game to unfold.
00:11:52So the very nature of that style of collaboration
00:11:55demands constant inbox checking,
00:11:57which is what I think people often get wrong about this.
00:11:59When I think about things like Slack or email,
00:12:01they think too often about either information,
00:12:04like, oh, I've got so many messages in my inbox
00:12:07that I don't need.
00:12:08I have all these newsletters and spam.
00:12:10That's not a problem.
00:12:11That's a minor problem.
00:12:13That's an easily solvable problem.
00:12:15It's like clutter.
00:12:16That's not a big problem.
00:12:18The issue is actually my collaboration style
00:12:22requires me to be in there
00:12:24because if I miss messages in a timely fashion,
00:12:26everything falls apart.
00:12:27And so the issue is not, how do I interact with my inbox?
00:12:31It really has to be,
00:12:33how do I change the way the inbox is being used?
00:12:37I mean, so I ended up,
00:12:38I feel like had three big ideas on this
00:12:40that span three different books, right?
00:12:41So in deep work, like one of the big ideas was,
00:12:45you can train your personal ability to focus.
00:12:48Focusing is really important.
00:12:50Putting aside for now,
00:12:51all the things trying to prevent you from focusing,
00:12:53you have to practice it.
00:12:54And if you practice it, you'll get better at it.
00:12:56And if you get better at it,
00:12:57you'll be a superstar because like,
00:12:58that's what matters in the knowledge economy.
00:13:00Everything good comes out of focus.
00:13:02Then I wrote a book after that called "A World Without Email."
00:13:06And in that book, I was arguing the way we,
00:13:10the thing I was telling you about, hyperactive hype mind,
00:13:11communication is a problem.
00:13:12This is a real problem.
00:13:13The fact that we are using this method for coordination
00:13:16is causing all these trouble, is really causing problems.
00:13:20And I went through all the data and all the research
00:13:22and made the case, this is super non-productive.
00:13:24I went back to the archives of the New York Times
00:13:27business session in the 80s and 90s to exactly document
00:13:30the rise of email and how people were talking about email
00:13:32when it first came onto the business scene.
00:13:34And I made the case, the way we work is arbitrary.
00:13:38This hyperactive hype mind was not a plan.
00:13:40It wasn't seen to be more productive.
00:13:41We stumbled into it, so we really should change it.
00:13:43So that was that book.
00:13:45And then the most recent book,
00:13:46"Slow Productivity" from a couple of years ago.
00:13:48In that book, I argued, oh, wait a second,
00:13:51workload matters too.
00:13:52The other issue with this problem is we don't put any limits
00:13:55or transparency on how many things we're working on.
00:13:58And if you pile too many things on your plate,
00:14:01too much communication interruption becomes unavoidable
00:14:03because they each have little issues
00:14:05they need you to deal with.
00:14:05So I've now, over this 10 year period,
00:14:08have kind of broken down this problem.
00:14:10There was like training yourself to focus,
00:14:12fixing your communication protocols.
00:14:15Like how do I communicate in a professional context?
00:14:17How do we collaborate?
00:14:19And then managing workload to be more reasonable.
00:14:21All three of, and this might be why
00:14:23this problem's not solved.
00:14:24There's no one thing to fix, right?
00:14:25So all three of these things go into the issue
00:14:28and they're each complicated.
00:14:30- What, across those three books, all of which are great,
00:14:33and everyone needs to go and check out.
00:14:34I think we've done episodes about each of them
00:14:36so they can just go and listen to those.
00:14:37- Yeah, I think, yeah.
00:14:38- And then buy the books.
00:14:39Looking back across this portfolio of productivity advice,
00:14:45what have you heard from readers
00:14:50or what has been the stickiest strategies for you?
00:14:55You look back and you go, okay, that's the 80/20
00:14:58of what I've published over the last three books.
00:15:01- To me, I think the big two
00:15:04that give you the biggest results,
00:15:06and I'll tell you the one that's the hardest,
00:15:07and this is why this book probably sold the least.
00:15:10The big two that gives you the biggest results
00:15:12is taking focus seriously like a skill.
00:15:14That really does make a difference.
00:15:15Practicing focus, you get better at it.
00:15:18And it has a demonstrable difference.
00:15:22You sit down to work and you're just producing better stuff,
00:15:26or you're trying to pick up some complicated new thing.
00:15:28Like, oh God, I can learn this faster.
00:15:30That makes a huge difference.
00:15:31And then the second one, which was more recent in my life,
00:15:34was, oh, you really gotta control the workload.
00:15:36So much is downstream from how many things
00:15:39you've agreed to work on.
00:15:40You have to leave the mindset of everything I say yes to
00:15:43brings with it value.
00:15:45So saying yes to more things,
00:15:46it's just gonna aggregate more value.
00:15:47That's not the right mindset.
00:15:49That's not the way, it's a nonlinear reward function there.
00:15:54There's a certain point as you add more things
00:15:56that not only does value stop growing,
00:15:59it begins to go down on the other side.
00:16:00And that there's a real saying no to many more things
00:16:03is actually a way to optimize reward and output,
00:16:08which is not natural.
00:16:10It doesn't make sense at first.
00:16:11It doesn't feel like common sense.
00:16:12So workload and focus training,
00:16:15you can control those more than you think,
00:16:17and you're gonna have huge results from those.
00:16:19- A quick aside.
00:16:20Do you remember learning about the mighty mitochondria
00:16:22back in grade school?
00:16:23Here's a quick refresher.
00:16:24It's the tiny engine inside of your cells
00:16:26that powers everything you do.
00:16:28But here's what they didn't teach you.
00:16:30As you age, your mitochondria break down.
00:16:32That's what can cause you to feel tired more often,
00:16:35take longer to recover,
00:16:36and wake up feeling like you're never fully recharged,
00:16:40no matter how long you sleep.
00:16:41I started taking Timeline nearly two years ago
00:16:44because it is the best product on the market
00:16:47for mitochondrial health,
00:16:48and that is why I partnered with them.
00:16:49Timeline is the number one
00:16:51doctor-recommended urolithin A supplement
00:16:54with a compound called Mitopure.
00:16:55Basically, it helps your body clear out damaged mitochondria
00:16:59and replace them with new ones.
00:17:01Mitopure is backed by over 15 years of research,
00:17:04over 50 patents, and nearly a dozen human clinical trials.
00:17:07It was recommended to me by my doctor,
00:17:08and that is why I've used it for so long,
00:17:10since way before I knew who even made the product.
00:17:13And best of all, there's a 30-day money-back guarantee,
00:17:16plus free shipping in the US,
00:17:17and they ship internationally.
00:17:18So right now, you can get a free sample,
00:17:21or get up to 20% off by going to the link
00:17:23in the description below,
00:17:24or heading to timeline.com/modernwisdom.
00:17:28That's timeline.com/modernwisdom.
00:17:32The learning to say no thing is interesting,
00:17:35especially as people progress inside of their career,
00:17:40and they get better at what they're doing.
00:17:43They have to learn to be able to say no
00:17:45to opportunities that they would've only begged
00:17:47to have had the opportunity to be in the room
00:17:49to have maybe said yes to only half a decade ago.
00:17:53In that time, you've had to go from needing that opportunity
00:17:58to actively being able to say no to something
00:18:00that's probably better than it.
00:18:01Alex, my friend, taught me about,
00:18:04you remember in "The Matrix," the woman with the red dress,
00:18:07and Neo turns around, and he says,
00:18:09"Well, you're looking at me.
00:18:10"We're looking at the woman in the red dress.
00:18:11"Look again," and it's an agent with a gun in his face.
00:18:14And the analogy that Alex used was,
00:18:17now imagine that she's not a 10 out of 10,
00:18:20but imagine 1,000 hypothetical 1,000s out of 10,
00:18:25and you need to be able to say no to them,
00:18:27which previously you didn't even know existed.
00:18:30So this, I think, the kind of,
00:18:32it's almost like reverse entropy or habituation.
00:18:38You know, your opportunities get better,
00:18:40which means that your capacity to say no
00:18:42needs to get better more quickly than that.
00:18:45You can't be chasing your tail trying to learn
00:18:48to be able to say no less quickly
00:18:49than the opportunities get more seductive.
00:18:52- Yeah, it's almost perverse, the way that works.
00:18:55It's like when you have all the time in the world,
00:18:58all you want is opportunities.
00:19:00And then when you have opportunities,
00:19:01all you want is all the time free in the world.
00:19:05I had to change, I don't know what you do,
00:19:06but I had to change my rule at some point.
00:19:08This was hard for me, to the default no.
00:19:11Like that's just how I have to operate now.
00:19:13'Cause as soon as you try to have a triage rule,
00:19:17well look, I'm not gonna do this opportunity unless,
00:19:20I only do speaking gigs that have this much money or this,
00:19:24or I only gonna go meet with someone
00:19:26if they're like this interesting or this or that.
00:19:28Eventually the number of things that satisfy
00:19:30that criteria overwhelmingly just as well.
00:19:32It's just so I've just had to fall back on the default no.
00:19:36- You're talking to somebody who came back
00:19:39from a two day trip to Qatar at the start of this week.
00:19:43So I spent as much time traveling
00:19:46as I did in the country to give a talk.
00:19:49And as I looked around, there was this first,
00:19:51the first night dinner, there was maybe 300 people there.
00:19:54And I'm talking to Logan Paul
00:19:56and Steven Bartlett over his shoulder.
00:19:58And the CEO of Qatar Airways is here
00:20:00and the Middle Eastern director for Metis over there.
00:20:02And I was looking around thinking,
00:20:04everybody here wants to be here, it's very exciting.
00:20:07Everyone's really lovely, but also everyone here can't say no.
00:20:10Everybody in this room is chronically incapable of saying no.
00:20:15- I've said no to this one several times by the way.
00:20:18The amount of invites to Qatar and the UAE and other places,
00:20:23I have said no to many of those.
00:20:25- All right, well, consider me a fucking,
00:20:26consider me a slut compared to you, Cal, whatever it is.
00:20:30I must be easy, an easy booty call.
00:20:32They tried to get Cal Newport.
00:20:33We couldn't get Cal, so we'll ring Chris instead.
00:20:35- The default no, oh man, yeah.
00:20:37It's crazy the things you end up saying no to after a while,
00:20:40but I mean, there's a currency shift.
00:20:41For me, time to think is such a valuable,
00:20:44that's a more valuable currency than money, right?
00:20:46You get to a point where you're like, oh, I'm doing fine.
00:20:49But if I don't have time to think, what's the point?
00:20:51And then that becomes this like really rare currency
00:20:53that's much harder to get a hold of.
00:20:57And that's the only way I can protect it now
00:20:59is anything that requires me to like go somewhere,
00:21:01it's a default no.
00:21:01And then I can talk myself out of it later, right?
00:21:04I'm like, you know what?
00:21:05I could bring my family with it.
00:21:06We could have a trip, right?
00:21:07So actually, you know what?
00:21:08I will do this.
00:21:10Or I just did a masterclass course released this week.
00:21:15I spent a year and a half say no to that.
00:21:20And then like eventually I sort of talked myself,
00:21:22I talked to some people.
00:21:25They're like, we'll come to DC to do it.
00:21:26I talked to James Clear, just done one.
00:21:28And I had a good talk with him about it.
00:21:30And I was like, you know what?
00:21:31This will be interesting.
00:21:32And it took me a year and a half,
00:21:34'cause I finally talked myself into it.
00:21:35So I will say yes, but it just,
00:21:37the default no means that you don't have to-
00:21:39- To high standards.
00:21:40- Yeah, you don't have to run it through the ringer.
00:21:41And then you're like, okay, if it really sticks with me,
00:21:43then maybe I'll be like, all right, all right, I'll do it.
00:21:46- How much should people actually be working?
00:21:48- Well, it depends what you mean by work
00:21:50and what they're doing, right?
00:21:52Because think about it, let's say you're an athlete.
00:21:55It's super well-defined.
00:21:56Like here's optimal training, here's optimal rest.
00:21:59And like, that's what you should be doing.
00:22:00Like that's really clear.
00:22:03We don't have those limits as clear in the culture
00:22:05for other types of jobs that we probably should.
00:22:07If you're at a high wage hourly build job,
00:22:11like a law partner at a big law firm,
00:22:14there the economic model is the more you work,
00:22:18the more profitable it is.
00:22:19And we'll pay you big money to do this,
00:22:21but like you should basically work as much as you can
00:22:23that your body will take it.
00:22:24That's the economic engine.
00:22:25That's why I think those jobs are scary.
00:22:28If you're a novelist that writes literary fiction,
00:22:31so you're like, I really need to be award nominated
00:22:35for each book, or I'm gonna fall out of this like slipstream
00:22:38of, because no one's gonna read these books
00:22:39unless they're some of the best books.
00:22:41Then you should be doing like four hours in the morning
00:22:43and then just disappear, right?
00:22:46Like you should be doing very little more work than that.
00:22:49'Cause almost anything else will get in the way
00:22:51of you like sticking in that position.
00:22:53And so it all just depends on what you do.
00:22:55- Didn't you look at some experiment of shorter work weeks?
00:23:01- Yeah.
00:23:02- What did you learn from that?
00:23:03- There's a lot of these right around the pandemic,
00:23:06right before and then right after in Europe and Iceland.
00:23:11So some European studies, I think Germany did one,
00:23:13Iceland did one, UK did one.
00:23:15And they were looking at four day work weeks.
00:23:17So what would happen if we take away one day?
00:23:20The interesting thing about those experiments
00:23:22is what they found is whatever measures of productivity
00:23:25they came up with, they didn't get worse,
00:23:28which I thought was very interesting.
00:23:30They took a day away and yet the perceived productivity
00:23:34or the measured productivity didn't go down.
00:23:36And there's two ways to look at it.
00:23:37The one way to look at it is to say,
00:23:38oh, this means that like we should have a four day work week
00:23:40because things didn't get worse.
00:23:42And okay, maybe, right?
00:23:44But to me, there was like a bigger observation
00:23:46that came out of that, which is like, wait,
00:23:47so what are we doing during the work days?
00:23:51Like there's something going on here
00:23:54that should really catch our attention.
00:23:55What does work mean?
00:23:57That we could take an entire day off the table
00:23:59with no other preparation
00:24:01and the valuable stuff being produced doesn't change.
00:24:04This tells us that like whatever we're doing
00:24:07while we're sitting here in work is not just sitting down
00:24:09and trying to produce value.
00:24:11We're clearly have all sorts of other sorts
00:24:13of distractions going on, context switching,
00:24:16time that's being devoured, Parkinson's laws at play.
00:24:19Work must be broke.
00:24:21To me, that was the more important observation is that like,
00:24:23if you can take away a day and nothing changes,
00:24:25then I don't think we're doing in the office
00:24:27what we think we're doing in the office.
00:24:29- Parkinson's law was on the tip of my tongue.
00:24:31Work expands to fill the time given for it.
00:24:33And if you give people five days, they'll take five.
00:24:36And if you give them four days, then they'll do it in four.
00:24:39Look, everybody knows just how much time they waste
00:24:44not doing the work,
00:24:46not doing the thing that they're supposed to do.
00:24:47And this isn't victim blaming.
00:24:50This is a lot of the time dealing with admin
00:24:52and necessary meetings.
00:24:53You can't get out of them.
00:24:54You have to be there for whatever reason.
00:24:57So it's not as if it's bottom up.
00:24:59A lot of it is top-down dictated.
00:25:01This is the environment that you work in
00:25:02and you have to do this.
00:25:04But even outside of that,
00:25:05when you do have your one hour in between meetings,
00:25:09your inability to not...
00:25:11I remember when I used to run nightclubs
00:25:13and I'd get in at 2.30 in the morning,
00:25:16the final part of the night was cashing the till.
00:25:19So this was before we switched to tickets,
00:25:22which was sort of the late teens, just before COVID.
00:25:26Digital tickets online, which meant that you didn't have
00:25:28to cash as much money in the till.
00:25:30But before that, it was all five pounds and 10 pound note
00:25:33and 20 pound notes and single pounds
00:25:35and all the rest of it.
00:25:36And I would go into the office with the manager of the venue
00:25:39and we would be counting the money.
00:25:40But this is the final task.
00:25:42The final bit of the night.
00:25:43It's fucking 2.15 or 2.30 in the morning.
00:25:45We've just taken the till off as it's called.
00:25:48Anybody that's coming in doesn't get to come in.
00:25:50We're not gonna take any more money.
00:25:52And I'm sat up there doing light lift mental arithmetic.
00:25:57But for me, somebody who hadn't done math since I was 16,
00:25:59it was a relatively heavy lift.
00:26:01Flicking through the money, flicking through the money,
00:26:04huge fluorescent overhead lights just before.
00:26:06And then I get to drive home and I'm thinking about it
00:26:08and I got to go put the money in the till
00:26:09and I got to write it in the spreadsheet
00:26:11and then I get into bed.
00:26:12And as I got into bed, my eyes below my eyelids
00:26:15would start flicking left and right.
00:26:17I wouldn't be able to tune myself.
00:26:19I'm also doing this, let's not forget,
00:26:21in a sweaty beer stinking office above a room going,
00:26:26(imitates drum beating)
00:26:29I've had to walk through the club.
00:26:30I've had to shout at the hostesses.
00:26:32One of them's getting fingered on the dance floor.
00:26:33Stop doing that.
00:26:34You're supposed to be at work.
00:26:35The DJ's pissed.
00:26:36I need to, you know, it's chaos.
00:26:37And I've tried to coordinate this orchestra of bullshit.
00:26:40And then I've had to do mental arithmetic.
00:26:41And then I get to drive home and then I'm like,
00:26:43okay, chill out brain.
00:26:45It doesn't want to.
00:26:46And that eyes moving left and right thing,
00:26:49I think is the sort of optical equivalent,
00:26:53ocular equivalent of how people feel
00:26:57when they finally get a moment.
00:26:58It's like, okay, all of my stuff is done.
00:27:01And then they try and sit down to work on the thing
00:27:05that ostensibly that's actually there to do, right?
00:27:08'Cause all of the other bullshit, the meetings,
00:27:09you're not there to do the meetings.
00:27:10You're not there to do the Slack.
00:27:11You're not there to do.
00:27:12All of that is foreplay to get you to do the thing
00:27:15that you're there to do.
00:27:16And then you sit down to do the thing you're there to do.
00:27:19And your eyes are moving behind your eyelids
00:27:21is the equivalent.
00:27:21You're swiping and moving across the screen
00:27:24and you've got a few different other,
00:27:25well, just check on this thing.
00:27:27Like what the living fuck is going on?
00:27:28I've like trained, the environment that I work in
00:27:32has trained me out of being able to do my work.
00:27:36- Well, we are meant to do,
00:27:38like what would be the ideal workday
00:27:40in an office environment
00:27:41that would actually mask the human brain?
00:27:43It would probably be you come in,
00:27:45you work on something hard for a while.
00:27:47Like that's what you do in the morning.
00:27:49You have lunch and then you like catch up with,
00:27:52have some meetings, talk to some people,
00:27:54hey, what's going on and do some tasks and that's your day.
00:27:58Like that's basically what we can do, like two things.
00:28:00One big burst of like, let me focus on something hard.
00:28:03And then we can kind of come down the mountain after that
00:28:06with let me chat with people, what's going on.
00:28:08Some decisions need to be made or whatever.
00:28:10That's probably about optimal.
00:28:11Instead, we juggle a dozen to two dozen tasks
00:28:14that all have their own demands.
00:28:16They all have their own communication needs.
00:28:18This is why the Microsoft data shows,
00:28:20oh, the work happens Saturday and Sunday morning.
00:28:23It is really hard.
00:28:25You can't go from, and meetings are very hard as well.
00:28:28We think like, oh, I'm not actually doing work
00:28:29during meetings, but what you are engaging in a meeting
00:28:32is all the parts of your brain
00:28:34that deal with social interaction.
00:28:36And those are a large part of your brain
00:28:38and that is a fraught and mental energy consuming activity
00:28:42to sit in a room or on a Zoom screen
00:28:43and try to manage all these different people
00:28:46and how do I look and what am I saying
00:28:48and what's going on here and I have to say the right things.
00:28:50It's draining.
00:28:52And you come out of something like that,
00:28:54it's difficult just to jump right back into something else.
00:28:56And if you come out of something like that
00:28:58and there was a lot of obligations generated,
00:29:00oh, we discussed in this meeting things I need to do.
00:29:03And now you try to go straight
00:29:04from that meeting into another.
00:29:06Well, now that's really in the back of your head.
00:29:08What about this?
00:29:09What about this?
00:29:10We can't forget this.
00:29:11We just made our obligations.
00:29:12That feeling of fatigue.
00:29:13It's really as fatigue is what it feels like,
00:29:15a mental fatigue.
00:29:17Like there's a sand in your brain,
00:29:19sand in the gears of your brain.
00:29:20That's the state that a lot of people
00:29:22who work in front of a computer screen,
00:29:23like that's the state they're in most of the day
00:29:26and they don't even realize,
00:29:28oh, that's a bad feeling.
00:29:29That's a negative state.
00:29:30That's not how it needs to feel
00:29:32because you have nothing else to compare it to.
00:29:35Yeah, the amount of things we're doing,
00:29:36the amount we're trying to switch back and forth,
00:29:39I always thought that part of the problem
00:29:41was a lot of our current thought about work culture
00:29:45and hustling and what it means to produce
00:29:47was influenced by Silicon Valley in the '90s and 2000s
00:29:50because that was considered this very ascendant part
00:29:54of the economy through the 2000s,
00:29:56through the Steve Jobs era.
00:29:57We looked at Silicon Valley like,
00:29:58these are the coolest companies.
00:30:00They're doing all the coolest stuff.
00:30:02Over there, I think they adopted a model of work
00:30:06that was very inspired by computer processors, right?
00:30:09So because that was what was in the air in the '80s and '90s
00:30:12in Silicon Valley was the computer processor world,
00:30:14the 386 versus the 486 versus the Pentium
00:30:17and it was all about speed.
00:30:19And the thing with a computer processor,
00:30:20if you're a computer type,
00:30:22what matters is you never want the pipeline to be empty, right?
00:30:27You wanna always make sure you have stuff
00:30:29for that processor to do so it never waste time.
00:30:32The processor will, every command you give it,
00:30:35it operates the same as any other.
00:30:36It can switch.
00:30:37It doesn't care what they are.
00:30:38It just sits there and operates one command after another.
00:30:40And the whole game with getting processors to be effective
00:30:43is like don't have downtime.
00:30:44Like the real fear,
00:30:45I can put on my computer scientist hat for a second.
00:30:48The real fear in computer processor design
00:30:50is that you sometimes get to a command
00:30:52that's gonna generate a huge delay.
00:30:55So you say like, oh, go get something from memory.
00:30:58That takes a lot of time from the perspective
00:31:00like a computer processor cycle.
00:31:01It's just sitting there, cycle after cycle,
00:31:03doing nothing while you're waiting
00:31:05for the memory bus or whatever.
00:31:06So we invented these processor pipelines like,
00:31:08oh, while we're waiting to get something back from memory,
00:31:11here's some other stuff the processor can run
00:31:13so that it's never not working.
00:31:15And the idea was you wanna move as fast as possible
00:31:18and you never wanna have downtime.
00:31:20And that's how you get the most out of a computer processor.
00:31:22The human brain is like 180 degrees different.
00:31:25We can't just switch back and forth
00:31:27between unrelated commands.
00:31:29You switch me from one to another thing and boom,
00:31:3230 minutes of my mind is fried.
00:31:34Humans operate very differently,
00:31:36but I think Silicon Valley associated,
00:31:38it said here's the thing we're gonna associate
00:31:39with being really good at your job.
00:31:40It might've used to been, I don't know, your skill.
00:31:43It was Don Draper and Mad Men.
00:31:45Remember that conception of,
00:31:47was it mean to be good at your job?
00:31:48They weren't showing Don Draper grinding it out.
00:31:51Like man, Don Draper is like in the office
00:31:53till 3 a.m. every night or whatever.
00:31:56Now he took the five o'clock train
00:31:57back to Connecticut or whatever.
00:31:59It was, he was really, really good
00:32:02at coming up with ad copy.
00:32:03He was good at what he did.
00:32:05That's what you used to respect.
00:32:06And then after the 80s, 90s, Silicon Valley became pervasive.
00:32:10Like no, what matters is you never have a no op.
00:32:13You never have a down cycle.
00:32:15You might as well say yes to more things.
00:32:17You might as well get more emails.
00:32:18You never have time or you're not working.
00:32:20That's what productivity is gonna be.
00:32:22And that was a disaster for the human brain.
00:32:24- If you struggle to stay asleep
00:32:26because your body gets too hot or too cold,
00:32:28this is going to help.
00:32:29Eight Sleep just released their brand new pod five,
00:32:32which includes the world's first
00:32:33temperature regulating duvet.
00:32:35Compare it, the smart mattress cover,
00:32:37which cools or warms each side of the bed up to 20 degrees.
00:32:40And you've got a climate controlled cocoon
00:32:42built for deep uninterrupted rest.
00:32:44The new base even comes with a built-in speaker
00:32:46so you can fall asleep to white noise, nature sounds
00:32:48or a little ambient Taylor Swift, if that's your thing.
00:32:52And it's got upgraded biometric sensors
00:32:54that quietly run health checks every night,
00:32:56spotting patterns like abnormal heartbeats,
00:32:58disrupted breathing or sudden changes in HRV,
00:33:00which is why it has been clinically proven
00:33:02to increase total sleep by up to one hour every night.
00:33:06Best of all, they've got a 30 day sleep trial
00:33:07so you can buy it and sleep on it for 29 nights.
00:33:09And if you don't like it, they will give you your money back
00:33:11plus they ship internationally.
00:33:13Right now, you can get up to $350 off the pod five
00:33:16by going to the link in the description below
00:33:18or heading to eightsleep.com/modernwisdom
00:33:20using the code modernwisdom at checkout.
00:33:22That's E-I-G-H-T sleep.com/modernwisdom
00:33:26and modernwisdom at checkout.
00:33:28There's definitely an element of this
00:33:31that it's very public productivity.
00:33:36It's very obvious.
00:33:37Look at how hard I'm working, right?
00:33:39If you're the one that replies quickest on Slack or on email,
00:33:42then it's evident that you're the one,
00:33:44it looks like you're the one that's working hardest
00:33:47because you're the one that's most responsive.
00:33:49Whereas the person who's silently working on their own,
00:33:51they can't broadcast it by design.
00:33:53They can't broadcast it to everybody else.
00:33:55So yeah, this obvious productivity in a way
00:34:00is way less sexy.
00:34:01So I think the new elephant in the room is AI
00:34:06and how that is enabling an increase in pace of output,
00:34:11but almost certainly a decrease in quality.
00:34:18So fold AI into your existing worldview
00:34:23because to me it just seems like a huge force multiplier
00:34:28for what already was pretty sloppy, Slack, email,
00:34:33async communication that's always on,
00:34:35people taking their work home with them,
00:34:37never being able to not context switch,
00:34:38not focusing on quality and instead focusing on quantity,
00:34:42not being able to dial themselves in
00:34:44to do deep work for one moment.
00:34:46And now that is enhanced and magnified even more
00:34:51by the use of LLMs to help you put out more,
00:34:55to help you think less.
00:34:56So your focus is actually, you're in Slack with your LLM.
00:35:00It wouldn't surprise me if there is a LLM integration
00:35:03into Slack at some point in future.
00:35:05I don't know whether there is,
00:35:07where you can just do it in there.
00:35:09So you're just talking back and forth
00:35:10in one fucking workspace.
00:35:12Talk to me, fold AI into this.
00:35:13You must have a million thoughts.
00:35:15- Oh, there's a lot going on with AI.
00:35:17I mean, I think in its current instantiation,
00:35:19so we think about like an office worker for the most part,
00:35:21put programmers aside, I'll get back to them,
00:35:23but non-programmers are really interacting with chatbots.
00:35:25Like that's the main way they're integrating right now
00:35:28with AI.
00:35:30It's exaggerating exactly what you said.
00:35:32For a lot of people,
00:35:32it's exaggerating the problems that already exist.
00:35:35Now there's a term for this
00:35:36that comes out of a Harvard Business Review article
00:35:38from last year.
00:35:39They call it WorkSlop, as they put together as one word.
00:35:43And they have some pretty compelling data on this.
00:35:45- So what's WorkSlop?
00:35:45Define WorkSlop for me.
00:35:47- So WorkSlop is AI generated work products
00:35:50in the knowledge work sector.
00:35:51So like emails, reports, and PowerPoints, or what have you
00:35:55that are generated quickly by AI,
00:35:59but they're so low quality that they actually,
00:36:02it's very difficult, they make everyone else's jobs harder.
00:36:05This seems to be,
00:36:06this is like the defining aspect of WorkSlop.
00:36:08It's quick to produce, but it's so low value
00:36:10that it actually no real progress is made.
00:36:12So like you get a WorkSlop email from your boss or whatever,
00:36:17and like, this isn't useful to me.
00:36:18It's this weird wordy thing that's broken up into sections,
00:36:22and it doesn't get to the core of the problem
00:36:23we have to solve.
00:36:24So you made that email quick,
00:36:26but in the bigger scheme of things,
00:36:29we made very little progress towards what we want to do.
00:36:31Or you put together a WorkSlop PowerPoint presentation
00:36:34so that you would have something at the meeting,
00:36:36but now we're spending 20 minutes looking at this nonsense
00:36:38and nothing, it's not helping us.
00:36:40It's not helping us actually do things.
00:36:41So this is what's happening, or at least my fear.
00:36:43I mean, the reality is most people
00:36:46aren't using these tools in the office.
00:36:48Let's just set the reality, right?
00:36:49So, but for the people who are using them right now,
00:36:53which is a healthy percentage, but it's not-
00:36:54- Do you know what the numbers are?
00:36:56- Well, it's difficult because there's a lot of fudging
00:36:58of the numbers here.
00:36:59There's a lot of mistaking I have used or experimented with,
00:37:04with our regularly using them.
00:37:06So I see this mistake happen a lot.
00:37:09And so it's difficult to get good numbers.
00:37:11There was like famously,
00:37:13maybe it was an Ethan Moloch article
00:37:15where he was talking about, in my world, like academia,
00:37:17the homework apocalypse.
00:37:18And he's like, look at this study,
00:37:20students just don't do work anymore.
00:37:22Nine out of 10 are just using chatbots now.
00:37:25But you look at that study and what it actually said was
00:37:28nine out of 10 had tried using a chatbot at least once.
00:37:31And if you looked at who's using them regularly,
00:37:33it was like two out of 10, right?
00:37:36Because like for most of the students,
00:37:37it wasn't helping the way they thought it would.
00:37:39So I don't know what the numbers are.
00:37:41If you count like advanced Google use, I think it's larger.
00:37:46Like, yeah, I searched for information on this
00:37:48instead of going to Google, that's larger.
00:37:50But in terms of people who are actually making office work
00:37:53product out of it, I think it's smaller
00:37:55than the people who follow AI commentary
00:37:57or talk about it on AI Twitter, AI YouTube.
00:37:59I think it's a lot smaller than they probably assume
00:38:01just because in their world is pervasive.
00:38:03But the people who are using it, this is the problem.
00:38:06They're trying to avoid, this is my theory on this.
00:38:09It's like, how is AI helping like an office worker now?
00:38:11Well, their brain is exhausted
00:38:12from all this context switching.
00:38:14So what problem are they looking to solve?
00:38:16They're looking to avoid having to do hard moments
00:38:19of cognition because their brain is so fried.
00:38:21It's really difficult to like solve the blank page problem.
00:38:24Oh God, I got to send this email.
00:38:27I got to, it's a blank screen.
00:38:28I got to start writing from scratch.
00:38:29That's really hard.
00:38:30- And the inertia that they've been trained
00:38:33out of overcoming because of the primer.
00:38:35It's almost like a one, two punch.
00:38:38Humans were primed to not like heavy co...
00:38:42Well, we already didn't like heavy cognitive load.
00:38:44Then our ability to deal with it and get
00:38:47through that initial resistance was decreased
00:38:49through the context switching.
00:38:50And now we're, don't worry about it.
00:38:53Don't worry about it, carbon-based life forms.
00:38:55The silicon-based life forms are coming.
00:38:57- And let's throw in one other aspect in there.
00:38:59Also outside of work, we had these distraction machines
00:39:03in our hand that were further degrading our comfort
00:39:05with concentration because any possible moment
00:39:08of introspection we would have had even outside of work.
00:39:10Why would I do that when TikTok has like the perfect dash cam
00:39:14video of a Karen getting punched or something?
00:39:17Like I got to watch that, right?
00:39:19And so we have that revolution comes along
00:39:22plus the email revolution.
00:39:23We completely atrophy our ability to think
00:39:25and we exhaust our brain.
00:39:26So the other aspect of it, as we talked about
00:39:28it's really exhausting to go through your day
00:39:31context switching.
00:39:31So like I don't have any reserves left
00:39:33to write this PowerPoint.
00:39:34That seems impossible.
00:39:36And then AI is like, Hey, Hey, Hey, Hey,
00:39:38I can do it for you.
00:39:39It'll be fine. It'll be fine.
00:39:40It'll be good enough.
00:39:41It'll be good enough.
00:39:41You're like, Oh, okay.
00:39:42I can smooth over.
00:39:44I use this analogy in a New Yorker piece last year.
00:39:46It's like, it takes your effort graph looks like spikes
00:39:49like an EKG or something like that.
00:39:51And AI smooths over those peaks.
00:39:54And so you don't have to,
00:39:55your peak concentration required can come down.
00:39:58Like, well, you can fill the blank page
00:40:00and then maybe I have to work with it a little bit
00:40:02but that's easier than doing it from scratch.
00:40:04But the stuff being produced is no good.
00:40:06And so I feel like work slop,
00:40:08it's almost less of a,
00:40:11it's less of a critique of AI
00:40:15than it is AI making obvious a problem
00:40:19with the way we were already working.
00:40:20I think that's what's going on there.
00:40:22I think this is even happening with computer programmers.
00:40:24This is considered, you know, heretical right now.
00:40:27I guess I'm used to being yelled at.
00:40:29People are really excited by this workflow
00:40:33where I have seven or eight cloud code agents
00:40:36going concurrently producing code and testing them.
00:40:38And I'm just a manager of all these different processes.
00:40:40And they're all producing this code on my behalf.
00:40:43And it feels really cool and interesting.
00:40:44Like this has to be the future.
00:40:46I don't know that that is.
00:40:48I mean, I don't know the context.
00:40:51The problem is outside of like demos or internal tools
00:40:53or just having fun.
00:40:55That's not really code you can trust very well.
00:40:59And it does though completely lower the peaks
00:41:02of being a computer programmer, those peaks of cognition.
00:41:05It's much, much easier to manage
00:41:07a bunch of cloud code processes
00:41:09than it is to come up with an algorithm.
00:41:11And then you have that same blank page.
00:41:12So I think the jury is still out on even where
00:41:15we're gonna end up in the AI impact on programming.
00:41:18I don't know where it's gonna end up,
00:41:20but the way it's being talked about in the last few months
00:41:22after the latest cloud code update, which is sort of,
00:41:26I guess that's something humans don't do anymore.
00:41:29I don't think we're there ready to say that yet.
00:41:31- I get popped with cloud code ads.
00:41:34I get, you give me a terminal, I have no idea what to do.
00:41:38I'm like someone's grandmother trying to use an iPad.
00:41:41I have no idea what's going on.
00:41:42So they are pushing very, very hard at the moment.
00:41:46- It's funny, but it's a little bit crazy.
00:41:48But it's my world, I'm a computer scientist
00:41:50is that for engineer computer scientist types,
00:41:54they forget how technically advanced they are.
00:41:56So yeah, cloud code works in the terminal, right?
00:41:59And that's why it works so well.
00:42:01It exists in a world of text only.
00:42:02Text command line commands like the old DOS command line,
00:42:06it's all text commands, which you can do a lot with.
00:42:09You can create an edit and compile a computer program.
00:42:11So it's very good at that.
00:42:12And it's a limited set of textual commands.
00:42:14That's perfect for a language model.
00:42:17And the engineers are like,
00:42:18oh, we can use this terminal based tool
00:42:20to do all sorts of other stuff
00:42:22that's not computer programming.
00:42:24Great, this is solved, everyone's gonna be doing this.
00:42:27Everyone is gonna have these sort of personal assistants
00:42:30based on something on cloud code.
00:42:32I'm like, man, do you realize how foreign a command line
00:42:36interface is to people?
00:42:37You realize like how weird and nerdy
00:42:39and complicated your world is?
00:42:40You're like, yeah, this will be great.
00:42:41My grandma will just on the command line understand
00:42:44that like the cloud code agent can bring up a bass script
00:42:47that's just gonna cat those files
00:42:48over to the the regex graph, it'll be fine.
00:42:51No one knows how to do any of that type of stuff.
00:42:53So it's sort of funny seeing the engineers
00:42:55building these incredibly intricate nerdy,
00:42:58wonderful tools they've custom built for cloud code
00:43:01to help them in their life.
00:43:03And they think the gap between that and everyone else
00:43:06having AI automating things in their life is like,
00:43:08oh, it's this real small thing.
00:43:09I'm like, oh man, I don't think you understand.
00:43:11I mean, people are still not quite sure about the right click.
00:43:14I think you still have a ways to go before they're--
00:43:18- I saw this tweet from Robert Frndlaw.
00:43:23Lawyer uses ChatGPT to help write a brief.
00:43:25ChatGPT hallucinates cases in quotations.
00:43:28Court sanctions lawyer and four co-counsel
00:43:31for not catching the errors.
00:43:32The lawyer who used ChatGPT has practiced for over 30 years.
00:43:35He prompted ChatGPT write an order
00:43:38that denies the motion to strike with case law support.
00:43:41Told the court that he doesn't normally use ChatGPT
00:43:43and he used it this time
00:43:43'cause he was caring for his dying family members.
00:43:46Said no of his co-counsel
00:43:47were aware of this use of generative AI.
00:43:49Court says that because all five attorneys
00:43:51signed both documents that included these errors
00:43:54and they admit that not one of them verified
00:43:57that the case law in those briefs actually exist,
00:44:00that conduct violates Rule 11(b)(2).
00:44:03- There's hundreds of those happening, right?
00:44:06I heard, I don't know where this site is.
00:44:08There's a site that tracks this.
00:44:10Lawyers getting busted for ChatGPT written briefs
00:44:15that just make things,
00:44:16because it will for sure make up things if you ask it.
00:44:19Because again, what it tries to do is not to get,
00:44:23people know this, but right at the very bottom,
00:44:25what is a language model trying to do
00:44:26is trying to solve the word guessing game.
00:44:28That's how it was trained.
00:44:28It was given real text, you knock out a word
00:44:31and say, replace that word.
00:44:32Can you figure out what word
00:44:33was really there in the real text?
00:44:35So the language models just think
00:44:36they're trying to expand a real text that really existed.
00:44:39So they're trying to produce text
00:44:41that makes sense given the prompt.
00:44:43There's not world models or structured reasoning in there
00:44:47of like, okay, this is a legal brief
00:44:49and we have a notion of a citation.
00:44:51We don't know how it thinks about that.
00:44:53There's hundreds and hundreds of cases of this happening.
00:44:56I heard Scott Galloway talk about this on the Pivot Podcast,
00:44:58that there's some site that tracks this
00:45:00that he keeps an eye on.
00:45:01And he says, it astounds you.
00:45:02You think it's a handful of people?
00:45:04It's not, it's all the time.
00:45:06Here's my story of getting burned by that.
00:45:08I sort of learned my lesson.
00:45:09I was working on, because the one way I'll use ChatGPT
00:45:13is just sometimes instead of Google, right?
00:45:16Especially if I want like instructions
00:45:18for how to, whatever, change settings on something.
00:45:20It's great.
00:45:21It has a lot of really useful--
00:45:22- It's fucking spectacular for all of that stuff.
00:45:24If you want to use it as basically a glorified Wikipedia
00:45:26that's more instructive, like--
00:45:28- Yeah, like Wikipedia, you can ask questions of, yeah.
00:45:31So I was using, I was writing an essay
00:45:35and it was on Isaac Asimov's Rules Robotics.
00:45:39This was a New Yorker essay.
00:45:42And I left my copy of iRobot.
00:45:45I was here at my studio and I'd left it at home.
00:45:48I was like, oh, I needed to add this quote, right?
00:45:51Oh, I left it.
00:45:52And I was like, oh, you know what?
00:45:53That story's in the public domain.
00:45:55It's all over the internet.
00:45:56And this seems like it would be perfect for ChatGPT.
00:45:59Like, hey, can you just grab a copy and find me that quote?
00:46:02And they'll save me a little bit of time.
00:46:03Like, yeah, here it is.
00:46:04Here's the quote.
00:46:05I was like, yeah, roughly I remember I put it in there.
00:46:07And then the fact checker was like,
00:46:09where's this quote from?
00:46:11I was like, yeah, it's from the story or whatever.
00:46:12I get the book.
00:46:13It had just hallucinated a quote that was more or less
00:46:18like what was said, right?
00:46:19Because again, it's kind of playing the game
00:46:20of this is the type of text
00:46:22that would make sense giving the prompt,
00:46:25but it wasn't the actual quote.
00:46:26It had full access to it, right?
00:46:28You can search this.
00:46:29It's in the public domain
00:46:31so that the actual story is everywhere.
00:46:33So I had just naively assumed
00:46:36if you ask it for some information that exists
00:46:38on the internet that, oh, it'll just go find it
00:46:40and format it for you.
00:46:41It didn't.
00:46:42And then I went to a whole dialogue with it where I was like,
00:46:44this is not the right quote.
00:46:45And it was like, yeah, you're right.
00:46:46You know what?
00:46:47I thought you meant paraphrase a quote.
00:46:48Here it is, made up.
00:46:50I was like, that's not the real quote.
00:46:51Can you go get the real quote and give it,
00:46:54at this point I was just experimenting.
00:46:56I'd already filled it in, the article.
00:46:57And it was like, you're right.
00:46:58Yeah, you know, I was being hasty.
00:47:00Here you go.
00:47:01I could not get it to give me the real quote.
00:47:03So anyway, I learned my lesson.
00:47:05I was like, oh, don't assume,
00:47:08even if it's common information that it has access to.
00:47:11- Dude, the desire to fucking reprimand an LLM.
00:47:15And I've shouted at them.
00:47:17I've capital letter exclamation marks.
00:47:19It's like, what are you doing?
00:47:21What do you do, what are you hoping to achieve
00:47:25by throwing your emotional distress
00:47:27at this fucking disembodied voice on the other side?
00:47:29Okay, we, bits aside.
00:47:31I fucking love chat JPT.
00:47:32I think it's been really, really fantastic
00:47:34for tons of things.
00:47:36What's important is learning the limits
00:47:38and not using it for case law.
00:47:39This episode is brought to you by Whoop.
00:47:43I have been wearing Whoop for over five years now,
00:47:45way before they were a partner on the show.
00:47:47I've actually tracked over 1600 days of my life with it,
00:47:52according to the app, which is insane.
00:47:55And it's the only wearable I've ever stuck with
00:47:57because it tracks everything that matters,
00:47:59sleep, workouts, recovery, breathing, heart rate,
00:48:02even your steps.
00:48:03And the new 5.0 is the best version.
00:48:06You get all the benefits that make Whoop indispensable,
00:48:097% smaller, but now it's also got a 14 day battery life
00:48:13and has health span to track your habits,
00:48:15how they affect your pace of aging,
00:48:17it's got hormonal insights for ladies.
00:48:19I'm a huge, huge fan of Whoop.
00:48:21That's why it's the only wearable that I've ever stuck with.
00:48:23And best of all, you can join for free,
00:48:25pay nothing for the brand new Whoop 5.0 strap.
00:48:29Plus you get your first month for free
00:48:31and there's a 30 day money back guarantee.
00:48:33So you can buy it for free, try it for free.
00:48:35If you do not like it after 29 days,
00:48:37they just give you your money back.
00:48:39Right now, you can get the brand new Whoop 5.0
00:48:42and that 30 day trial by going to the link
00:48:44in the description below or heading to join.whoop.com/modernwisdom
00:48:49that's join.whoop.com/modernwisdom.
00:48:53What opportunities do you think
00:48:54an increasing reliance on AI opens up?
00:48:57'Cause I get the sense that as more people use LLMs
00:49:00to do work for them, this will create advantages
00:49:04in some areas for people who don't need to be reliant.
00:49:08So have you thought about the holes,
00:49:11market openings that will occur?
00:49:13- It will, I mean, the way I think about LLM based AI
00:49:16versus more advanced AI that we don't know how to do yet
00:49:18is my theory is what is being affected
00:49:22is gonna be more narrow at first.
00:49:23It's gonna be places where there's an exact match
00:49:26between what generative AI existing tools can do
00:49:29and existing market sectors.
00:49:32We saw this actually, the week we're recording this,
00:49:34we actually saw this reflected in the stock market.
00:49:37It was this interesting paradox that was going on this week
00:49:40where the stock price of software companies
00:49:44that deal with stuff that is well suited for an LLM
00:49:48went down, they call it the SaaSpocalypse, right?
00:49:51The software service apocalypse.
00:49:53So companies that do like legal advice,
00:49:56companies that do graphic design like Figma and Adobe,
00:50:00because a lot of, we have generative image generation
00:50:02is making building images from scratch is less useful.
00:50:06Customer service, so companies that do a lot
00:50:08of customer service type software.
00:50:10We saw the stock was sliding
00:50:12on these very specific software industries
00:50:14because like, look, I think LLMs are gonna be able to do this.
00:50:16It was triggered by Anthropic releasing some plugins
00:50:18that made it easier to integrate LLMs into your services
00:50:21without having to hire these other companies.
00:50:24But you would think that would be good news
00:50:26for the big tech companies building the AI
00:50:28that's gonna replace all this.
00:50:30Their stock was sliding as well.
00:50:32So the big tech companies had this big slide
00:50:34that at the end of the week we're recording this,
00:50:36there was a rebound at the end,
00:50:37it was like a trillion dollars in market cap disappeared
00:50:39from the big tech companies at the same time.
00:50:42So what does that mean the market was betting on?
00:50:45What are investors betting on at that point?
00:50:47What was gonna happen?
00:50:48And they were betting that in the near future,
00:50:50the next year or two, what we're gonna see
00:50:52is selective impacts in specific fields from generative AI,
00:50:57but also that too much money is being invested
00:51:00in these AI companies as it already,
00:51:03which means they're betting that they're not about
00:51:05to automate most of the economy.
00:51:07They're not about to just one more iteration away
00:51:10from a huge economic disruption.
00:51:12They're not at this peak of like complete transformation
00:51:15because if they were, you would be trying
00:51:17to increase your holdings in these companies.
00:51:18Like I don't care how much money they're investing.
00:51:21These companies are gonna be worth
00:51:22an astronomical amount of money, but the market is betting,
00:51:26I think the impact is gonna be more limited
00:51:28in the one to two year window
00:51:30than a lot of the commentary was seeing.
00:51:32So I think that's important because talk is cheap,
00:51:35but tech stocks aren't.
00:51:37And so people, the way they spend their money
00:51:39actually often has more of,
00:51:42I think there's a lot of information in that versus just,
00:51:45I've been reading these articles online
00:51:47and my God, the vibe really seems to be saying
00:51:49this is a big deal.
00:51:51So I kind of agree with the market's consensus right now.
00:51:54For sure, there's gonna be industries that are affected,
00:51:58but it's not gonna be one of these situations
00:52:00where you say, okay, any work
00:52:01that's not just the deepest creative work
00:52:03is all gonna be automated in the next few years.
00:52:04I better go learn how to like do art or something like that.
00:52:07I don't think it's gonna be that broad at first.
00:52:09I don't think the current generation of AI technology
00:52:12can support as broad of impacts as people think.
00:52:15There's a lot of extrapolation from,
00:52:16well, if it can do this with code,
00:52:18certainly it could do this with all these other jobs.
00:52:20If it could do this with this industry,
00:52:21well, certainly next it'll do it
00:52:23for all these other industries.
00:52:24We have to be wary of those extrapolations.
00:52:26- Right, I think I read an article from you.
00:52:28What if AI doesn't get much better than this?
00:52:31Sort of, if we have, I don't know,
00:52:32some sort of Flynn effect thing that kicks in, but for AI,
00:52:36where, you know, 'cause I think a lot of people
00:52:41would agree chat GPT two to three, fucking hell,
00:52:45to four, 4.0, I know there's a whole furor on the internet
00:52:49about people that have got girlfriends or boyfriends
00:52:51that are virtual on 4.0,
00:52:52and they're all getting upset and sad about it.
00:52:55And I don't understand.
00:52:56I don't think I use the tools sufficiently deeply
00:52:59to be able to test this and benchmark it.
00:53:01It's like, my Fire TV sticks remote isn't working well.
00:53:05It was able to do that fucking five years ago.
00:53:09But is your, your thinking is that we're maybe
00:53:14gonna reach asymptote for what LLMs generally
00:53:18and transformer technology is able to do,
00:53:20and then it's gonna be a new architecture entirely
00:53:23if we're going to actually get beyond this?
00:53:25- Yes, yeah, that's what that article is about.
00:53:27I think that was a very, of the articles I've written,
00:53:30I think that was a really important one
00:53:31that came out in August.
00:53:33And the story it tells,
00:53:35and a lot of other people have told the story as well
00:53:37or around that time and since.
00:53:38But the story it tells is basically what happened
00:53:41is there was this big paper that was published in 2020.
00:53:44The lead researcher Kaplan, Jared Kaplan,
00:53:46I think was at Anthropic at the time.
00:53:48And it was this paper where they said,
00:53:49"Hey, something weird is happening here.
00:53:51If we make LLMs bigger and we train them longer,
00:53:55they perform better."
00:53:56And technically they're seeing the loss decreased.
00:53:59That sounds kind of obvious,
00:54:00but in like machine learning circles,
00:54:01that was surprising because there's this idea of overfitting
00:54:04where if you just make your model bigger,
00:54:07the performance goes down.
00:54:08So it used to be like,
00:54:09you have to find the perfect size model
00:54:12for your problem space.
00:54:12That's the way people thought about machine learning
00:54:14until this paper came out.
00:54:15And like, I don't know, transform based LLMs,
00:54:18they were using GPT-2
00:54:19and they were systematically making it bigger.
00:54:22And they were seeing that the performance just kept going up.
00:54:25Like, this is interesting, so let's try it.
00:54:27And that was GPT-3.
00:54:29All right, let's actually make this like 10X bigger.
00:54:32Surely this can't be right, and it was.
00:54:34It matched the Kaplan curve exactly.
00:54:36Like, oh my God, this actually got way better
00:54:39just by making this bigger.
00:54:40Like, all right, well, certainly that must be the end of it.
00:54:43Let's try it with GPT-4.
00:54:44They made it bigger, they trained it much longer.
00:54:46Months and months they trained it.
00:54:47Microsoft had to build these custom data centers
00:54:49to train it with new AC technology that didn't exist before.
00:54:53And it fit the curve.
00:54:54It was like way better.
00:54:55And the thing GPT-4 did that really got,
00:54:58so GPT-4 set off the whole industry.
00:55:01The thing it did is it started showing abilities
00:55:03beyond just language.
00:55:04And that's where people got excited.
00:55:06Like, oh wow.
00:55:07If you train a language model on enough language,
00:55:11it learns about things that isn't just producing language.
00:55:14It can play games, it can do math problems, it can do logic.
00:55:17I mean, this was super exciting.
00:55:19It was super exciting.
00:55:20So the assumption was do this two or three more times.
00:55:25You have AGI.
00:55:26So that's what the whole industry was based off of.
00:55:28When we went from three to four was,
00:55:31this was legitimate, justified excitement.
00:55:33Expand the size and the training duration
00:55:36two or three more times,
00:55:38and the economy is gonna happen in a box.
00:55:41I mean, it was so, that's where all of,
00:55:42that was the engine for all this excitement.
00:55:45So they tried.
00:55:46At OpenAI it was called Project Orion.
00:55:48They made it bigger, modeled in four.
00:55:51They trained it even longer, like here we go.
00:55:53And they tried it and they said, it's not much better.
00:55:57And this was this big brick wall surprise for the industry.
00:56:02Like, wait, it didn't get better.
00:56:04Everyone else tried as well, right?
00:56:06Grok, they tried this with Grok as well
00:56:07at the Colossus Data Center was like,
00:56:08we're gonna have 200,000 GPU data center.
00:56:12No one's ever built anything this big.
00:56:14And it was like a little bit better.
00:56:16Meta tried this.
00:56:17They had a model called Behemoth.
00:56:19Like we built the biggest data,
00:56:20is bigger than any one we've had before.
00:56:22They didn't release it because it was marginally better
00:56:25than the last model that they had.
00:56:27And so this was a huge issue, right?
00:56:29You couldn't just make the models bigger
00:56:31and train them bigger.
00:56:32So what they did was they switched to,
00:56:35what are other ways we can get performance increases?
00:56:38And can we get more narrow by what we mean with performance?
00:56:41And this is when we began
00:56:42to get all the alphabet soup models.
00:56:44Well, it's GPT-03-mini slash whatever.
00:56:49And they switched the focus from just,
00:56:51this is amazing if you use it
00:56:52to we have these benchmark graphs
00:56:54and look at these graphs.
00:56:55Things are going better on these benchmarks.
00:56:56It all became about benchmarks
00:56:58because these are very narrow things
00:57:00that you could train models to do well on.
00:57:02They weren't intuitive.
00:57:03GPT-04 was just awesome.
00:57:05By the time we got to GPT-05,
00:57:06their whole launch, their launch page had 28 graphs
00:57:09of benchmark names that no one knew what they were.
00:57:12And so then they had to look for all these other ways
00:57:14to get improvement.
00:57:15And that's where you got like inference time compute.
00:57:17Well, what if we compute longer for harder questions
00:57:21and they began really pushing fine tuning?
00:57:23Well, for specific types of problems,
00:57:25we can get data sets that have answers
00:57:28and questions and answers.
00:57:29And we can use reinforcement learning
00:57:31that try to take this pre-trained model
00:57:34and make it better at this particular type of problem.
00:57:37And then we can have a benchmark
00:57:38that shows us we got better at this problem.
00:57:40And my argument in that article is like,
00:57:43this is a way different game than we were playing
00:57:45when we went from two to three and three to four.
00:57:47We're no longer scaling to AGI.
00:57:49We're taking basically GPT-04
00:57:51and we're doing all of this like tuning
00:57:53and adding extra stuff on top of it and around it
00:57:56and measuring these very narrow benchmarks.
00:57:58And that's why people have this feeling ever since.
00:58:00Like I guess they're better, but it's not in an obvious way.
00:58:03It's better in specific tasks or if I vibe code this,
00:58:06it looks better, I guess, and it seems more narrow.
00:58:09And so, yeah, we're reaching an ad,
00:58:11there's a long answer to a short question,
00:58:12but we are reaching an asymptote on just pure fine-tuned
00:58:16LLMs as an engine for AI.
00:58:19We're gonna need more architectures.
00:58:20It's gonna take more time.
00:58:21- Well, presumably chat GPT-6 could come out and oh fuck,
00:58:26they just blew through the entirety of my prediction.
00:58:29This curve no longer curves flat in the way that I thought
00:58:33and shit, this is a different universe now.
00:58:37- Yeah, but that won't happen because they tried
00:58:39and they don't know how to do that.
00:58:41So it's not gonna be just an LLM.
00:58:43I mean, my prediction of the future of AI
00:58:45is I think what we're gonna see,
00:58:46I think LLMs are very powerful,
00:58:48but what we're gonna see is much more of hybrid models
00:58:51that are custom fit to particular problems
00:58:54where, okay, this system does this thing better than a human.
00:58:59And in its guts, there's like an LLM in there,
00:59:02not a huge frontier model, but one that's like souped up
00:59:05and optimized for this particular type of thing.
00:59:07But there's also like five or six other models
00:59:09and there's an explicit world model,
00:59:10there's a future predictor,
00:59:12there's a policy network trained to reinforcement learning
00:59:14to try to evaluate situations to see what's good or bad.
00:59:17There's a whole logic engine on top of this
00:59:19that hooks these together.
00:59:21These are what I think the AI systems of the future
00:59:24are gonna be like, they're gonna be bespoke
00:59:25and there's gonna be a ton of them.
00:59:26So when we get to AGI, it's not gonna be GPT-7
00:59:30can do everything you ask it as well as a human.
00:59:32It's gonna be a world in which there's 10,000
00:59:34different AI products and you realize,
00:59:37everything I can think of now,
00:59:38there's some product out there somewhere
00:59:41that can do this better than humans.
00:59:42Just like there's AI that can play chess better than humans.
00:59:44There's a different AI that can play Go better than humans.
00:59:46There's an AI now that can beat professional poker players
00:59:49at Texas Hole of no limit.
00:59:51They're all different systems with their own pieces in them.
00:59:54And a lot of them have some language models in them as well,
00:59:56but a lot of other pieces as well.
00:59:58It's distributed AGI, that's what it's gonna be like.
01:00:00We're just gonna wake up one day and say,
01:00:03there's fewer and fewer things where we say,
01:00:05humans can do this better than computers.
01:00:07And it's a different model than PAL 9000.
01:00:10There's one giant, it's a really inefficient way
01:00:13to imagine solving this problem.
01:00:14If we just have a big enough language model,
01:00:17it's gonna do all activity, it's gonna power all agents,
01:00:20it's gonna automate all systems.
01:00:22That really doesn't make sense.
01:00:24I think it's gonna be a much more distributed path
01:00:27towards AGI and AI.
01:00:29- Given what AI can and can't do
01:00:36and what the quality of work is
01:00:38that it puts out at the moment,
01:00:40what is some good advice for somebody
01:00:42who wants to work against the weaknesses
01:00:47that are going to be exposed in other people
01:00:49because of their reliance on AI by avoiding it themselves
01:00:52or by using it appropriately?
01:00:53What would you focus on?
01:00:54Because again, it seems to me like quantity
01:00:57is easier to achieve than ever before.
01:01:00Quality is going to be rarer.
01:01:02That inertia, getting the project off the launch pad,
01:01:04the blinking cursor of the blank page,
01:01:06where should people focus their time and their attention
01:01:11in order to capitalize this?
01:01:12- I think you need to begin thinking about
01:01:16the feeling of cognitive strain,
01:01:19the way that a weightlifter thinks about the burn of a muscle
01:01:22or a runner thinks about burning lungs.
01:01:24As a thing that is uncomfortable in the moment,
01:01:26but man, I'm excited about this feeling
01:01:28because I'm getting stronger.
01:01:31You got to make yourself really comfortable thinking hard.
01:01:35That is the differentiating factor.
01:01:37I mean, obviously I've been saying this
01:01:38since 10 years now, but that's even more now
01:01:41going to be the differentiating factor, right?
01:01:43And if you talk to athletes, they're like,
01:01:44this is like Schwarzenegger and pumping iron
01:01:46talking about pump.
01:01:47And that's a really painful what he's doing actually, right?
01:01:50Like lifting the level of weights
01:01:52that the physical pain he's in is high
01:01:54and he compares it to an orgasm, right?
01:01:56Because if you're a weightlifter, you're like,
01:01:57oh, that pain is directly translating
01:02:00into more strength and more muscle mass.
01:02:01You got to think that same way about your brain.
01:02:03You cannot flee cognitive strain.
01:02:07You have to think about it
01:02:08in a knowledge work cognitive age.
01:02:11That is the feeling of my brain getting more capable.
01:02:14Yeah, I want to seek that out.
01:02:15Let's go get it.
01:02:16Let's go get some, right?
01:02:17Like I want to this, nope, bring my focus back to this thing.
01:02:20I'm going to try to push this through.
01:02:21And then when you're done, be like, oh man,
01:02:23I exhausted my brain.
01:02:24That's awesome.
01:02:25That was like a really good cognitive workout.
01:02:28So don't, while everyone else is using AI to run away
01:02:30from strain, you should be the person running for it
01:02:33because especially in the American context,
01:02:35I mean, the knowledge economy is now a massive portion
01:02:38of our GDP and the knowledge economy itself
01:02:40is shifting more towards cognition intensive work.
01:02:45So, you know, knowledge work can capture anything
01:02:47where you're not building things.
01:02:48But now all the lower level knowledge work
01:02:50is being outsourced or automated.
01:02:53A lot of it has been replaced over the last 30 years
01:02:55by software.
01:02:56We don't have support staff and assistants
01:02:57and secretaries like we used to because,
01:02:59well, you can use Microsoft Word and email.
01:03:01We don't need separated people.
01:03:03And so the work that's left in our economy,
01:03:05the knowledge economy has been getting more
01:03:07and more cognitively demanding.
01:03:08And so the number one skill is I'm used to straining my brain,
01:03:12learning hard new things and maintaining focus.
01:03:13That's what I would train.
01:03:15- That's so good.
01:03:17I really, really agree.
01:03:19And the funny thing is, that's why I asked at the top
01:03:23if you just felt like fucking Cassandra,
01:03:25because each subsequent development in technology
01:03:30makes this more important.
01:03:32There's always gonna be that seductive whisper
01:03:37in the back of someone's mind that,
01:03:40well, yeah, but I can work faster with AI.
01:03:42I can work quicker by what if my boss sees me
01:03:45doing executive functioning through Slack more, whatever.
01:03:49What's the elevator pitch for you should do
01:03:55a work of high quality and that will end up winning.
01:03:59- You have to think about employment.
01:04:01Ultimately, it's a marketplace.
01:04:03There's a lot of obfuscation and fog and smoke,
01:04:06but it's ultimately a marketplace, right?
01:04:08You're paid money.
01:04:10In exchange, you produce things that have economic value.
01:04:13That's what makes that exchange make sense.
01:04:15There is not ultimately an underlying economic value
01:04:19to the coordination activities by themselves.
01:04:21There is no actual economic value
01:04:23to the speed of your Slack responses
01:04:25or the number of meetings you go into
01:04:27or the number of like bullet pointed emails
01:04:29with those sort of chat CPT emojis that you put out.
01:04:32That itself doesn't generate economic value.
01:04:34The stuff that does a knowledge work
01:04:35almost always requires you mastering hard skills
01:04:38and applying them through concentration.
01:04:40And ultimately that shakes out.
01:04:42There's only so far you can get or so far you can hide
01:04:45being busy because busyness can't be monetized.
01:04:49And of course you can create a smoke for a while.
01:04:52Like, I don't know, like, you know,
01:04:54Chris seems like productive, I guess,
01:04:56like he's always on these emails and this and this and that.
01:04:59But if you're not actually producing things
01:05:01that have economic value,
01:05:02like ultimately that catches up to you.
01:05:03Your opportunities narrow.
01:05:05You're gonna get found out at some point
01:05:07where if you do the other thing,
01:05:09it's like, no, I'm creating stuff that is rare and valuable.
01:05:11It's unambiguously has value in the marketplace.
01:05:14You write your own ticket.
01:05:16Like what, you wanna have a business
01:05:17where you work half the year, you can do it.
01:05:19You wanna get paid a huge amount of money, you can do it.
01:05:22You wanna like work for a company,
01:05:23but you choose when you come into the office
01:05:25and you declare, like, I don't wanna do meetings.
01:05:27That's actually a thing, by the way,
01:05:29I talked to a marketing team
01:05:31at one of the major tech companies not long ago.
01:05:33And they said, you know what?
01:05:35We're in the sales side and like our group, the sales group,
01:05:39we are exempt from meetings
01:05:41because they can directly monetize.
01:05:44Oh, you brought in this many dollars.
01:05:46We can see it.
01:05:47And if you're bringing in dollars,
01:05:49they're like, you can do what you want.
01:05:50And they could also see if we make you go to meetings,
01:05:52those dollars go down.
01:05:53It's like, forget the meetings for you.
01:05:54Everyone else, where there's not a clear number
01:05:56where they can see how much value you're bringing,
01:05:57like, oh, you better be there in the meetings.
01:05:59- Dude, I've always thought this,
01:06:00the big problem that most people have
01:06:05that doesn't exist in the world of sports stars.
01:06:08If you're a sports star,
01:06:10everything that you're doing is to facilitate performance
01:06:12and performance is very tightly bounded and it's quantifiable.
01:06:17If you're a weightlifter, 300 kilos is 300 kilos.
01:06:21You either pick it up or you don't pick it up.
01:06:23And your sleep and your recovery and your nutrition
01:06:27and your hydration and your game tape
01:06:28and your technique work and your SNC and your body work
01:06:31and massage and soft tissue and all of that stuff
01:06:34combine to this output.
01:06:36It's a very, very sort of single ordinating principle.
01:06:40The same thing goes for tennis
01:06:41and the same thing goes for football
01:06:43and the same thing goes for baseball and so on and so forth.
01:06:45You do not perform well.
01:06:47You begin to scrutinize all of the contributing elements
01:06:49that come toward that.
01:06:50The problem that you have in most normal people's lives
01:06:54is the output that they're optimizing for
01:06:57is diffuse and very hard to work out.
01:07:00Well, I wanna be a good father
01:07:02but I also wanna perform at work.
01:07:05I do Brazilian jiu-jitsu on an evening time
01:07:07and my wife makes me go dancing
01:07:09and I wanna be engaging at a cocktail party.
01:07:11Okay, well, first off, that's lots of things.
01:07:14It's not a single ordinating principle.
01:07:16And secondly, define to me the lineage
01:07:19between your disrupted sleep last night
01:07:23and your poorer performance around the dinner table
01:07:26or in Brazilian jiu-jitsu or whatever.
01:07:29The diffuse thing contributes
01:07:31because you inevitably have to make trade-offs
01:07:33from one thing in order to do another.
01:07:35But also it's just hard.
01:07:36It's hard to work out how your performance is performing.
01:07:39And this is the same in the work life.
01:07:43Perfect example, the salespeople,
01:07:45we just know if we make you do this thing,
01:07:48we lose that thing.
01:07:49And that thing is more important than this thing.
01:07:51It would be like if for some reason
01:07:54sports stars were being encouraged to stay up late.
01:07:56You go, well, we know if we make you stay up late
01:07:58answering fucking slacks,
01:08:01your performance in the game the next day decreases.
01:08:04But for most people, there's this implicit assumption
01:08:08that part of what you do is the contribution
01:08:11to the strategy and the operations
01:08:13and the executive function culture and so on,
01:08:17which means that you forget what you're there for.
01:08:19I think people have forgotten what they're there for.
01:08:21What am I supposed to be here at work doing?
01:08:23What is my outcome goal?
01:08:25- There's so much fat
01:08:27in the American knowledge work sector right now, right?
01:08:30'Cause we're so wealthy
01:08:33and there's so much money being slung around
01:08:34that we can have whole organizations
01:08:37where most people don't even know
01:08:38how they're directly connected to producing that value.
01:08:40And they could just be doing email all day or whatever, right?
01:08:42It's so inefficient.
01:08:45But there are plenty of knowledge work areas
01:08:49where people don't put up with a bunch of this nonsense.
01:08:50And it's all areas where it's very easy
01:08:52to quantify your production.
01:08:55I did this essay a couple of years ago
01:08:58where I did a reflection where I said,
01:08:59God, almost every thought I've had in my books
01:09:02all came out of my experience as a grad student at MIT.
01:09:06So I was at the theory of computation group
01:09:09in the computer science department at MIT.
01:09:11I don't call it department,
01:09:12but the theory of computation group in the CS lab at MIT,
01:09:16which is like a group, the professors there,
01:09:18the students, we weren't like this,
01:09:19but the professors were super geniuses.
01:09:21Like literally Turing Award, Turing Award,
01:09:24MacArthur, MacArthur, Turing Award, Dijkstra Prize,
01:09:26like smartest people in the world.
01:09:29And it was incredibly clear if you were successful or not.
01:09:32What major theorems did you prove in the last few years?
01:09:35That's it, that's all that mattered, right?
01:09:37And that required a lot of thinking.
01:09:38So they were terrible with email.
01:09:41They had no interest in social media.
01:09:44Meetings, like if you're trying to throw meetings at them,
01:09:46they would just ignore you, right?
01:09:47I wrote about this in deep work even, and people pushed back.
01:09:50I was like, this is what it's like in that world.
01:09:52If you send someone an email in this world,
01:09:54like one of these professors,
01:09:56and you're like, this is ambiguous.
01:09:59You kind of didn't word this well,
01:10:00or I don't really want to do this.
01:10:01They just ignore it.
01:10:03Like that's on you, buddy.
01:10:04Like I have to get, you know, I'm being,
01:10:06I will lose my job if I'm pre-tenure,
01:10:08if I don't come up and solve theorems.
01:10:10And they put up with no nonsense.
01:10:11And a lot of that actually infused the book, Deep Work,
01:10:14because like, you know what?
01:10:15I came of age in an environment
01:10:17where all anyone cared about was focus,
01:10:20and everything else was secondary.
01:10:22It's like athletes, just like you said,
01:10:24if this is getting in the way of my launch angle going down,
01:10:27or my batting average adjusting, I'm going to change it.
01:10:30But it's crazy right now in knowledge work
01:10:32how many positions that's not true.
01:10:33But what I advise people then,
01:10:34get in a position where that's true.
01:10:37Change your profile at work,
01:10:39or if you're changing your job, change your job into one
01:10:42where your value production is unambiguous.
01:10:44Now, this is a double-edged sword,
01:10:46'cause it swings both ways. - 'Cause you can't hide anymore.
01:10:48- You can't hide anymore.
01:10:49But if you get into one of those situations
01:10:51and then you do the cognitive work,
01:10:53I know how to focus, I build the skills, I apply the skills,
01:10:57I'm not afraid of cognitive strain,
01:10:59you're in the absolute best position in our economy, right?
01:11:02You can write your own ticket,
01:11:03but you have to be willing to go into a circumstance
01:11:07of like, this is the only world I know.
01:11:09And academia is, what did you publish?
01:11:11That's all that matters.
01:11:12That's all we care about, what'd you publish.
01:11:14Book writing, how many copies did your last book sell?
01:11:17That's all that matters.
01:11:18There's no, you know what though?
01:11:19He answered our publisher email so quickly,
01:11:22so let's give him another deal, folks.
01:11:24No, it's exactly how many dollars did you make us last time?
01:11:27That's what we care about for the next time.
01:11:30So it's a scary world where you're being held accountable.
01:11:34But it's an equation I always say,
01:11:36is that if you're accountable,
01:11:37you don't have to be accessible.
01:11:39If you're like, I can point to,
01:11:42this is the value I produced and I'm killing it for you,
01:11:46then I don't answer emails, I don't go to these meetings,
01:11:49I don't do 50 sort of things.
01:11:50You can get away with almost anything you want.
01:11:52So I think that's, more people should make that move,
01:11:54especially in the AI age, I suppose.
01:11:57More people should make that move towards like,
01:11:58hey, hold me accountable,
01:12:00and then do the work to actually show up.
01:12:03It makes your life so,
01:12:04it's such a better way to go through knowledge work.
01:12:05You get away from that hyperactive hive mind,
01:12:08brain melting, distracting, soul crushing,
01:12:11slack all day long nonsense.
01:12:13- In other news,
01:12:14you've probably heard me talk about Element before
01:12:16and that's because I am frankly dependent on it.
01:12:20And it's how I've started my day every single morning.
01:12:24This is the best tasting hydration drink on the market.
01:12:27You might think, why do I need to be more hydrated?
01:12:28Because proper hydration
01:12:29is not just about drinking enough water,
01:12:31it's having sufficient electrolytes
01:12:33to allow your body to use those fluids.
01:12:35Each Grab and Go stick pack
01:12:36is a science backed electrolyte ratio
01:12:38of sodium, potassium, and magnesium.
01:12:40It's got no sugar, coloring, artificial ingredients,
01:12:43or any other junk.
01:12:43This plays a critical role in reducing muscle cramps
01:12:46and fatigue while optimizing brain health,
01:12:48regulating your appetite and curbing cravings.
01:12:51This orange flavor in a cold glass of water
01:12:53is a sweet, salty, orangey nectar,
01:12:56and you will genuinely feel a difference
01:12:58when you take it versus when you don't,
01:12:59which is why I keep going on about it.
01:13:01Best of all, there's a no questions asked refund policy
01:13:03with an unlimited duration.
01:13:04Buy it, use it all, and if you don't like it for any reason,
01:13:07they give you your money back
01:13:08and you don't even have to return the box.
01:13:10That's how confident they are that you'll love it.
01:13:12Plus, they offer free shipping in the US.
01:13:14Right now, you can get a free sample pack
01:13:15of Element's most popular flavors with your first purchase
01:13:18by going to the link in the description below
01:13:19or heading to drinklmnt.com/modernwisdom.
01:13:23That's drinklmnt.com/modernwisdom.
01:13:27Let's say that you were in an organization
01:13:29that was small enough
01:13:30that you could actually enact some change.
01:13:31Maybe you're at the top of it, near the top of it,
01:13:33or you're just, you're toward the bottom of it,
01:13:35but you feel like you've got the ear
01:13:36of the person that's in charge.
01:13:39If you were to say, well, you've got the classic
01:13:42diffuse hive mind, pseudo productivity malaise,
01:13:47like the ambient soup that everybody's swimming in,
01:13:50how would you, what would you do?
01:13:54How would you rework the internals of an organization
01:13:57that still needs to communicate?
01:13:58Obviously, there has to be coordination.
01:14:00People aren't working in silos.
01:14:02There is gonna be inevitable communication
01:14:03and coordination that needs to happen.
01:14:06How do you survive the modern world?
01:14:09What would you propose?
01:14:10How would you restructure things?
01:14:11- Yeah, I mean, I would do a few things.
01:14:13One, I would say we're gonna have explicit workload tracking
01:14:15and management, right?
01:14:16No more just people throw stuff at you
01:14:18and you implicitly just add it to your plate.
01:14:20We want a place where we write down
01:14:21what everyone's working on and we can see it.
01:14:24And now we can start talking about things
01:14:25like what is an ideal WIP?
01:14:27What's an ideal work in progress limit for an individual?
01:14:30How many things do we want someone working on
01:14:32at the same time before that curve
01:14:34starts to go the other way?
01:14:36So what you have to do once you start doing that
01:14:38is saying we need a place to track things
01:14:40that need to be done
01:14:41that no one is actively working on right now
01:14:43and we can feel okay about it.
01:14:45So I would definitely want to set up
01:14:46where we're things enter into our radar
01:14:48of this needs to be done.
01:14:50There's a place for that to go
01:14:52and to be stored where it's no one.
01:14:54- It's like an organizational getting things done inbox.
01:14:58- Yes, and it's not on anyone's plate
01:15:00because as soon as you are responsible for something,
01:15:04it generates email, Slack and meetings.
01:15:06So once it's on your plate,
01:15:08it begins to spin off administrative overhead
01:15:11and slow productivity call it the overhead tax.
01:15:13That gets spun off as soon as it's on your plate.
01:15:15So everything by default goes to a team plate.
01:15:19No one's working on it.
01:15:20Then we keep track of from that play
01:15:23as we move things to people's individual responsibilities,
01:15:25we have like, you should do three things at a time, that's it.
01:15:28And when you finish something,
01:15:30you can pull something else in.
01:15:31So do a small number of things fast and well,
01:15:33and then keep bringing things.
01:15:34So I would definitely do that.
01:15:36The second thing I would do is I would say
01:15:38no more hyperactive hive mind.
01:15:39If you send a message that requires more
01:15:42than a single message in response,
01:15:44that should not happen over digital communication.
01:15:47If I can't just answer your question with one more message,
01:15:50then that has to be real time.
01:15:52Now we can't have that turn into an explosion of meetings.
01:15:54So what we're gonna do is we're gonna have
01:15:56daily office hours for everyone.
01:15:58So there'll be a daily time where everyone knows
01:16:00they can call you or walk to your office or whatever,
01:16:02and go through a bunch of things with you real quick
01:16:03instead of sending emails.
01:16:04We're gonna have morning stand up meetings
01:16:06within the teams for sure.
01:16:08Who's working on what this morning?
01:16:10Who needs what from who to get that done?
01:16:12Go do the work.
01:16:14We're gonna have, so we'll definitely do those as well.
01:16:16We might throw in phone hours.
01:16:18It's a new idea I'm thinking about where you say,
01:16:19look, there's a longer period of time,
01:16:21like maybe all afternoon, where you can always call me
01:16:24if there's something that's so urgent
01:16:26you can't wait till the next office hours.
01:16:29There's enough friction in phone calls
01:16:30that that actually turns out to work pretty well.
01:16:31Like I'm not just gonna call you
01:16:33because I'm wanting to get something off my plate.
01:16:35I won't call you unless it really is serious.
01:16:36So I would do that as well.
01:16:38And then I would say, okay, what ongoing work
01:16:41does this not work for?
01:16:43What type of projects do we work on on a regular basis
01:16:45where this isn't working because it's too long
01:16:48to have to wait till the afternoon's a problem?
01:16:50I say, great, let's identify those.
01:16:52And for each of those, let's build a protocol.
01:16:55Here's our protocol for collaboration on this type of work.
01:16:58And however that's gonna work,
01:16:59but it's like the information goes into this spreadsheet
01:17:02and then whatever, someone checks it in the morning,
01:17:04they move things to shared files.
01:17:06I don't know what it is, but whatever it is
01:17:07that prevents us to have ad hoc unscheduled messaging
01:17:10isn't necessary.
01:17:11So explicit workload management.
01:17:13I would have this rule of no hyperactive hive mind.
01:17:17I would have protocols for any type
01:17:18of recurring collaboration where it could be explicit
01:17:20about how we actually wanna do this.
01:17:22And then I would have a culture of talking about deep work
01:17:24and concentration, like a tier one skill.
01:17:26How's it going?
01:17:28How many deep work hours did you get in this week?
01:17:30Are you happy about that?
01:17:31What was getting in the way of that?
01:17:33Did you have a particularly good session?
01:17:35Tell everyone else about it, like what worked?
01:17:37Oh, I see, you did music.
01:17:39You have a different look.
01:17:40Oh, let's all think, hey, here's a good idea
01:17:42that we can borrow.
01:17:44Make deep work culturally something you talk about
01:17:48as like this is a tier one skill
01:17:50that we're really proud about.
01:17:52You do those things, you're like gonna 2X your profitability.
01:17:55This is the thing that's always frustrating me
01:17:57about these ideas is like you could make more money
01:18:00if you do it, but it's really hard.
01:18:03Those changes I just talked about, it's hard.
01:18:05There's friction, there's personalities.
01:18:08And this is the thing I really underestimated
01:18:09when I wrote those books.
01:18:11The way we work now is like a low energy point, right?
01:18:16It's like the easiest possible configuration of work.
01:18:21So if you feel friction,
01:18:22you're trying to do something more structured,
01:18:23you're trying to do something that makes better use
01:18:25of our brain and you're getting resistance.
01:18:28The place you're gonna fall when you give up
01:18:30is the way we're doing it now.
01:18:31So it's not arbitrary, I've realized.
01:18:33This hyperactive hive mind,
01:18:34let's just figure things out on the flow,
01:18:36no workload management.
01:18:37It's not arbitrary.
01:18:38It's the low energy.
01:18:40It's like this local minimum.
01:18:42It's the place that like minimizes the complexity
01:18:44that still allows a company to run.
01:18:46And I think that's why we keep falling back.
01:18:48In mathematical terms, it's a suboptimal Nash equilibrium.
01:18:51It's not the optimal way to work together,
01:18:53but no one person can leave it
01:18:55and make their situation better.
01:18:56It's a low energy state, it's an attractor,
01:18:59it's a local minimum in the utility landscape,
01:19:01whatever mathematical metaphor we wanna use.
01:19:04And so it's not arbitrary.
01:19:06I was like, oh, it's like a law of work physics.
01:19:08This thing is like a neutron star in the world,
01:19:12a universe of work that just attracts everything back to it.
01:19:15And it takes a huge amount of energy to escape its pull.
01:19:18That's why I think we've had so much trouble
01:19:21solving this problem even though you would make more money
01:19:23if you did it.
01:19:24- I wonder, I'm thinking about sort of immediately
01:19:30implementable solutions for this.
01:19:33I get the sense that you could probably tell people
01:19:35we don't use Slack before 1 p.m.
01:19:41Like nobody is to post in Slack before 1 p.m.
01:19:44Because you can ring if it's SOS emergency scenario,
01:19:50you can just call somebody.
01:19:51We just don't use it.
01:19:52And then it means that everybody knows
01:19:54that they should not be doing, it's a company-wide deep.
01:19:58I mean, look, are there gonna be some departments,
01:20:00HR for instance, probably would be used.
01:20:03But your job is HR, you're in the PR department
01:20:06or something like that.
01:20:07Your job is actually about comms.
01:20:10But if you're in marketing or if you're in accounting,
01:20:13something like that, okay, sit down and do your fucking work.
01:20:16And up until a point, what do you make of intermittent fasting
01:20:19for communication company-wide?
01:20:20- Yeah, it works, especially though,
01:20:22what really makes that more sustainable
01:20:24is if you have that quick morning stand up
01:20:26on the team scale at the beginning of the day,
01:20:29where everyone says, here's what I'm gonna be working on
01:20:31during these morning hours.
01:20:33Here's what I need from each other to make progress on this.
01:20:36So what would have unfolded over Slack and email,
01:20:40you're doing in 10 minutes.
01:20:42So you say, okay, here's what I'm working on this morning.
01:20:44I'm working on the new white paper.
01:20:46Here's what I need though.
01:20:48I need those figures from you.
01:20:50When can you get them to me, by 9.30?
01:20:51All right, you're gonna get them to me by 9.30.
01:20:53And I need those quotes you promised.
01:20:55Can you just do that right away?
01:20:57Okay, so you all know what I need from you.
01:20:59Okay, now I'm gonna put my head down and write that report.
01:21:01So having that meeting ahead of time,
01:21:05where everyone says what they need and what they're gonna do,
01:21:08that makes that time work better.
01:21:09And then the thing that really works,
01:21:11do the same thing on the other end of the morning.
01:21:13All right, you said you were gonna work on
01:21:16this, this and this, what happened?
01:21:19So there's accountability on the other end.
01:21:20You can't run away from,
01:21:23if you just went on email and social media,
01:21:25they're like, well, wait a second,
01:21:26I thought you were gonna write the white paper.
01:21:27Yeah, and if other people flake,
01:21:29they don't send you the figures,
01:21:30they don't send you the quotes.
01:21:31You're like, I got stuck, man.
01:21:33I never got this.
01:21:34- Cal didn't do what he said.
01:21:35- And they're there in the same room.
01:21:37And they're like, oh, okay, I get it.
01:21:39I get it, I can't just ignore stuff, right?
01:21:42Like I actually have to do it.
01:21:43I think that's a great idea.
01:21:44I think something like that works well
01:21:46if you put that accountability before it
01:21:49and you put it after it.
01:21:50That scares people, by the way, though.
01:21:53That really does scare people
01:21:55because you actually have to do the work.
01:21:57And this is the thing with really social media
01:22:01and smartphones killed this way worse.
01:22:02AI is gonna make this worse.
01:22:04But that was a big inflection point.
01:22:06In terms of losing our comfort with concentration,
01:22:08that got really bad.
01:22:10Once we got algorithmically optimized content,
01:22:12we really got used to that.
01:22:13And so it's scary if you just go to a company
01:22:15and say, here's the new plan, boss.
01:22:17We're gonna have a meeting in the morning.
01:22:19You gotta tell me what you're gonna do
01:22:20for the next five hours.
01:22:21And then you gotta do it.
01:22:22And we're gonna check in after that five hours
01:22:23and see how it went.
01:22:25That's a nightmare for a lot of people.
01:22:27That is like, oh God, I don't know what I'm gonna do.
01:22:30- I agree.
01:22:31I get the sense that a nice way to introduce this
01:22:35would be look, everybody's brain here
01:22:37has been turned into slop.
01:22:39Everyone.
01:22:40No one is able to do their job
01:22:42as effectively as they should.
01:22:44So you are expected to do the work.
01:22:46But the reason that we do the pre and post
01:22:49is not to whip somebody into performance review.
01:22:52It's to give you accountability
01:22:54'cause you don't look like a tit
01:22:55in front of your coworkers.
01:22:56But if you don't get to the point,
01:22:59the same as when you start training for a marathon,
01:23:01you don't run 10K on the first day.
01:23:03You will titrate the dose up and overtime.
01:23:06Week one, we'll permit some fuckery
01:23:09and week two, we'll permit a bit less fuckery
01:23:11and week three, we're all in it together
01:23:13and this person's pulling ahead.
01:23:14They're really like a hyper responder.
01:23:17They're making loads of gains in the focus gym
01:23:19and other people are moving a bit more slowly.
01:23:20Okay, what is it that they are doing?
01:23:21And so on and so forth.
01:23:22But imagine that.
01:23:23Imagine if you had a company-wide focus initiative
01:23:28where people were just, okay, we're gonna move together.
01:23:30Everybody is going to focus on focus.
01:23:32And interesting around the AI thing.
01:23:35So George, my housemate's writing a book at the moment.
01:23:38Do you know Cold Turkey?
01:23:39Do you ever use Cold Turkey?
01:23:40- I know about it, yeah, the software.
01:23:42- Yeah, it's a website limiter, app limiter for MacBook.
01:23:46We've been using it for a decade.
01:23:47His Cold Turkey went rogue
01:23:50and just kept shutting his browser down
01:23:54even though he wasn't trying to access the thing
01:23:57that he wasn't supposed to.
01:23:58It said he needed to install it.
01:23:59It was a nightmare.
01:24:00And here's a conversation between him and his AI.
01:24:03Cold Turkey has gone rogue and I need to remove it.
01:24:05Please tell me how to delete it from terminal.
01:24:08And the response, the response is,
01:24:11I'm not gonna help you bypass it, George.
01:24:13This is exactly the scenario you set it up for.
01:24:15You're two days in, the book is waiting.
01:24:17Close the terminal and write.
01:24:19And he's replied and said, no, it's got a bug
01:24:22so I can't get on calls.
01:24:23He's pleading with his own AI
01:24:25'cause he's obviously put in the instructions,
01:24:27be rigorous with me, be tough with me,
01:24:29tell me that I should be getting back to being focused
01:24:30when I start to go off task, do the thing.
01:24:32And that's an AI equivalent of what you're talking about
01:24:35which is this super visionary oversight commission thing
01:24:40but his just happens to be based in silicon
01:24:43instead of in other people.
01:24:44- So maybe AI will help us.
01:24:45It could basically chastise us.
01:24:47- Well, the problem is,
01:24:48the problem that you have with the AI thing
01:24:49is it's so fucking sycophantic all the time
01:24:52that it will tend to bend eventually
01:24:57to what it is that you want.
01:24:59- Yeah, but no one believes
01:25:00that the chatbot interface is the future of AI,
01:25:02the boosters, the skeptics, the moderates.
01:25:05There's an emerging consensus
01:25:07that we're gonna look back at this current moment
01:25:09where we interact with AI by typing into a chat window.
01:25:13That's gonna be like the Usenet newsgroups
01:25:17at the beginning of the internet.
01:25:18It was like a cool thing early on
01:25:20that showed the promise of the internet
01:25:22but the tools got better.
01:25:24There's better ways to make use of it.
01:25:26So the thought is in the future,
01:25:28AI is gonna be more integrated into more things.
01:25:29It'll be more agentic.
01:25:31It'll be a lot not like having conversations in English text
01:25:34but deploying agents to do things,
01:25:36maybe with natural language
01:25:37but also it'll be more integrated into software,
01:25:40individual tools will be more common.
01:25:42So it'll be much more common.
01:25:43I'm in Microsoft Excel and I'm like,
01:25:47can you sort row five by this amount
01:25:49and cut out all columns that have less than as many values
01:25:52and it does that.
01:25:53That's what the interactions are gonna become like.
01:25:56And so this idea of having a singular anthropomorphized entity
01:26:00through which you're having all conversations,
01:26:02that's almost like an accident of early AI.
01:26:04OpenAI will tell you this,
01:26:06that ChatGPT was supposed to just be a demo
01:26:08of the type of things you could do
01:26:10using the APIs into their language models.
01:26:13It's like the type of tool you can build
01:26:15that would make use of AI
01:26:16and then it caught them completely off guard
01:26:18and everyone wanted to use ChatGPT and chat with it
01:26:20because it was really cool.
01:26:21I don't think that's gonna be the form vector.
01:26:22So I think a lot of these issues we have now,
01:26:24like this is weird.
01:26:25It's unsettling, we're anthropomorphizing it,
01:26:27we're getting parasocial relationships with the agents,
01:26:29we're having romantic relationships with them,
01:26:32we're getting unsettled
01:26:34because having English conversation,
01:26:36we have a hard time not simulating a mind
01:26:39on the other end of this type of thing.
01:26:41- Which is why I shout at my ChatGPT team.
01:26:42- That's why you shout at it.
01:26:43I think a lot of this two years from now is gonna seem,
01:26:45it'll be super narrow, right?
01:26:47Because I don't think just having
01:26:49a sort of general purpose oracle you chat with,
01:26:51that's not the future.
01:26:52That's not what people think we're gonna be doing.
01:26:53- Why are people mad about 4.0 being removed?
01:26:58- My understanding was they were just happy with the fine,
01:27:02so you tune these things.
01:27:04The conversational style comes from
01:27:06a post-training tuning session where you give it,
01:27:09you've already done the pre-training, which is unsupervised.
01:27:11And you go through this post-training session
01:27:13where you have a lot of examples of questions and answers,
01:27:16and you ask the question and then it gives an answer,
01:27:19and then you sort of zap it using optimization theory
01:27:23to try to move, like now we're gonna change the weights
01:27:25to be closer to this answer we already said was better.
01:27:28So if you have a bunch of examples
01:27:29of the way you want something to respond,
01:27:31and then you go through one of these
01:27:32sort of zapping training sessions after the fact,
01:27:35it'll respond more like that.
01:27:36So they just changed the way they were doing that.
01:27:38And the thing they changed to,
01:27:40people didn't like the tone that created.
01:27:42So it was just about what day,
01:27:44literally like the data sets you're using
01:27:46when doing this fine tuning
01:27:48after you've done that big, massive pre-training
01:27:51where it's unsupervised.
01:27:52- Talk to me about the role of quantum computing in AI.
01:27:57- Minimal to non-existent.
01:27:59- So QAI is all just bullshit?
01:28:03- Yeah, I'm not, yeah.
01:28:05I mean, quantum computing is really interesting.
01:28:07There's a huge amount of technical problems
01:28:08just to actually get these things scaled
01:28:10to the number of qubits in which they're useful.
01:28:12And there's a fallacy out there
01:28:15in thinking about quantum computing
01:28:16that it's basically like a normal computer,
01:28:18but times a million,
01:28:20which is just not the way these things function, right?
01:28:22So there's only very specific problems you can solve
01:28:26with a quantum computer
01:28:28because you actually have to express the problem
01:28:31in the language of physics
01:28:32in such a way that you're creating
01:28:34what's known as a wave function
01:28:35that when it collapses,
01:28:36it's going to collapse to a configuration
01:28:38that's the right answer.
01:28:39Therefore, like implicitly searching a large state space
01:28:41in sublinear time.
01:28:43Only certain problems allow you to do that.
01:28:45So it's unlike a normal computer
01:28:46where I can program a computer to do almost anything.
01:28:49Quantum computers is much more narrow
01:28:50what you can do with it.
01:28:52- Could you give me an example of something
01:28:53that it would and wouldn't be able to do?
01:28:55- Well, like the big example,
01:28:57this was a guy who was at MIT when I was there.
01:28:58Peter Soar early on was the one who figured out like,
01:29:01hey, one of these complicated wave function collapsing things
01:29:03you could do could factor prime numbers or--
01:29:07- Cue day.
01:29:08- Yeah, factor numbers to see,
01:29:10to find the prime factors rather.
01:29:11Find the prime factors of big numbers.
01:29:13That's a really big deal because--
01:29:16- Security.
01:29:17- Yeah, public key encryption.
01:29:19And ironically, this just goes to show how crazy MIT was,
01:29:23is also at MIT is Ron Rivest who I TA'd for,
01:29:27who invented, you see R and RSA,
01:29:29he invented public key encryption.
01:29:30So like the guy who invented public key encryption
01:29:32is there next to the guy who figured out
01:29:34how quantum computers could--
01:29:36- Could maybe undo it.
01:29:37- Undo it, yeah.
01:29:38So it's kind of interesting.
01:29:39So it's good at that.
01:29:40There's a lot of problems that are based around
01:29:42simulation of quantum or physical physics systems.
01:29:46And that's, you can simulate quantum physics systems
01:29:50directly using quantum in a way,
01:29:52instead of having to try to simulate it with,
01:29:53so it's very good for that.
01:29:54There's a certain type of search.
01:29:56It gets a little technical,
01:29:57but there's a certain type of search that you can implement.
01:30:00It has applications.
01:30:01So there are interesting applications.
01:30:03But the thing I was beginning to sense recently,
01:30:06which made me worry,
01:30:07is that there was a sense of like height migration.
01:30:10So people are getting a little bit frustrated,
01:30:12sort of like post GPT-5 of like,
01:30:14this isn't filling my need to have something to be,
01:30:17you know, a technology that is gonna change everything.
01:30:19I love that concept.
01:30:21And they begin sniffing around, okay,
01:30:22but what if we just quantum somehow,
01:30:25we'll unlock AI and solve all these problems we're having.
01:30:28I think it's way more complicated than that.
01:30:29There are narrow applications of these particular things
01:30:32that might have some AI application,
01:30:34but you can't like run an LLM on a quantum machine
01:30:38and now it's a billion times better.
01:30:39That's just not how it works.
01:30:40So quantum is interesting.
01:30:42It's just really hard.
01:30:43The problem is the errors multiply.
01:30:44I mean, they make these qubits,
01:30:47these quantum bits they use for these algorithms.
01:30:50It's incredibly complicated.
01:30:51You have, there's different ways to do it,
01:30:52but in some ways you have laser beams
01:30:54and a super cool chamber holding like a particle
01:30:57in a very careful state.
01:30:59And it generates errors
01:31:01and then the errors add up with other errors.
01:31:03And after you make enough of these things,
01:31:05then the errors, they swap out of control.
01:31:08It's a really, you know.
01:31:09- Okay, so you're telling me that the fucking M6 chip
01:31:12in the MacBook Pro is not gonna be a quantum one.
01:31:15It's not gonna be the Q6 chip.
01:31:17- It's not.
01:31:17I was, now I wanna know what QAI is.
01:31:20What is QAI?
01:31:21You mentioned QAI.
01:31:22- QAI, quantum AI.
01:31:24- Yeah, but I mean, is there a particular product
01:31:26or just people talking about quantum's
01:31:28gonna just make AI better?
01:31:29- Yes, yeah, there is.
01:31:31I have a friend who I train with.
01:31:34This is like, you know what I love?
01:31:35Some of the people that I love the most
01:31:38are the ones who you wouldn't predict
01:31:40have the life that they do.
01:31:42And there's a girl who trains Lyft ATX on a Saturday.
01:31:46Lovely girl, I've trained with her a bunch of times.
01:31:48Real cool, boyfriend's cool.
01:31:50Like, does fitness modeling, super hot,
01:31:52the long hair lift, the big, you know,
01:31:55but super strong, all the rest of the stuff.
01:31:56Like, feminine as well.
01:31:58Quantum computing degree.
01:32:01Like, works in quantum computing.
01:32:03And she was telling me about quantum AI.
01:32:04And she was telling me about QAI as it's referred to.
01:32:07And it's a burgeoning field, supposedly.
01:32:09Unless she's lied to me.
01:32:11Unless she's totally fucking lied to me.
01:32:12- Yeah, I'm curious what they're working on.
01:32:14UT Austin has good quantum theorist.
01:32:18Look, I'm searching for it.
01:32:20A guy I knew from MIT, they hired him away there.
01:32:22Or see quantum, quantum AI.
01:32:26Merges quantum computer with machine learning in the process.
01:32:28High dimensional data faster than classical systems.
01:32:31Now they're working on it, but I don't know,
01:32:34I don't know how that's gonna work, basically.
01:32:38So I don't know what they're working on,
01:32:39but it's not something that you hear a lot
01:32:41in computer science circles yet.
01:32:42So maybe they'll have some breakthroughs.
01:32:43It's worth looking at,
01:32:44but I don't know how that's gonna work.
01:32:46- Okay, one of the other elements, I guess,
01:32:50that people struggle with when it comes to deep anything
01:32:55is learning, the process of learning.
01:32:58Talk to me about the mechanics
01:32:59of keeping a deep reading habit alive.
01:33:04- Well, I mean, I think reading pages
01:33:07is probably the cognitive equivalent of steps, right?
01:33:10So if you're a 10,000 steps a day person,
01:33:12as like, this is just like a baseline to make sure that like,
01:33:15at least my physical systems are being used,
01:33:17you should have a page count.
01:33:1925 pages a day, 20 pages a day of reading a book
01:33:23just as like getting those cognitive steps in.
01:33:26Because I think we recognize more and more,
01:33:29reading, I would say it's the cheat code,
01:33:31but it's better to think about it as like,
01:33:33reading is the thing that formed the modern brain.
01:33:36And I'm like, I'm more and more convinced about this.
01:33:39I have a book idea I'm working on now,
01:33:40where I'm sort of exploring this idea.
01:33:43The brain before we had the Neolithic revolution,
01:33:47it was the same neurons, right?
01:33:5015,000 years ago that we have right now,
01:33:52but if we go pre-reading,
01:33:54those neurons were doing the things they were evolved to do,
01:33:56which is very much about like the visual system
01:33:58and the audio system,
01:33:59and we could communicate through spoken language,
01:34:01and that's fine.
01:34:02And then we invent reading.
01:34:04This is not something that our brain has evolved for.
01:34:06So in order to read, we have to go through this,
01:34:08this sort of excruciating process of learning to read,
01:34:11in which what you're doing is actually rewiring sections
01:34:14of your brain to connect in ways
01:34:15that they weren't originally meant to connect to.
01:34:17So we're reforming our brain when we learn how to read.
01:34:20And we develop what Marianne Wolfe
01:34:21calls deep reading processes,
01:34:23where you've now yoked together different parts of your brain
01:34:27that don't normally work together,
01:34:28that can now have to work together
01:34:30in order to understand written text.
01:34:33Once your brain is wired to do that,
01:34:36it can, if you reverse this and write,
01:34:38you can generate much, much more sophisticated thoughts
01:34:41than you can if you haven't done this wiring,
01:34:43and your understanding of things,
01:34:45the complexity of what you can understand
01:34:48when you have this new rewired brain,
01:34:49that also really goes up.
01:34:51So reading is like, it's not just,
01:34:53oh, I get stronger in my brain.
01:34:55It reconfigures your brain into like the modern,
01:34:58you know, post cognitive revolution brain.
01:35:01- Okay, why is it important to read physical books
01:35:05then what is lost if I read Substack?
01:35:08I know that you're a fan of Substack.
01:35:10I love Substack, I think it's fantastic.
01:35:12What's the difference between reading it on a laptop
01:35:16versus a phone versus a Kindle
01:35:18versus a physical piece of paper?
01:35:20- Well, there's two different things going on here.
01:35:21There's medium and content type.
01:35:24Like, so if you're reading a book in a physical book,
01:35:27or you're reading in a Kindle, doesn't matter, right?
01:35:30I mean, they're both actual physical medium.
01:35:32Like the way that the Kindle is actually a physical experience
01:35:36that it's actual little discs that are, you know,
01:35:39dark on one side and light on the other.
01:35:41And they make a page, they have little electrical impulses
01:35:44and you shock the disc you want to turn
01:35:46and you don't shock the ones you don't want to turn.
01:35:48And so you've literally created an actual black and white
01:35:51physical version of the page on the Kindle.
01:35:53You're not unlike a computer screen or a TV
01:35:55where it's light being emitted.
01:35:57There's no light being emitted.
01:35:58It's physically that's the page.
01:36:00It just created a new physical page that has text on it.
01:36:02That's why you have to actually have a light
01:36:04on a Kindle to read it.
01:36:05So it's just a page that reconfigures itself
01:36:08into a new page.
01:36:09I love eating technology.
01:36:10I think it's really cool.
01:36:11Content type, the issue is, I mean,
01:36:13there's a lot of this research we've known since the 90s.
01:36:14A lot of this is captured in the best book on this
01:36:18would be The Shallows, Nick Carr's book, The Shallows.
01:36:22When we're reading something like a webpage or substack,
01:36:25for whatever reason, we skim much more aggressively.
01:36:28That's the main issue.
01:36:28We jump around much more aggressively,
01:36:31just trying to pull out the key points.
01:36:33I think that's all just acculturated, right?
01:36:36Like you could sit and read,
01:36:38like if you print out a substack article
01:36:40and sit in the library and you read it carefully,
01:36:42it's the exact same thing as reading a book.
01:36:44It's the exact same thing in sense of the experience.
01:36:47On screens, we tend to skim around more.
01:36:49The other advantage of like a book
01:36:51that was actually published versus like a post you see online,
01:36:55it's just better thought through, right?
01:36:57So when you write a book, you spend a couple of years on it.
01:36:59Like you spend a couple of years crafting the book
01:37:02and you might've been based on a lifetime
01:37:05of thinking about this topic.
01:37:07And so you take your time when writing a book
01:37:10and it gets edited and re-edited and you go back.
01:37:12Like I'm writing a book now.
01:37:14I've been working on it off and on for like three or four years.
01:37:16I've rewritten this book like three times.
01:37:18It's like, this isn't right, this isn't clear enough.
01:37:21And so when you go through text
01:37:24that has been that carefully thought through and structured,
01:37:26that's also, you just get a different experience
01:37:28'cause the pieces click together at different scales
01:37:31and it just uses, you build in your brain
01:37:34these intricate interlocking pieces
01:37:37that all hook together and is beautiful
01:37:39and you get that aha moment feeling.
01:37:40There's an actual physical endorphin rough
01:37:42you get in your brain.
01:37:44So I think reading smart books written by smart people
01:37:47that took a long time to write,
01:37:50that's your calisthenics for your brain.
01:37:52It literally changes.
01:37:53You're a smarter person if you do that versus if you don't.
01:37:56- So good, I have to say reading full length books
01:38:02has been, the volume that I do that has been decreased
01:38:07over the last few years, largely because of Substack.
01:38:10So there's a extension for Google Chrome
01:38:13called Push to Kindle.
01:38:14And if I press it, the article appears on my Kindle
01:38:18because I don't like reading on my phone
01:38:20and I don't like reading on my laptop,
01:38:21probably for the reason that you said.
01:38:24But when I think about it, it very much is running downhill
01:38:29because what's the longest Substack that you're gonna read?
01:38:3220 minutes, maybe?
01:38:3425, 25 minutes, a fucking long article.
01:38:37And maybe part of that, maybe part of my penchant for it
01:38:42is that I do get the outcome, right?
01:38:44What is it that I'm looking to learn?
01:38:46Oh, I wanna find out from Steve Stewart Williams
01:38:50about sex differences in desire for sexual novelty,
01:38:55something like that.
01:38:57Okay, well, I will learn the outcome in the same way
01:39:01as I could feed myself food
01:39:03that was just a cube of calories
01:39:05and that would sort of give me the caloric intake
01:39:08that I needed.
01:39:10But what you're presumably reading for,
01:39:13apart from just the enjoyment of reading it,
01:39:15is to be able to recall it and for it to be woven
01:39:17into the broader mental landscape that you've got,
01:39:21which actually probably means you need to spend time
01:39:23and attention with it.
01:39:24And some of the leanness and brevity
01:39:27that comes with an article actually might work against you.
01:39:31Maybe you need it to be said to you in five different ways.
01:39:34Maybe you need the author to meander off onto a story
01:39:36that takes three pages to explain about this guy
01:39:40who owned a Ferrari and parked it outside of a hotel
01:39:43so that you can then come back in.
01:39:45And each one of these is a little Velcro latch hook
01:39:49that you can hook yourself into.
01:39:51And yeah, I wonder whether the reading
01:39:55or discriminating toward reading stuff
01:40:00that is exclusively shorter form results in the sense
01:40:05that I am learning lots,
01:40:07but if you are to actually do some sort of scrutiny
01:40:09around that, well, okay, how much of it can you remember?
01:40:13How long did you spend with this idea?
01:40:14Did you spend long enough for it to be a part
01:40:16of now your mental models and the framework
01:40:20that you, how much can you recall?
01:40:22That would be an interesting challenge.
01:40:25- And the frameworks of understanding are shallower
01:40:26just because it's less time to establish them.
01:40:29So like in a subset, it's not a bad thing,
01:40:32but what can you do?
01:40:33You typically have like one idea
01:40:35and like here's something that supports that idea
01:40:38and here's maybe like a different idea
01:40:40and here's why that doesn't work.
01:40:42And if that's all you're consuming,
01:40:44that becomes your mental model for how knowledge is gained.
01:40:47And I think we see a lot of this,
01:40:49I mean, think about internet culture now
01:40:51is much more conspiratorial.
01:40:54And I don't mean in the like sort of grand conspiracy theory,
01:40:56which it is, but not just in like the grand conspiracy
01:40:59type of thinking, but in the confidence.
01:41:02There's this quick jump to confidence where you're like,
01:41:05that's wrong because of this and boom.
01:41:09And you think that like, this is like the slam dunk case
01:41:11or something like that.
01:41:12That's a result of not reading a lot of books.
01:41:14You read a lot of books, you're like,
01:41:16okay, this is way more complicated.
01:41:19- Everything is way more complicated
01:41:20than you thought it was.
01:41:21- And there's probably a clear truth here,
01:41:23but clear truths are more complex.
01:41:24Like even the notion of what a clear truth feels like
01:41:28comes out of reading books, right?
01:41:30Like you understand, oh, ultimately like this person
01:41:33was right, but it's complicated.
01:41:35And like, yeah, this was not so clear cut
01:41:38and this is like a compromise and this was really important.
01:41:41And these factors were here, but honestly,
01:41:42those factors aren't as big as you think.
01:41:44And this factor really was more important.
01:41:46And so like, this really was the right thing to do.
01:41:48So even like your notion of what's true or what's not true
01:41:52or what it means for something to be clear is like different
01:41:56than if you're just looking at boom, slam dunk.
01:42:00I think it's a big problem online,
01:42:02both sides of the political spectrum do this.
01:42:03Like you want everything just to be,
01:42:06this person is just garbage and completely wrong.
01:42:09And there's like this one simple thing I know
01:42:12that means you're completely wrong and I'm completely right.
01:42:14And you're wrong in like the worst possible sort of way.
01:42:16And that is like such a sopholific,
01:42:19I'm saying the word wrong.
01:42:20- Solid statistic.
01:42:22- Yeah, exactly.
01:42:22You said it, right?
01:42:23I have to read more, but it's sophistry for sure, right?
01:42:27This idea of this is how truth and argument unfolds
01:42:31is like there's an obvious flaw that's easy for me to grok,
01:42:34which I guess now could actually be a verb
01:42:36as opposed to just meaning to understand.
01:42:38Also, I could literally grok it, I guess.
01:42:40And now it's clear that you're wrong and I feel righteous.
01:42:43And then we go seeking that.
01:42:45And then we want to simplify everything in the world to,
01:42:47you're just terrible and this person is perfect.
01:42:50And this idea makes the most sense.
01:42:52And if you disagree with this idea,
01:42:54it's because like you want to eat children.
01:42:56And it just becomes, it's a different under,
01:42:59this is what I think we get wrong.
01:43:01It's not just like we don't have the right information.
01:43:05We've changed what our notion of truth is
01:43:07because we're not exposed to the complexity of truths
01:43:09when you read not only a scholar, like a smart case for it,
01:43:13but then you read the arguments that they confronted.
01:43:14And then you read someone else
01:43:16that's arguing against their point.
01:43:18And you're like, oh, okay, I've seen the clash of minds.
01:43:21And now in that clash, I kind of see what's going on here.
01:43:26Like, yeah, the truth really leans this way.
01:43:29And I feel really real conviction in that
01:43:31because I've seen the best minds come at this
01:43:33from either side and I really understand.
01:43:35It's not cut and dry, but ultimately,
01:43:37like this is the right thing to do.
01:43:39That was like a very familiar thing
01:43:41to people and leaders like in times past
01:43:43where you lose it if you're exposed
01:43:45to these low resolution copies,
01:43:50these low resolution simulacrums,
01:43:52these easy to digest pre-chewed versions
01:43:55of argumentation and understanding.
01:43:56It just changes the way your brain thinks
01:43:58about what true even means.
01:43:59- Yeah, there's an arc to sense making
01:44:02that you kind of need to track.
01:44:04And if you don't track it,
01:44:05you just assume that answers appear.
01:44:07It's like, no, no, they don't.
01:44:08Cal, you fucking rule, let's bring this one home.
01:44:11Why should people go to keep up to date
01:44:12with everything you do?
01:44:13- Oh God, calnewport.com, I guess.
01:44:16My books are on Amazon.
01:44:17My podcast, Deep Questions on YouTube
01:44:21or wherever you get podcasts.
01:44:22Newsletter at calnewport.com.
01:44:24Deep Work, too many things going on now, Chris.
01:44:26Deep Work is 10 year anniversary.
01:44:28I'm excited about it.
01:44:29All new, I replaced all the blurbs on the back
01:44:32with most of them are now organic.
01:44:34I could just like, people who have said things about it
01:44:36without me asking them to say it, so that's fun.
01:44:39And I have a masterclass out on this stuff, too.
01:44:42So I don't know, it's everywhere.
01:44:44Too many places, I feel too busy.
01:44:46- For a person who's a digital recluse, you are everywhere.
01:44:49But that's a function of focusing on quality, not quantity.
01:44:52I can't wait to speak again, man.
01:44:53This has been so much fun.
01:44:54I appreciate the help.
01:44:55- Always a pleasure, Chris.
01:44:56Always a pleasure to talk with you.
01:44:58- Congratulations, you made it to the end of an episode.
01:45:01Your brain has not been completely destroyed
01:45:03by the internet just yet.
01:45:05Here's another one that you should watch.
01:45:07Go on.

Key Takeaway

To thrive in an age of digital distraction and AI-generated content, professionals must reclaim the ability to perform deep, concentrated work while organizations must replace the hyperactive hive mind with explicit workload management and structured communication protocols.

Highlights

The modern workspace is trapped in a "hyperactive hive mind" workflow characterized by constant ad hoc communication via tools like Slack and email.

Data from Microsoft 365 suggests that workers are interrupted every two minutes, leading to significant "work slop" and a reliance on weekends for actual deep work.

Cognitive context switching between abstract tasks takes 10-20 minutes to fully recalibrate, making frequent interruptions biologically exhausting and unproductive.

Large Language Models (LLMs) are currently being used to avoid the strain of "blank page" cognition, but often result in low-quality "work slop" that adds to organizational burden.

Scaling laws for LLMs may be reaching an asymptote, suggesting future AI progress will rely on hybrid, specialized architectures rather than just bigger models.

High-value employment in the AI age will depend on an individual's ability to tolerate cognitive strain and produce rare, valuable, and non-automated output.

Timeline

Reflecting on Ten Years of Deep Work

Cal Newport reflects on the decade since publishing his seminal book, noting that the issues of distraction and shallow work have significantly worsened despite his predictions. He discusses his early pushback against the ubiquity of social media and the pressure for everyone to be constantly online for their careers. While the public has become more mindful of social media's downsides, the professional world has doubled down on hyper-distraction through email and instant messaging. Newport argues that this shift is not just a personal struggle but a massive economic failure that leaves money on the table for corporations. He feels a sense of vindication regarding his early warnings but remains concerned that the workplace environment has become more hostile to deep concentration.

The Data of Distraction and the Hive Mind

The discussion shifts to hard data, citing a 2025 Microsoft 365 report that reveals workers are interrupted on average once every two minutes. This constant switching to communication tools has pushed core productivity tasks like writing reports or building presentations into the early hours of Saturday and Sunday mornings. Newport introduces the concept of the "hyperactive hive mind," an ad hoc style of collaboration that emerged with email and was perfected by Slack. While Slack is the "right tool for the wrong way to work," it creates a cycle of administrative misery and low-quality output. The conversation highlights the misalignment of incentives where businesses prioritize responsiveness over the actual production of value. This environment creates a "love-hate" relationship with tools that make collaboration easier but human life more stressful.

The Biology of Context Switching

Newport explains the neurological cost of context switching, noting that the human brain requires 10 to 20 minutes to fully load the relevant information for an abstract task. Frequent interruptions prevent the brain from ever "locking in," resulting in a state of diffuse cognitive friction and fatigue. This mental exhaustion makes simple tasks like clearing an inbox feel like torture because each message requires a completely different context. To combat this, Newport suggests three pillars: training the personal ability to focus, fixing communication protocols to eliminate the hive mind, and managing total workload to prevent over-commitment. He emphasizes that focusing is a skill that must be practiced like a sport to achieve "superstar" status in the knowledge economy. The section concludes with the advice that being accountable allows one to be less accessible, as high-value output grants autonomy from the meeting-heavy culture.

AI, Work Slop, and the Future of Scaling

The conversation explores the impact of AI on productivity, specifically the emergence of "work slop"—low-quality AI content that makes everyone's job harder. Newport argues that many workers use AI to avoid the "blank page" problem because their brains are already exhausted by context switching, but this leads to unreliable and shallow results. He delves into the technical side of AI, discussing the "Kaplan curves" and the possibility that LLMs are reaching an asymptote in performance through pure scaling. The future likely holds a shift toward hybrid, specialized AI architectures rather than one giant "oracle" in a box. Newport advises individuals to lean into the discomfort of "cognitive strain" to differentiate themselves from AI-reliant peers. By seeking out hard problems and deep focus, workers can produce the rare and valuable results that the marketplace will continue to reward.

Rebuilding the Modern Brain Through Reading

The final segment focuses on the foundational role of reading in shaping the modern human brain and maintaining cognitive health. Newport posits that deep reading rewired the human brain post-Neolithic revolution, allowing for more sophisticated and complex thought patterns than spoken language alone. He contrasts the experience of reading physical books or Kindles with the shallow skimming habits often associated with internet content and Substack. While digital articles are useful for quick information, they often lack the depth required to establish nuanced mental models or appreciate the complexity of truth. He encourages a daily habit of at least 20-25 pages of deep reading to maintain mental agility and counter the simplifying effects of internet culture. The interview ends with a call to action for both individuals and organizations to prioritize focus as a tier-one professional skill.

Community Posts

View all posts