Ep. 372: Decoding TikTok’s Algorithm

Last week, it was announced that Oracle would take over operation of TikTok in the US. One of the primary reasons proposed for this deal is that it was in national interests for us to take over control of TikTok’s fabled “algorithm.” But what is this algorithm? How does it work? To what extent can it be controlled? In today’s episode, Cal looks deeper at these questions and arrives at a broader philosophical point about the role these services, in general, should play in our civic culture. He then answers listener questions on the same topic and reacts to the three recent news articles about TikTok’s influence on our culture. Below are the questions covered in today's episode (with their timestamps). Get your questions answered by Cal! Here’s the link: bit.ly/3U3sTvo Video from today’s episode: youtube.com/calnewportmedia Deep Dive: Decoding TikTok’s Algorithm [0:03] Should I quit social media, even if I don’t live on the slope of terribleness? [44:44] Is it bad to use TikTok to keep up with the news? [49:41] How does quantum computing relate to AI? [53:48] CASE STUDY: Trying to become a great composer [1:02:29] CALL: Tips for dealing with growing kids [1:08:58] CAL REACTS: Three Articles about TikTok (and Three Sighs of Exasperation) [1:18:04] Links: Buy Cal’s latest book, “Slow Productivity” at calnewport.com/slowGet a signed copy of Cal’s “Slow Productivity” at peoplesbooktakoma.com/event/cal-newport/Cal’s monthly book directory: bramses.notion.site/059db2641def4a88988b4d2cee4657ba?the-independent.com/news/world/americas/crime/tiktok-surfing-challenge-manslaughter-b2832760.htmlyahoo.com/news/world/article/why-people-on-tiktok-think-the-rapture-is-coming-and-that-the-world-is-ending-this-week-202334215.htmlbbc.com/news/articles/c4gj5lqq52loyoutube.com/watch?v=z7Df9yGofLoyoutube.com/shorts/xWIVA0ZJm-w Thanks to our Sponsors: monarchmoney.com (Use code “DEEP”)reclaim.ai/calwayfair.commakeheadway.com/deep Thanks to Jesse Miller for production, Jay Kerstens for the intro music, and Mark Miles for mastering. Learn more about your ad choices. Visit podcastchoices.com/adchoices

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:03Cal Newport?

  The big news in the world of social media recently is the announcement made last week that US interest would be taking over operation of TikTok in America. Now, this deal is complicated, so I'm going to read you here from a New York Times article that's explaining it. So, the article says the following. "The software giant Oracle will oversee the security of Americans' data and monitor changes and updates to TikTok's powerful recommendation technology under a new deal to avert a ban of the service," according to a senior White House official. "A copy of the algorithm, the recommendation engine that powers the app's addictive feed of short videos, will be licensed from China to an American investor group that will oversee the app in the United States," the official said.

• 0:47Cal Newport?

  All right, so we're going to have this new entity, TikTok in America, that will be created to run and monitor all of this, and American investors will have an 80% share in it. Here's some questions that come to mind. Why does our government care about this? Well, TikTok's mysterious and acclaimed algorithm seems to know so much about its users and is so good at serving up videos that meet their interest that there's a general fear that a foreign government could be gaining too much influence over our population. But this begs a follow-up question. How does this algorithm actually work? Is this something that can be fixed once the right people have control over it? In other words, is our problem with social media today largely one of bad algorithms? That's what I want to look at today.

• 1:32Cal Newport?

  I'll put on my computer scientist hat and take a closer look at what's going on underneath the hood of services like TikToks, and then I'll put on my digital ethicist hat to use what we learn to better understand what role these services should have or should not have in both our lives and our civil culture. As always, I'm Cal Newport, and this is Deep Questions.

• 2:03Cal Newport?

  Today's episode, decoding TikTok's algorithm. All right, let me tell you this, as a computer science professor who specializes in algorithm theory, I mean, this is what I studied in grad school as distributed algorithm theory. It's what most of my academic CS papers are on. I teach algorithms at both the undergraduate and graduate level. So it's been sort of amusing and pleasing for me to see how often I'm hearing these days, non-computer science people use the word algorithm. It's sort of like my little secret world is one that everyone has been exposed to. But when it comes to discussions of social media platforms like TikTok, this term algorithm

• 2:44Cal Newport?

  has seemed to taken on some sort of almost mystical power and capabilities. I want to play a clip here of a sort of recent discussions from last week of in the news, people talking about the TikTok algorithm. Jesse, let's hear this, this first clip.

• 3:01Speaker 01

  And they will basically be helping TikTok to retrain its algorithm and also to make sure that US data of people, all the Americans, you know, the roughly half the country using TikTok, that all of it is secured.

• 3:14Cal Newport?

  We can cut it there. All right. And then I think we have another clip, right? Mm hmm. All right.

• 3:19Speaker 02

  ... from surveillance or interference by foreign adversaries. And the algorithm, I know this is a question many of you have had, will be secured, retrained, retrained and operated in the United States outside of ByteDance's control.

• 3:33Cal Newport?

  All right. So there we see the way the algorithm, putting square quotes around that, is being discussed in the media right now. It's this thing, and we want Americans to be in control of it, that we're going to retrain it once it's over here. And with that sort of supervision, we can have some assurance that whatever it is that we are uncomfortable about happening with service like TikTok, we can have some assurance that at least our interests are being preserved. But what is this algorithm that is at the core of this new deal? Well, I want to start with the mental model that I think most people have when we talk about a social media content curation algorithm. I think most people imagine it's basically like a digital version of a newspaper editor.

• 4:20Cal Newport?

  So we have like a newspaper editor who makes decisions. What's going to go on the front page today? What's worth focusing on? No, put that aside. This is going to have a banner headline. We imagine an algorithm like we're building a computer version of that, a digital version of that that can work at like really high capacity and make like building a sort of like a custom newspaper for each person. A computer program that makes decisions about what we see in the same way that like a computer editor might do.

• 4:48Cal Newport?

  Okay. So in that model, if that's our mental model for algorithms, transferring control of the TikTok recommender algorithm to US control makes sense, right? We wouldn't want a foreign country playing the role of the editor for the social media newspapers that half of the US population is receiving through TikTok. We want an editor who has our values, who has American values, who's not going to be doing not only promoting values of not American, but maybe conspiratorially trying to Manchurian Candidate style influence Americans to think one way or the other. Right? We would, we would not be happy if the New York Times was edited in a dark room in Beijing. And so we sort of feel with this mental model, we want our algorithm for this popular service to be here. But is that mental model correct?

• 5:32Cal Newport?

  Well, I'm going to take out my, my computer scientist hat here, which is Jesse and I have discussed as an awesome hat. It has Spock ears, Jesse. Let's be honest.

• 5:42Cal Newport?

  And we're going to take a closer look at what really is going on and we're going to correct our mental model for thinking about these social media algorithms. All right. So what is Oracle going to discover when they finally get their hands on TikTok's algorithm as part of this deal? You know, like most private companies in the social media space, ByteDance has not published detailed looks at exactly how their systems work. I mean, why would they? Most companies don't if they don't have to, but it's not like they've been completely secretive either. We, we have a pretty good sense about more or less how TikTok works. There's two sources we can draw from. One, ByteDance researchers in the last five years, they published two different major papers in academic venues that looked at major components of their, their system architecture

• 6:26Cal Newport?

  they use for multiple products, including TikTok. So that gives us a really good insight. And two, We know how recommendation, the evolution of recommendation system algorithms, that this is something we just know from a data science, computer science perspective. There's, there's only so many ideas you can be pulling from. No one suspects there's an entirely new idea that they're using. So we put these two things together. I think we can create a pretty reasonable understanding of what's actually going on here. All right, so let's get technical. First of all, when we say algorithm, that's not really the right word. We should be talking about recommender architecture or recommender system architecture.

• 7:05Cal Newport?

  So what drives TikTok is a, a massive system that stores and updates information about hundreds of billions of videos and whatever it is, over one to two billion users. And all this stuff is stored and accessible. And this, in the end, what this system has to do is for each individual user, as they do each swipe, is have a recommendation of what next video to show them. So it's actually a very large global distributed system, not a single algorithm that we need to look on. Now, we know a lot about building these type of systems. This idea of using computer driven recommendation systems really began to get a lot of attention in the late 90s and into the 2000s with Amazon and then Netflix really being pioneers and figuring out, how do we use data about stuff we want to show or sell to people?

• 7:51Cal Newport?

  Data about the people themselves? Put these two together and make good recommendations. So by the 2010s, we were getting pretty good at these. What type of recommender architecture does something like TikTok probably use? Almost certainly, they're using a general approach that's known as a two tower system. Let's get into this a little bit. Let's set the, I want to set the, the stage here before we get to the two towers that, that they probably utilize. All right. So imagine, you know, I am building a system to recommend short videos to you. You have a bunch of short videos, you're a user. I want to recommend videos to you. One way I might want to do this would be to come up with a master list of properties.

• 8:35Cal Newport?

  And I want this to be on the order of a thousand, few hundred, maybe a couple thousand, but kind of in that order, right, of a few hundred, maybe a thousand properties, right? That I can go to any particular video and I can assess it on each of these properties, right? So maybe you have a property that's like, it's funny or not. Or there's a property that is, you know, the content is conservative, or there's a, there's a property that's like, this content involves like, South Korean K-pop music. Now, it could be more complicated than just singular properties you have or don't. It could be complicated combinations, like maybe one of my properties I assess videos on is, okay, it's like conservative K-pop content that's not funny. Because maybe it turns out like, that's a pretty important property.

• 9:18Cal Newport?

  But this is my challenge, is have like, a pretty limited list of properties that captures enough stuff about videos, like the stuff that people really tend to care about, that I can start to make good recommendations. And now, I'm not going to necessarily just have like, a bit for each, is it funny or not? But maybe I have like, a one to 10, or one to 100 scale, where you know, how, how close, how much of this property does it have? All right, so imagine I could do that. And I go through each of the videos in my system, and I have my big list of properties, and I give it a score on each of those. Now, assume when you join this service, I say, okay, you're a new user. I'm going to sit down and have like, a detailed user interview with you. I'm going to talk to you about all sorts of stuff, what do you like, what you don't like, let me show you some videos, is this the type of thing you like or not? And what I'm going to do is take that same list of properties.

• 10:03Cal Newport?

  And now, I'm going to go through, and for each of those, try to measure, write down with a number, how important each of those properties is to you. In other words, when it comes to videos you like, do you want a lot of this property or not? So like, I find out, oh, you love funny stuffs. I'm going to, in that same category, I'll put down a pretty high number for, you know, funny stuff. But maybe you're really left-leaning, so you're like, I do not want to see conservative content, so I'm going to put a zero, big zero there, right? Like, that's not something you care about. So we have this master list of properties, every video is ranked on them, they have a little this tag we put on every video, and every user has the same properties. We say, okay, here's how much they care about each of these properties.

• 10:42Cal Newport?

  Now, assume I was just able to do that. Don't ask me how, but just say, I was able to do that. This now gives me a pretty good way of recommending videos to you. What I can do is say, look, here's what I want to go find. I have 100 billion videos, fine, but what I want to look for is videos that strongly overlap properties you care about. So where you have like a high score in a property, if the video also has a high score in that property, I'm like, that looks good. But I also want to make sure that it doesn't have high scores and properties you don't care about, right? So the more it matches your preferences. So it has the things you like and it doesn't have the things you don't like. The better candidate I'm going to say that video is to show to you. And so that's what I'm going to do to figure out a video to show to you and then I'll show it to you.

• 11:27Cal Newport?

  If I could do that, the recommendations I make are going to seem like really good. Like, yeah, man, how do you know? Like, I like baseball and, you know, like, like left-wing politics and I'm really into Lord of the Rings. And my God, there was a video somewhere in these like hundred billion, you know, videos where you have Frodo making a sort of like left-wing argument using a baseball analogy, right? You're like, wow, you know me really well. But it's what we really know is this video matched a bunch of things you cared about and didn't have a lot of the things that you don't. A two-tier recommendation system is one way of building an automated system to do something like this.

• 12:10Cal Newport?

  Now, the way this works, I'm going to, God help us, draw a picture, Jesse. This is like an impossible thing to diagram and I'm a, I'm a bad. Actually, I take that back. Or I'm using the wrong pencil. But you can all tell I'm a fantastic technologist. I couldn't even figure out how to use an Apple pencil. I was actually using the wrong pencil. Okay, so what I'm going to do here is I'm going to draw a picture here. Make the two tower system, make this make more sense. All right, so on the screen here, I'm going to draw to on the right over here, a tower. All right. Perfect diagram of a tower. Right, so this is our first, is our first tower. This one is going to be, inside of it is going to be a collection of machine learning type of tools.

• 12:55Cal Newport?

  There are going to be things in here. They often think about layers, neural networks and transformers and embedding matrices. And it's just mathematical manipulation type stuff that's complicated, but you don't need to worry about it. All right, so what we have as input to this, this tower is we have all of these videos. So people have just uploaded all of these videos. So we just have billions of videos. Okay, and then we're going to take these videos and they're one by one, we input to this tower. And what's going to come out on the other side of the tower is going to be a property list for that video. And I'm going to indicate that with, I don't know, I'll just put like a a big purple box.

• 13:37Cal Newport?

  All right, so that purple box is a, it's a vector of numbers, but just think of it as if we have our thousand categories and it has a number for each of it. So it's just describing the video using whatever sort of the sort of master list of things we care about. So we can do this for each of the videos. And in the end, we'll have this, think of it like a big database of billions of videos. We can do this like when they're uploaded, like the videos don't change. And we have a way of describing this big, long list of, of cat numbers that describes the properties of each of those videos. That's tower number one. We want videos through it and it gives us this description of the, these videos with the properties we care about. All right, but what's tower number two? The two tower situation here.

• 14:22Cal Newport?

  And there's no way, by the way, Jesse, that this terminology did not come out of Lord of the Rings fandom within computer science. Let's be honest. Like, do you even know that reference, the two towers is the name of a Lord of the Rings book?

• 14:35Cal Newport?

  No. You spend too much time doing sports. You gotta. I might have, but I would have lost it. All right, so here's what we have over here. We have a second tower. And the way this tower works is now our input is going to be. People, and by people, I mean their user profiles. So it's like a description of everything we know about them, including primarily, and this is the key thing, their behavior on the platform. So things they've watched before, things they haven't watched before, how long they watched various things. So it's all this information about these people. And we run each of the people through, they have their own tower, which again, inside of it's mathematical stuff. There's neural networks, there's transformers, there's embedding matrices. Don't worry about the math.

• 15:15Cal Newport?

  And what we get on the other side is it will also explain, describe each person with their list of when it comes to these same properties, hey, how much do they care about each of these things? So we have a common way, one tower that does nothing but describe videos. It's called the item tower. And one tower that does nothing but describe the interest of users. We call that the user tower. And the key thing is, is we describe both these things with the same way, the same list of categories. All right. So that's the idea. Now, how do we teach these towers to do it? Here's the important thing. This is done in a largely like semi-supervised manner using machine learning techniques.

• 15:55Cal Newport?

  So it's, humans don't sit down and decide what goes in each of these, you know, what should these categories be? What are these things we are rating? Humans don't, we don't decide. Instead, what we do is we train both of these towers at the same time, in the same ways we train other sort of neural network-based systems like language models or visual, vision models or other stuff that you're used to. And here, the data we have, the training is, we have a lot of examples of, okay, we know this user likes this video, we know this user didn't like this video. You can use that data to train these things together. And what is the goal that you're training them for? You say, come up with, I don't know how you're creating these, these, these, these categories and these numbers and how you're describing things. I don't know. It's not for me as a human to know.

• 16:39Cal Newport?

  But what I want to see is... The vectors you use, the descriptions of things that people care about, should be pretty close to the vectors you use to describe the stuff we know they like, and not too close to the things they don't like. So I don't know what's in them, but I have a bunch of examples of stuff that real people, real people did, and what videos they really like or don't like. And I want you to keep nudging and changing your internal descriptions, until you get pretty good at describing people and describing things in such a way that the vector describing the people is close to the things they like and not close to the things they don't. So we don't know what's in these vectors. We don't know how, in these two tower recommendation systems, what they're looking at,

• 17:23Cal Newport?

  what these neural networks and transformers, etc., what they're noticing in these videos, what they're noticing in our human behavior. We have no idea. It's just a list of numbers. But we know it does a good job. That when we test it and say, "Okay, we know this user likes this video." We say, "Yeah, you did, these things overlap pretty well." And we know this user doesn't like this video. "Yeah, they don't overlap very well." So you train these towers together. They learn, in this diagram here, the pink stuff. They learn what, some useful way of describing users and describing videos, so that it does a pretty good job of matching them. Okay, so then if we step back, how does the final recommendation work? All right, so again, now we can draw from some of these architecture papers

• 18:06Cal Newport?

  that ByteDance themselves have published. But what typically happens in these systems is you have, you have so many, you have so many of these items, so many videos, that what you do is you say, "Okay, we're gonna do like a really rough first path to get some candidates of what to show the user, and we'll use just entirely something like a distance metric, like a way of just, here's a list of numbers, here's a list of numbers, how close are these list of numbers? There's different ways of measuring this. You can do pretty quickly mathematically. And we're just gonna go through and grab a bunch of videos that are, have a pretty close, they're close by the sort of mathematical notion of close, to the user's, to the user's description, to their vector.

• 18:48Cal Newport?

  And then there's a little bit of proprietary stuff at the end is how do we then rank these candidates? And describe what's the actual one to return? And that's actually a place where recommendation systems can have a little bit of human oriented heuristics and rules of thumb. And this final step, we're like, okay, here's 100 videos that are like a good match. Now, which one do we want to show them? That's where you can throw in some actually like hand coded, like final little rules or tweaks or rules of thumb. That would kind of happen at the end. So that's basically how these systems work. So why is TikTok so uniquely successful? If that's an architecture that like other systems use, Spotify probably uses something like that. Some other social platforms are using something like this. So why is TikTok work so well?

• 19:33Cal Newport?

  Well, we kind of have an answer to that as well. There seems to be a few things going on here. All right. One is just what that service does. So short form video is a best case scenario for building one of these recommendation systems. Right? All it does is deliver you stuff. It doesn't have to deal with TikTok doesn't have to deal with other complicating factors that other social services have, like your friend graph and who you follow and trying to like mix in the stuff you said you're interested in with the stuff the algorithm thinks you're interested in. TikTok basically ignores that because everyone uses the for you tab. That's what we're talking about here. And so it's just pure. We're just showing you things that we think you'll like. Nothing else matters.

• 20:19Cal Newport?

  Second, because they're short, you get a lot of feedback. An average TikTok user might go through 30 plus videos in a typical session, each one generating feedback about what they like and don't like. So you have a huge amount of data with which to get better and better at making these recommendations. Now compare that to like Netflix where I might watch one series and one movie in a given week. It's a much, much slower data cycle. Right? And I'm not going to try, by the way, if I'm on Netflix, I'm not going to try most of the stuff you recommend. I'm probably going to end up watching something someone told me about anyway. So I get very slow stream of data. If I'm a service like Netflix, but TikTok is optimal because you, you only can look at what they show you. So you're, you're giving them feedback on every single recommendation they get. All right. So that's part of it.

• 21:02Cal Newport?

  The format is super well suited for these type of systems to work. Really well. The other part of the advantage is actually architectural. It's a really smart and powerful distributed system that ByteDance actually built for their products like TikTok. So one of the things that they do that's really impressive is they can update. They update the training of the user tower almost in real time. They don't just train these two towers once, then go deploy it. Now let's go use it. As you're using the app, you're getting more data. You're generating more data about yourself. Which videos did you watch and how long did you watch them? Well, they built a system. This is really pretty amazing from a distributed systems point of view

• 21:47Cal Newport?

  that can essentially be constantly trying to retrain your part, the user tower piece, using this new data so that it can sort of respond in real time. This is really hard to do, but the way they do it is with this massive distributed system where, you know, there's all, it's fragmented among all these different systems. There's probably a system near you that's working on it. In the U.S., most of this is an Oracle Cloud infrastructure. So there's probably some local machine doing it. And then they transfer over the new train parameters to the production model that's actually making recommendations really frequently. And there's this whole fault tolerance system they have built up so that if this gets partitioned, the system can still run. It's really hard computer engineering, but it allows them to continually update how it labels your info, how it labels you and what you care about,

• 22:31Cal Newport?

  like almost immediately in reaction to the stuff you're doing. This is what gives TikTok, for example, its amazing cold start capability, where if you're a new TikTok user, you just start watching things and swiping. And within 10 minutes, like, how is this already showing me stuff that I really care about? It's because they built this architecture that can retrain parts of the towers in real time alongside of what you're doing. The other thing we know they've done in their system is that it's not a pure user-based, history-based recommendation system. They have a parallel system. That's doing nothing but studying what's popular. Hey, what's doing well on our network, maybe worldwide, or what's doing well on our

• 23:12Cal Newport?

  network in a particular region or among like a general group of users? And they mix this in. They call this the short-term profile versus the long-term profile, which is the user description. They join these two things together. So when they're trying to figure out what to show you, yeah, there's a bunch of candidates that really match your expressed interest, but there's also candidates that maybe are a looser fit to your expressed interest, like a reasonable fit, but a much looser fit, but are trending or really popular right now. And so those get mixed in. And then so you can get shown, not everything you're being shown is just here's the best match to what you've shown interest in before. It's also like, hey, this thing is really popular right now. People like it. It roughly overlaps stuff you care about. Let's throw that into the mix.

• 23:57Cal Newport?

  And then this becomes a feedback mechanism that allows you to see things that aren't like an exact fit for things you've seen before. And maybe you like it and it begins, you watch it for a while and it allows the model, when it's describing you to sort of learn about other interests you may or may not have. So it's why you get this mix. I mean, you know this when you use TikTok, you get this mix of, oh, this is straight shot matching to one of my clear interests, but also like, oh, this is weird and kind of compelling, but kind of off the wall. And maybe half of those things you see, you end up watching them. That it's it putting in this sort of real-time popular stuff as well. So they have sort of a secret sauce for mixing those two things together. So basically, it's just no big new ideas. It's just a very well-executed system.

• 24:41Cal Newport?

  It's a system that was built at the highest level in a format, short video, one-by-one algorithm recommendation, that is perfect for this type of recommendation system. You put those two things together and the whole thing seems pretty eerie, like, my God, it learns me so fast. It knows more about me than I thought I knew about myself and it works really well. All right. So that's what Oracle's going to find. There's no magic in there. It's going to, these are a well-implemented distributed system. There is no magic description of a, you know, newspaper editor in somewhere that you can tweak. Just a really well-built distributed system that runs these machine learning-based categorization algorithms.

• 25:22Cal Newport?

  So what does this teach us? Well, modern recommendation architectures like the one run by TikTok are not digital newspaper editors. They're not things that we can easily configure to reflect particular values or interests or philosophies. The machine learning techniques used in these two tower architectures are completely agnostic to what they're trying to describe. It could be videos, it could be data from a science experiment, it could be, you know, descriptions of shopping behavior, moving watching behavior. These algorithms do not care. They've just been optimized in a relentless training model for I am assigning list of numbers to things. And if these numbers are close to the numbers for this thing over here, the user,

• 26:08Cal Newport?

  and the system tells me that's good, then I think my numbers are good. And if the system tells me they're bad, then I adjust how I do it until the system tells me it's good. There is no intention in here. There is no visibility into how things are being described or what matters or what the values are. It's just trying to win this training game of I don't know in advance as the two towers, you know, who this user is or what they like. So I better have described them in a way that ends up matching the things that they liked. The way these systems actually work in terms of if we want to think about what are they actually doing? If you talk to a machine learning or data scientist, they'll say, yeah, what these techniques do, you give them enough data, what they're trying to do mathematically is build approximations,

• 26:48Cal Newport?

  mathematical approximations of what whatever underlying process or systems best describe the patterns it's fed in its training data.

• 26:56Cal Newport?

  This is why if you feed a bunch of information about, you know, traffic times or something into one of these models, and it gets good at predicting what's going to happen at given times, it has approximated maybe some sort of reality about the underlying traffic system. Like, you know what? There's a lot more cars between 4:30 and 6:30 because that's when a traffic lets out from buildings. It sort of learns these underlying, approximates these underlying systems and processes so it can do better at predicting what's going to happen. So when it comes to serving content like videos that feature other people to other people, this method of curation, I believe, is something that should give us some pause. And the reason is, if we think about this historically, humans have always been

• 27:39Cal Newport?

  a little uneasy and a little wary about the production of mass content.

• 27:45Cal Newport?

  We worry about it because we know content has a real impact. People talking to other people and mass content since the beginning of the printing press has had both good and, you know, bad impacts. Just look at like the witch trials that happened all throughout Europe and eventually making its way to colonial America. A lot of this came out of some, you know, printing that got people thinking about this or that. We worry about the power of content. And the reason is, is the human psyche has dark elements. We have a hardwired affinity for hatred or violence or dehumanization, an attraction to the grizzly, an attraction to the purient. You know, we, we have a lot of dark parts in our brain and we try to appeal to our better angels, especially when it comes to mass content. So we have all sorts of guardrails we put up.

• 28:30Cal Newport?

  If I'm editing a newspaper, if I'm a producer for a television program, if I'm, you know, producing a podcast, you know, what I'm going to say or not going to say to accomplish my goals, we have, we're careful about it. We're careful in about it because we know there's a lot of stuff in the human brain, a little more primitive, that we don't want to appeal to. We integrate human values into how we curate content. And when we don't do that, we get really upset, you know, worried about it. I mean, this was like World War II propaganda. What was that? If not, basically a group saying, throw those guardrails aside. What matters is our goal is more important. And then we look back at like World War II era propaganda, like, ah, that's not great. We don't be like, hey, what great communication.

• 29:13Cal Newport?

  Like, no, no, no, we don't go there, even if it could help our cause. Careful about it. We have human values. These type of recommendation architecture don't share our values because they don't know what values are. They don't know what they're doing. They're just producing numbers to win a game or getting positive or negative zaps from a, a machine learning algorithm. So when we ask a system like this, hey, do a good job of recommending stuff. It's like, great, I will do what I'm mathematically supposed to do, which is build these mathematical approximations of the underlying systems and processes that help explain the patterns I've seen. That means it's going to be building models of the dark impulses. It's inevitable. It's going to build models about like what the, the, the affinities for hatred or dehumanization or violence or purience, or whatever it is that we're sort of, we, we're, we embarrassed

• 29:58Cal Newport?

  to admit that we're wired for. These algorithms have no embarrassment. They just say like, there's a thing here. We show stuff that appeals to this, it does well. You get basically not a digital newspaper editor, but a digital propagandist of the worst kind. That's what happens when you allow blind mathematical models to start doing content curation. And I think that territory, this is what we should worry about. There's not something here we easily tweak. We're not replacing one country's values with another. There's no knobs to turn about different properties we want or don't want. Machine learning based recommendation algorithms, architectures, again, just to summarize, they are just going to, they're going to model the systems at hand blindly to win the game of finding matching interest.

• 30:43Cal Newport?

  And when it comes to content, that really pushes back against the last 500 years of human experience with how we should deal with content production. All right, so let's move on here to my takeaways. This is excuse to hear some good music.

• 31:05Cal Newport?

  All right, let's go back to the original question.

• 31:08Cal Newport?

  Am I glad that TikTok in America is coming under American control? Well, sure, right? This can't hurt. And there's places where a foreign government maybe could mess with these architectures to screw with us, especially at that last phase where you do some heuristic tweaking in the final ranking of candidates. But will this somehow allow us to fix TikTok in like a fundamental way? If we have the right people controlling the algorithm, can we make these platforms behave in the right ways? No, the answer is no. Machine learning algorithms deployed in this context will relentlessly learn how best to summarize the human condition and exploit us to get it to do what they, whatever it is, whatever goal we have given them.

• 31:51Cal Newport?

  This technique is going to exploit our dark sides just as much as our bright, because it doesn't know the difference between them. When it comes to technology in recent decades, I think we have underestimated the degree to which we, we just sort of implicitly integrate our human values and how we operate in many areas of both our personal and civic lives. And as we start conceding more control of these things to technology, technology that cannot by definition share those values, we begin to learn how unsettling things get.

• 32:21Cal Newport?

  We don't realize how much we depended on these just humanistic moral rules of thumb, these normative standards about what's good and bad. And so it's not so innocent to say, let's let an algorithm serve our news. Let's let an algorithm serve our entertainment. Let's let an algorithm be at the center of the town squares. An algorithm is very different than a person and we don't miss what we have in human types of moralistic thinking until we take it out of our system. So that's my main takeaway about the TikTok situation. I'm sure there's lots of national security concerns and this and that, privacy concerns, great, but we're far away from solving the problem of social media stark impulses. It is baked into the mathematics of how these things execute.

• 33:03Cal Newport?

  It is not an accident or a bad feature someone added late in the process that we can remove.

• 33:10Cal Newport?

  All right, so we have a, I want to keep this conversation going. I want to get Jesse's reaction. We have some questions from listeners that are about this topic. And then I have a few recent articles about TikToks. We can kind of see like, well, what, what really is happening with this technology now? It ain't pretty, but let's all get there. So stay tuned, but first we need to take a quick break to hear from one of our sponsors. I want to talk about our friends at Monarch. You know, I have a theory about financial planning that I want to share here. It's based on my own personal experience. When it comes to financial things like your spending, but also like your 401ks, your stock investments, maybe properties you own, you know, the places where you have money or you're spending money or trying to grow your money.

• 33:53Cal Newport?

  The number one issue that matters is often visibility. What I mean by that is if you can easily get updated on how all of your financial assets and investments are doing, you're going to be motivated to keep going. Like you're seeing, oh, I want this number over here to grow. Let's make sure we keep investing. Let's invest more. Let's invest on, you know, the right, the right track. Or if you see you're spending a lot of money clearly here on something that you don't really want to be like, oh, let me adjust that behavior right away because I'm getting that clear feedback. I'm getting that visibility into it. It also allows you to do things like, I need to change my asset allocations for the year, like all this stuff that really helps you get the most out of your money. The more visibility you have into what's going on, the more you act and the more you act.

• 34:38Cal Newport?

  The better you're going to be, the big mistake, especially for people at my age who have so much going on in their lives, how do you, what's the money you're leaving on the table? It's the money you're leaving on the table because you've been too busy to like invest it properly, to get it out of your bank account and into this, to re-shift out the allocations, to cut out this extra spending that you don't really need to do, but you didn't know it was happening. And thousands of dollars are going out the door that could have otherwise been put into something that was going to grow farther. This is why I've never understood when people are wary of like spending a little bit of money on a service or advisor that can bring more visibility to their finances, that really misses the forest for the trees. Because having more visibility into what's going on in the short term might cost you a little bit, but return massive investments, massive returns in the long run.

• 35:21Cal Newport?

  So this brings us to today's sponsor, Monarch Money, an all-in-one personal finance tool that brings your entire financial life together in one clean interface on your laptop or your phone. And I want to say right off the bat, just for our listeners, Monarch is offering 50% off your first year, so stay tuned for the promo code that will get you that 50% off your first year. All right, here's why I like Monarch. It optimizes exactly the visibility issue I just discussed. It's a beautiful interface that shows you everything, how everything is doing, how you're spending, what's going on, how your different asset classes have been returning. You can do it right there on your phone. The visibility is immediate and complete. And whether or not you also work with an advisor or manage your money yourself, this is a tool that will help you keep taking the actions you need to take.

• 36:06Cal Newport?

  For me, for example, there's like an annual retire investment I do because of my, my company I have, it's something I have to remember to do. Having more visibility into these accounts means I make that investment as soon as I can now each year, which means I am getting more return on it. Before, I would procrastinate, not because I didn't know it was important to do, but I just didn't have a lot of data, you know, visibility in what was going on in my finances. Seeing those numbers got me acting, and that's going to make me so much money in the long term because I'm investing the money early and as much as I can.

• 36:38Cal Newport?

  So when you're looking for visibility in your finances, Monarch is the way to go. Here's two other reasons. One, it's easy to use. You can link all your accounts together and get going in minutes. And two, it's a tool that professionals love. The Wall Street Journal named it the best budgeting app of 2025. So don't let financial opportunities slip through the cracks. Use code DEEP at monarchmoney.com in your browser for half off your first year. That's 50% off your first year at monarchmoney.com when you use the code DEEP.

• 37:08Cal Newport?

  I also want to talk about our friends at reclaim.ai. This is a tool that I'm super excited about. Reclaim is a smart calendar assistant built for people who value time as their most precious resource. It automatically defends what they call focus time, but we know what this really is, deep work time. It will auto-schedule meetings, resolve conflicts, and protect your time for habits, tasks, and breaks that matters. It's a tool for people who care about deep work and worry about meetings making deep work difficult to get done. Basically, it's custom-built for listeners for this podcast. It's the stuff we care about, guys. Reclaim gets us. I've been messing around with Reclaim recently, and here's the best way I can summarize it. It's like having an assistant who sits next to you and helps you manage your calendar,

• 37:53Cal Newport?

  and the same assistant has a prominent Cal Network tattoo. That's basically what you're getting with Reclaim. They know the way we think about things, and it's going to help you act on that. Here's a stat. The average Reclaim user gains seven extra productive hours every week, cuts overtime nearly in half, and sees major drops in burnout and context switching. It's not just about scheduling better, it's about living and working better. So you've got to check this out. Here's the good news. You can sign up 100% free with Google or Outlook Calendar if you go to reclaim.ai/cal. Plus, Reclaim is offering my listeners an exclusive 20% off 12 months if you use the discount code CAL20, name CAL to number 20.

• 38:36Cal Newport?

  So visit reclaim.ai/cal to sign up for free. That's reclaim.ai/cal. All right, Jesse, let's get back into our discussion here.

• 38:46Cal Newport?

  So first of all, Jesse, how do you rate my artwork on this one? I think this was pretty good as my artwork goes.

• 38:52Speaker 03

  I think it's great. I mean, you navigate that pen really well.

• 38:55Cal Newport?

  All right, there we go.

• 38:56Speaker 03

  With the coloring and well, it must be because of all the teaching.

• 38:59Cal Newport?

  I do teach with the iPads now. Yeah, that's probably helping me. You're not a TikTok guy.

• 39:04Speaker 03

  No.

• 39:04Cal Newport?

  Yeah.

• 39:05Speaker 03

  So when you got the ByteDance papers, is that how you got most of that information? Or did you know most of that before?

• 39:11Cal Newport?

  So I, I looked up, I mean, I know about this before. I mean, I know some machine learning. So I know that the two tower system. Basically, there's ways to do recommendations. The original ways that were, you know, came out of the machine learning community, just they don't scale when you replace thousands with billions. It's just too computationally expensive. So this sort of two tower approach is something where you can run it with really large amounts of data. The, the papers get into, the one paper gets into the architecture. They call it the I F, I F S information, something system that they use for TikTok and also two other products. That really gets into their mixture of the two tower recommendations and the real time stuff, like what's popular. And it really gets into the architecture of how they do it, like in the nuts and bolts.

• 39:55Cal Newport?

  And then there's another system that really gets into the, the weeds about how they like very efficiently look up and manage the data that they've, there's a type of hash table called a cuckoo hash you've learned about in computer science circles. And they've modified it and they have all these like interesting ideas they use. And that, that also got in on what they were using here. So like almost certainly it's like some, it's some mix of two tower system plus integrating stuff that's popular. Plus just a really, really good distributed system design. So there's no magic there. There's like no secret sauce to discover. I mean, I really think the reason why it performs better than other social platforms is just the platform. A lot of it is just what, what they're doing. One by one sequential short video serving. That's like the ideal circumstance for a recommendation system.

• 40:41Speaker 03

  Is it normal for technology companies to publish papers about what they're doing?

• 40:45Cal Newport?

  They do a little bit. Yeah, it's interesting. They do a little bit. And I so we see this and then they stop sometimes. It's like OpenAI stopped publishing papers. They used to. And then at some point, like, we don't want to publish papers. Meta AI's group publishes a lot. Like, here's what we're up to. Anthropic publishes what they want to publish. It's more like research papers, studying, like, they want to be seen as like a real academic company, but they don't get into weeds about how their systems work. But yes, you'll you'll discover this at computer science conferences, that there's always a non-trivial fraction of papers that come from technology companies.

• 41:18Speaker 03

  Interesting.

• 41:19Cal Newport?

  Based on their own data, like what they've observed. And I don't know why. I mean, I don't know if it's recruiting. I don't know if it's community oriented. I don't know if it like helps to research scientists. They they want to publish, you know, it's it's interesting, but it's definitely true. They do publish this stuff.

• 41:34Speaker 03

  So if you have doctoral students that want to keep on writing papers forever, they can go to a private company and potentially still do that.

• 41:39Cal Newport?

  Well, it's part of the problem is like the social media companies did this, like Twitter did this. So Twitter at first had an interface where if you were a professor, right? You could get a bunch of data to do research on. Like, I want to do research on, you know, tweets and what they tell us about X, Y, and Z, right? There was an interface, you could get a license as an academic institution, but like, yeah, I want to give me the last 20 million tweets that were on this topic. And then you could get them and then, you know, run research on them. And then at some point Twitter turned that off. Like, no, no, our researchers will write papers on our data. You can't have access to it. And so partially, it allowed them to have really good papers because now Facebook do something similar with their social graph, like only we have the data.

• 42:23Cal Newport?

  So you had to be at one of these companies to write papers about like their data. And that gave them some control over what was published or not published because they get to approve every paper. So they kind of figured out at some point like, hey, maybe this is not, we don't want people writing about our stuff if we can't control what they're going to publish. There's a lot of that going on, or you can go be like a fellow at Facebook or something and use their data, but they get, give you the thumbs up or thumbs down on anything you publish. And there's like a whole period with Facebook, they don't do this as much anymore, but there's a whole period where the original papers were coming out about the, the mental health impacts of some of their products, where they would put a Facebook data, you would have to have a Facebook

• 43:02Cal Newport?

  data scientist on your paper to get access to their data. And the only papers coming out all had these like positive spins on everything and they all had Facebook data scientist. And then eventually like the field was like, we're just going to do these papers without Facebook. And lo and behold, like the outcomes were a lot more negative.

• 43:19Speaker 03

  Yeah. Well, given a third party is kind of like a New York reporter, if they're like following some, you know, somebody around for months and then you don't really know what they're going to write.

• 43:27Cal Newport?

  Yeah. You don't think they're going to write. It's kind of scary. Yeah. Yeah. So anyways, it's interesting how that works, but that's what's going on. So there's no magic. I just, this is a big thing in digital ethics is the role that ethics plays in the way we run systems and algorithms don't have ethics. They don't have like an obvious way of having ethics inserted. So then it, it leads to these weird things. I mean, it's again, it's not like people are tuning these algorithms to do bad things. It's just these algorithms operate blindly. Recommendations is machine learning. I don't even like the word algorithms. It's like machine learning systems approximate the underlying sort of processes and systems that explain the patterns they've given as training. They don't know what they are. They just build mathematical equations that have approximated the systems well enough to properly predict, you know,

• 44:11Cal Newport?

  will this person like this thing? That's all it is. There's no values or ethics in there. And so you're, again, you're approximating the dark as, as easy as you are the, as you are the good. We should be wary about that.

• 44:22Speaker 03

  It's gonna be a big field.

• 44:24Cal Newport?

  I think so. I think so. Digital ethics is probably gonna be, yeah.

• 44:28Speaker 03

  The fact that you guys have the first program is pretty cool.

• 44:30Cal Newport?

  Yeah, we're training up a lot of students that know CS and know ethics and they're trying to figure these things out. So, you know, we'll see what happens. All right, we got some questions here that sort of follow the same thread about TikTok and social media and algorithm recommendations. Let's see what you care about. Jesse, what's our first question?

• 44:44Speaker 03

  First question is from Cara. I truly do not live on the slope of terribleness. I use social media for about one hour per year. Do you think I have an ethical obligation to stop that usage because it's continuing even a little to these companies and maybe helping fuel others addiction?

• 45:00Cal Newport?

  So for those who didn't listen to last week's episode, the slope of terribleness is the model I use to explain the impact of social media platforms, especially curated conversation platforms on your sort of flourishing as a human. And basically, the idea was it's not like these platforms have a constellation of problems and you want to steer your way, steer yourself around them. It's better understood as a slope and you start with the smaller problems, but the gravity of the slope is going to pull you continually to worse and worse problems. Until at some point, hopefully you exert enough mental energy to arrest your slide, but that also takes up a lot of energy. You could be doing and applying to something else. So your flourishing goes down and then you waste a lot of energy. So it doesn't go down further and it doesn't make sense to stop using the platforms. That's what Cara is talking about. See last week's episode for more.

• 45:46Cal Newport?

  Okay, I'm going to say something maybe people in my field might not agree with. I'm not a big believer in the ethical sorting approach to thinking about our engagement with like consumer products. For example, this idea that picked up steam in the last decade, first on the left and now on the right, that what we need to do is we have to sort of ethically rate people and organizations, you know, based on other things they've done. And then be very careful about, uh, this is going to be our expression of our ethical values is like, well, I gotta be very careful. Can I use a product by this company? Where are they in my ethical ranking? Can I like talk to this person or talk about this person? Where are they on my ethical ranking?

• 46:30Cal Newport?

  And it's this idea that our sort of performative or visible ethics are based on who we interact with or who we associate with. So we have to constantly be self policing and trying to assess so that we can be very careful that all of our interactions and engagements with the outside world carefully, carefully match our sort of like internal ethical ideas. I mean, there's a, you know, there's a general principle there that makes sense, but I think it's untenable in practice. And I think it ends up, you end up in sort of like capricious and tribal territory when you try to do this more broadly. So sure, if like there's a massively important principle that a company is like really violating, you know, there's like a company producing roller blades, but they use the bones of puppies to make the wheels.

• 47:13Cal Newport?

  You're like, I, I don't want to buy those roller blades. Right? But for the most part, I think it's, it's just not tenable to do this with all things in your life, and it leads to tribalism and capriciousness. So I, that's not my approach to things. My approach, especially when it comes to technology is to have a value centric approach. What do I value? Right about what do I value? What do we value in terms of like our civic culture? And that's what I want to assess things through. If this product is against, it reduces the things I value in my life without like a commensurate positive upside, then I don't want to use it. If this product is not just making my life worse, but really hurting like civic life

• 47:55Cal Newport?

  or this or that, then, then I don't think it's a good product and I want to find another thing to use if I, I don't want to engage with it. Like in other words, is it hurting things I value myself? Is it hurting like these values, like values I hold true? And if so, I don't want to use it or I want to use it in a way that sidesteps those value violations. So this seems more functionalist or pragmatic. It's basically the approach to technology I give in my book, Digital Minimalism, but I think it's the right way forward. I don't need to assess everything this company has ever done. Is it good? Is it bad? And what does my tribe think? I just want to say this doesn't make my life better or worse. And if it's worse, sorry buddy, I'm not using you. Hey TikTok, you haven't made a compelling case why I have to use you. Your ads during the Super Bowl about it helped the vets bakery business or whatever,

• 48:40Cal Newport?

  not convincing me. That's not what I'm seeing on here. I'm not helping vets sell cupcakes. I'm seeing the, you know, videos about the, the puppy bone rollerblading company and it's kind of making me sad. So I don't need to use you. That's the way I think you should look at it. So if you're using social media like one hour per year, they have some hyper focused functionalist use like, hey, this helps my company to do this each year when we have some announcement and so whatever, that's fine. The key is that's not making your life worse. And if we just have that assessment, a lot of companies that we might like ethically value is like, actually I don't like them. I don't like Mark Zuckerberg. I don't like, you know, whoever they're, they're not going to do too well because deep down, why don't we like them? It's because like their products are making our lives worse.

• 49:25Cal Newport?

  So you, you kind of end up in the same place, but I'm not a big believer of trying to ethically sort everything. I just think it's, it's not a way to run a civil society. Use the stuff that makes your life better and be confident to use the stuff that doesn't, even if everyone else says it's important. All right, who do we got next?

• 49:41Speaker 03

  Next up is Jamie. My son uses TikTok to keep up on the news. A lot of people from his generation do the same thing. Is it good that they stay informed? So how else could they do this?

• 49:53Cal Newport?

  I think staying informed is useful, especially like as you go from your adolescence into adulthood, it's one of the, the, the good skills to learn is like, you should know what's going on in your world. Right? Because that's important as an adult, especially in like a participatory republic like, you know, the US, like you need to know what's going on. TikTok is a terrible way to do that. And now we know why, because of what we were just discussing in the deep dive. There's no human curation. The curation is being done by a blind machine learning algorithm, which is just building approximations of the systems and processes that help explain the patterns it sees. And that could have no correlation to things like truth or to civic value or to like inflaming darker impulses.

• 50:37Cal Newport?

  It doesn't care about any of that stuff. It's like the world's most reckless newspaper editor. You know, it's like a psychopathic newspaper editor who like really has like no empathy for humankind and doesn't really know what's going on and is drunk all the time and has some weird fetishes. It's, you know, it's like the worst possible curation is algorithmic curation. Well, they should keep up with news, but here's the simple rule. I want human curation because all of the things that we value, human values, human ethics, these normative standards and guardrails we've had since the Gutenberg printing press, those are all there, one degree or the other, sometimes more rigorous than other times, you know, sometimes more strict than other times, but they are there and unavoidable when it's human curation.

• 51:22Cal Newport?

  That could mean a lot of things. It could be like various newspapers. It could be websites that aggregate news, like people like All Sides, for example, which will be like, here's an issue, here's some publications on the left and on the right and some center and they're all covering it and you get a pretty good side. Podcasts, email newsletters, I think these are all great, right? The goal here, because right now I can hear everyone's concern. There's often this sort of tribal concern, no, no, no, but what matters is like bias because, you know, my side's right, the other side's wrong. And if they do that, they might be hearing from someone. I mean, everyone thinks that like their side is very neutral and fact-based and the other side is crazy. And so like, it's not just get information from a human, but like the right humans. But a lot of stuff is biased. That's not my concern right now. My concern right now is not biased.

• 52:06Cal Newport?

  My concern right now is not this person really doesn't like that person, you know, it's, you know, Jimmy Kimmel does not like Donald Trump. Greg Gutfeld does not like Joe Biden, right? It's, I'm not trying to get rid of, find neutrality. I just want humans in the loop. Let's just start there. You know, let's just start there that like, actually there is deep down some sort of culturally attuned, you know, normative ethical frame. Let's just start there. It could be completely biased or political content, but even that is much better than letting an algorithm make these selections. So yeah, teach your son ways to stay informed that it's a human making the ultimate decision and talking. And then you could teach them about bias, like, look, people have different points of

• 52:50Cal Newport?

  view and different goals, right? And you could teach them about that. And I'm a big believer in dialectic content consumption. This would be a great thing to teach a kid is like, if they're really into like this type of news source, like, I don't know, it's like a young man who really likes listening to like Joe Rogan to hear about things, right? To really connect with him, that's fine. But be like, look, let's take this issue. Hey, you heard this issue on Joe's podcast. You're kind of worked up about it. That's good. You're involved. You're you care about things. You know, here's another person who maybe comes at this a different way like let's, let's listen to both. And when those things collide, like, oh, the world's complicated and interesting and all these sort of great things. None of that happens if it's just algorithms. So anyways, stay informed, but have human in the loop. Let's start there. Then we can solve the problem

• 53:29Cal Newport?

  of all humans. Humans are super flawed. But again, I'll take a flawed biased political hack any day over the combination of transformers and neural networks that goes into the user tower and a TikTok recommendation formula. All right. What else do we got?

• 53:48Speaker 03

  Next up is Sean. You often discuss the architectural limits of LLMs. One way to overcome these limits and those of traditional computer processing is quantum computing. Is this a more promising route towards a phase shift in AI and even superintelligence?

• 54:03Cal Newport?

  All right. So first of all, why am I placing this question in today's episode? Because today's episode is about social media algorithms. It's because what I want to do here is zoom out slightly on an application of the same underlying principle. So what was the underlying principle that's driving today's conversation? It's a principle that's at the core of my work as a digital ethicist, which is you have to understand technologies and how they work before you can start to make decisions about the role of those technologies in your life or in civic society. The more you actually understand about the core of a technologies operation, the more effectively you can bring whatever humanistic value matters at bear to figure out what to do about that technology. So this is sort of a broader example of that. There's a lot going on with AI where the underlying technologies are complicated.

• 54:49Cal Newport?

  And when we don't try to understand at least a little bit about what's really going on, we can get pulled left and right into all sorts of different corners. So here's a great example, because several people, I don't know where people are seeing this, Jesse, but several people. In the last week, I'm thinking about one example from yesterday. I've been asking me about quantum. Quantum and AI, so there must be something going on.

• 55:10Speaker 03

  Several people in Georgetown or some people just on the street?

• 55:12Cal Newport?

  No, people I know.

• 55:13Speaker 03

  Okay.

• 55:13Cal Newport?

  Like family members or friends. So there must be, I'm sure I'm missing some sort of like big thing that went viral somewhere. But there's this idea out there like, well, quantum computing. That's what's going to unlock super intelligence because you know, these party poopers like me wrote these articles a couple, you know, a month ago that was unfortunately telling people, hey, guys, these language models are not just going to keep scaling till super intelligence by 2027. Like they're not, they've hit it, they've hit a wall in scaling. The technology isn't there anymore and the companies are stuck. And a lot of people recognize this and it was kind of a disappointment for those who are betting their whole sort of self-interest and worth and, and you know, their whole philosophy of life on the idea of a cataclysm, a technology cataclysm, either positive or negative.

• 55:58Cal Newport?

  So they need to get that hype back alive. And so they got it back alive by saying like, well, quantum computers can do crazy stuff that regular computers can't. So maybe quantum combined with AI will break through these scaling limits and we'll have something like super intelligence. All right. So bad news, again, I hate to be the party pooper about this. That's not going to happen. Quantum computers are not, as a lot of people think, just like a super powerful version of a regular computer. But you can take anything you could do with a regular computer and now do it in a much more super powerful way. If you move it to a quantum computer, it's like going from a 386 chip to a Pentium chip. If you appreciate my 1990s Intel references. But that's not how quantum computers work.

• 56:41Cal Newport?

  Quantum algorithms are a mix of computer science, logic and actual quantum mechanics. It's where you very carefully make connections between these quantum bits, right? In such a way that when they, the wave functions collapse, it has the equivalent of searching a large search space that collapsed to an answer so that you can, you can extract an answer from a search space all at once in a sort of constant time collapse, as opposed to having to do a sort of linear search of the whole search space, which could be computationally infeasible. But there's only certain problems that you can set up quantum algorithms to work for. So Peter Schor, who was, you know, at MIT when I was there, he had this big breakthrough where like, hey, factoring large prime numbers, you could do this way. And that matters.

• 57:26Cal Newport?

  Like not to get too technical, but RSA public key encryption is sort of the core way that we do key exchange for, say, secure commerce on the internet is based on having really large prime factors that are so large, you can't easily figure out what they are. So if you could figure out prime factors quickly, that would be a problem for a lot of cryptography. That's something you can do with quantum computers. Once you have enough quantum bits. There's also a certain type of search you can do. A different way of searching the search space that's really useful, especially for like simulating a lot of like chemical or physical systems and quantum computers could then do that in a way, do these sort of like physical simulations in a way that would be hard for a regular computer to do. But these are kind of the big use cases right now. So no, there's no obvious connection between this narrow things you can do right now with quantum and what we

• 58:11Cal Newport?

  need to do to run like a generative AI model. And to make matters worse, it's incredibly complicated and expensive to try to make these quantum computers have enough bits to actually even do those problems on big enough input sizes, because the more quantum bits you have, the more error you produce and the errors add up. And we don't know how to, how to get to the many thousands of bits we need to handle like really big problems. And there's all sorts. They're working on it. It's a cool technology, but no, it's not coming anytime soon to suddenly break through the scaling barriers in generative AI. Now, yes, there are some like very specific things around gen AI where you could connect to one of these capabilities, like some sort of like, you know, it was the search through uncertainty with the sample space of things that the, the model is going to look at. Like

• 58:54Cal Newport?

  there's things in theory in some future, a quantum machine could help with, but no, like fundamental, we're going to break through barriers and what, you know, training these things or building these things. So no, quantum is not going to give us superintelligence. But the reason why I bring that up is when you understand that technology, it gives you a completely different valence on this recent emphasis on this issue. And it tells us, right? I think there's a great data point we suddenly learn as, as observers of technology and society. I think it is taking down the mask of a lot of people who are given a lot of content, online in particular, about all of these coming AI apocalypses.

• 59:32Cal Newport?

  The fact that they switched right away to like, well, quantum is going to do it says, oh, well, that's clearly nonsense. So I see all along, you're just working backwards from a psychological need to need some sort of massive technology-driven disruption to happen. And then you'll just try to fill in the blanks any way you can. It's like, when you understand quantum and know that that's kind of a crazy thing to say, it reveals a lot about the people who were saying things a year ago that seem much less crazy. Like, well, they know a lot about AI and they're saying that, you know, in 2027, we might have humanity extinguished by, by superintelligence. We were taking that seriously, but now we can realize with this pivot to quantum that some subset of those people, it really is just a fulfilling a psychological need of, of something disruptive,

• 1:00:15Cal Newport?

  because that will bring some sort of like, meaning or focus to their life. So there we go. Understanding the technology helps us better understand our cultural's interaction with technology. Quantum stuff is hard, Jesse.

• 1:00:28Speaker 03

  Yeah.

• 1:00:29Cal Newport?

  I had a couple of friends at MIT. So, so quick story. So if you're a computer science doctoral student at MIT, you have to have a minor. So you have to go take a few classes in something that's not computer science as part of being your doctorate. And I had one friend in particular, but a lot of people did this, I think. They're like, oh, I want to do quantum algorithms. So I'll take some courses in physics so that I can do, I'm, I'm a computer scientist at MIT. I know a lot about algorithms. I'll learn some quantum physics and then I can do quantum computing. And they completely failed. As it turns out.

• 1:00:58Speaker 03

  They actually failed?

• 1:00:59Cal Newport?

  Not failed in the sense that they failed the course, but they failed to make any sort of meaningful, to actually learn enough to do quantum computing work. Because it turns out, here's what I learned from them, to really work with quantum computers, you need to understand quantum mechanics. Can't just go take a course on quantum mechanics. You gotta like, learn all of physics and related math and end up at quantum mechanics. You basically have to, it's a multi-year, like, this is have to be what you're studying. It's like, okay, as I get towards like, year four of my doctoral program, I finally know enough stuff, enough math and enough traditional physics to start doing the quantum mechanics. You couldn't learn it in three classes.

• 1:01:40Speaker 03

  Kind of similar to what you were talking about, the theme with Brian Keating.

• 1:01:43Cal Newport?

  Yeah, like how long these things take. Yeah.

• 1:01:45Speaker 03

  And just focusing on one thing.

• 1:01:46Cal Newport?

  Yeah. Jesse's referencing, would have been a few days ago when they hear this, right?

• 1:01:50Speaker 03

  Yeah, last Thursday.

• 1:01:51Cal Newport?

  My episode with Brian Keating, who knows a lot about this type of math, but as you'll see.

• 1:01:56Cal Newport?

  College and multiple postdocs before it took him a long time to really learn all this stuff. So anyways, as an aside, quantum computing is really hard and really cool and it's going to do some cool stuff, but it is not a miracle machine. Don't listen to anyone who says, because I bet you the same people are saying quantum is going to give us super intelligence or saying LLM scaling is going to give us super intelligence. And I would bet you go back four years and they'd be talking about blockchain based internet was going to be the future as well. That just, there is something exciting about massive disruption, but it's hard transformative technology is hard. Most stuff is more complicated than it seems. All right, we have a case study here. Let me just set this up real quick. So again, we're deviating a little bit from the theme, but not really. So like what I like to do with these case studies is talk about people

• 1:02:38Cal Newport?

  who have used the type of advice we talked about on the show successfully in their own life. Right? So what is the relevant advice on this show that would match what we're talking about today, which is about how these algorithms work and this sort of the dangers of algorithmic curation? Well, one of the big ideas I talk about is that an effective way to not have your life be captured by the digital, especially by the algorithmically curated world on your phone is to make your life outside of your phone so deep that the stuff on your phone becomes a lot less interesting. If your life is weird and shallow and chaotic and uncertain, you're like, I might as well number distract myself with the phone. But when you construct,

• 1:03:18Cal Newport?

  intentionally construct a life outside that on your own terms that's deep, then suddenly becomes a lot easier to step away from things like the Tik Tok curation algorithm. So that's the case that I'm going to, I'm going to give you today is one about, as many of my case studies are, someone who figured out through trial and error, how to actually start engineering a deep life. It's not specifically about technology, but it's also all about technology, because when your life is deeper, you have much more say over what you let into your life and not. Let's do this case study. Do we have case study music? Yeah. Now let's hear that. We need some music.

• 1:04:05Cal Newport?

  All right, here we go. Today's case study is from Kiran. Kiran says, "A friend of mine introduced me to your work in 2021, and ever since then, I've been a voracious reader of your writings and an evangelist of all things New Portonian. I've always been a bit of a Renaissance man, and within a few years of getting my master's degree in 1997, I had a pretty cool, good double career going in classical music composition and computer consulting. The IT gigs paid better than the composing, so I leaned mostly into computer work, taking semi-regular commissions to keep my artistic soul happy. I had no conscious plan. I just stumbled unintentionally through an enjoyable and lucrative 15-year stretch of contract-based self-employment. In 2012, just a few years after my wife and I bought a house and had our first child, I decided I was done with the dual-career

• 1:04:50Cal Newport?

  thing. Finally deciding to follow my passion, I set out to become the great composer," I should say Kiran capitalized "great" and composer here, "that I believed I was destined to be. I told all my computer clients I was closing up shop, wound down all those projects, and cast my net for bigger and bigger music commissions. Long story short, that led to a steep decline in income, and as the primary breadwinner with a family and a big Toronto mortgage to pay for, I fell into a rapid downward spiral that left me completely burnt out. It took me a full year to recover health-wise and another two years to rebuild my networks and regain the trust I had broken in every facet of my professional life. I then vowed to work intentionally towards being happy and living the lifestyle I wanted,

• 1:05:33Cal Newport?

  rather than focusing on any particular aspect of my career. Today I have a truly fantastic groove going. Most days I punch out at 3:00 PM, whether in computers or music. I spend the vast majority of my time in deep mode, and I'm producing the highest-quality work of my life. I collaborate almost exclusively with people who let me work at my own pace. My income is the highest it's ever been, and I have a lot of quality time and energy left over to spend with my family." As a great example of some core ideas of lifestyle-centric planning, which is at the core of my theories about how to cultivate a deep life. So what Karen did first is what most people think you need to do to build a deep life, which is have some sort of radical change. The way we think about it is the magnitude of the radicalness of my change directly corresponds to the magnitude of how much better

• 1:06:18Cal Newport?

  my life will be. So if I want my life to be five times better, I need a five times more ambitious radical change I'm going to make. That often doesn't work. And why? Because what matters in your day-to-day satisfaction is not any one factor. It's many, many different factors that affect your day-to-day subjective well-being. I call those many, many factors, your lifestyle. So if you go for one radical change, often what you're doing is a combination of getting the short-term fix of just doing something radical feels good in the moment, which is true. There's nothing to sneeze at, it's true. And then you're taking at best like one aspect of your life that's important to you and making it bigger. But you're ignoring all the other aspects of your lifestyle.

• 1:07:00Cal Newport?

  So either those stay the same, or what's more likely is you make this one aspect of your lifestyle better, but as a result, these other aspects get worse. And when you add it all up, you're worse than when you began. And that's what happened to Karen. He's like, I like this inspiration of writing. I get a rush thinking about the idea of being a full-time composer. And I like doing creative work and pushing myself. So let me take that aspect of my lifestyle, put all my chips on the table there. The problem is it made these other aspects, like your financial security, time with your family, you know, these other aspects, your stress, your health, they all got worse. And you know what? That added up the net negative. So one radical change very rarely makes all the difference. So what he did instead was essentially lifestyle-centric planning, where you look

• 1:07:43Cal Newport?

  at your whole lifestyle, and you systematically make changes, you assess them how they're going to affect all aspects of your lifestyle, and you're looking to move everything up, or at the very least, not bring anything down. And when he looked at his whole lifestyle, he's like, oh, if I keep my computer consulting, I can still do music. And if I get better at my computer consulting, instead of taking on more work, I can raise my rates, and now I can make more money and not have to spend, can spend less time, he's done by three, and I can bump up this need for creative pursuits and improving my craft, I can bump that up. Like I can build up my ambition for the composition I'm doing. Now, I'm not bumping that up as much as just going full-time to be a composer, but I can still improve

• 1:08:25Cal Newport?

  that while keeping these other things higher, better as well. And in the end, he had a much, much better. Lifestyle. His life became deep because he cared about all aspects of his lifestyle, not just one. All right. So again, what does that have to do with TikTok? Kind of everything. Because once you take control of your life and make your life deeper, what's going to happen is the attractions of the diversions, the digital, the attention merchants wears is not going to be as impressive for you anymore. And it's going to be much easier to make these sort of intentional, those are intentional decisions. So there you go. Karen, thanks for that case study. Let's see, we often try to do calls when we can as well. Do we have a call this week?

• 1:09:01Speaker 03

  We do.

• 1:09:02Cal Newport?

  All right, let's hear this.

• 1:09:06Speaker 04

  Hey, Cal, my name's Steven. We met last year, actually, when I brought you in for a recording on my own podcast, Two Dads, One Car. And thank you so much for that wonderful conversation. Since our chat, many of my recent guests have talked about how their kids need us even more as they get older and head into the high school years. And I think you shared that as well. Have you adopted any new practices recently to accommodate this extra pull on your time from your kids on top of your role at Georgetown? And of course, writing. So let me know. I continue to be a huge fan. Thank you so much, Cal. It means a lot. Cheers.

• 1:09:41Cal Newport?

  Oh, it's good to hear from Steve. That was a cool show. I think that episode's maybe on YouTube. It's like, remember comedians in cars getting coffee?

• 1:09:49Speaker 03

  Yeah, Seinfeld.

• 1:09:49Cal Newport?

  Yeah. So this guy does something similar. We drove around Tacoma Park, and he sets up cameras in the car. And they kind of interview you. He came down, I think, from Canada or something. Cool interview. So you should find that, Two Dads and a Car. It's somewhere, I don't know, on YouTube or something like that. Okay, a good question. I have a few thoughts about it, including thoughts that connect us back directly to the technology themes we're talking about, as well as some broader thoughts as well. So I have three boys. One of the things I'm, here's one of the changes I'm making to account for my boys are all in the age now where like dad time really matters. They're not really young, and they're not out of the house yet, and like completely just with their friends.

• 1:10:28Cal Newport?

  They're my hobby. And what I mean by that, that sounds trite, but there's actually something deep there. That is the priority.

• 1:10:36Cal Newport?

  Over other things I might be interested in doing like there is a phase of life if you have multiple kids where the priority you know you have work you have to do the you know provide for the family but you keep that reasonable you need to keep yourself healthy and then it's your kids right I'm coaching three things right now Jesse I coach all the time hours and hours I am I am sore as we record this because me and the other coach of one of my kids little league teams did a two and a half hour practice the other day and when you're my age two and a half hours is a long time to be like hitting ground balls and catching balls and throwing balls like I am sore my hamstrings are sore as I sit here I spend a lot of time I'm on the board of my kids school I'm giving a talk there next week a lot of time with my kids doing things with

• 1:11:21Cal Newport?

  my kids because you know that has to be that is my hobby and that's what I mean by my hobby is like that's actually my priority when it comes to spending time and doing things outside of work and then that'll change again as you get older and your kids move on you gotta fill that in with other things or you know things get sad before that point you have to have other things going on as well but that's the main thing I'm doing now is like that's priority one outside of the fundamentals of just making sure the lights are on and I'm not having repeated heart attacks but I think that matters now what do you do in that time well I think it's really important if we're talking about like kids coming up to high school age do not let them be exposed to algorithmically curated values do not let them be on with unrestricted access to the internet do not let them be

• 1:12:04Cal Newport?

  on things like tiktok do not let them be on things like instagram they're at an age where they need these sort of human values that those algorithms do not have they need repeated examples of those values applied and the only way to do that is to be there with them doing things with them let me tell you about this thing we just did and what I noticed there here's why I like this person here's why I don't like this person I really appreciate I think that person is really impressive this person is this book is really good this book is not what are you getting out of this but all of those type of discussions when you're doing things involved in your kids lives are your chance for helping them learn and understand the normative principles by which you live you have to implant those values in them they do not get those values from their phone if you give them a phone

• 1:12:46Cal Newport?

  that becomes their main source of interaction with the world you have put them into a an upside down world in which it's all there is no humanistic core

• 1:12:58Cal Newport?

  Don't put them in that world. They might come out of it okay. Most kids do, but a lot of kids don't. We know all about, we talked about the slope of terribleness leads you to all sorts of weird or bad places. So to me, that's like another key principle. I want to help be their filter and exposure to their world or the interpreter of what's going on in the world. What's interesting, what's not, what's valuable, what's not, what's heroic, what's not, who's impressive and who's not. I want to be that filter, not a two-tier, two-tower recommendation system running within Oracle's cloud infrastructure for TikTok. That doesn't have my values in mind. There's no values that I want my kid to have. So I think the technology issue here is key. You should be the curation algorithm in your kids' lives, at least in, you know, especially through a middle school and like upper elementary school.

• 1:13:40Cal Newport?

  Not technology, not these algorithms, right? It is not harmless. It is imprinting in their mind, what values, what matters and what doesn't. And the things that is imprinting is almost certainly not going to be the things you want. So there you go. I connected it back to TikTok. Yep. All right. We got a final segment here. I've got some articles. I haven't even read these yet, Jesse. It's gonna be interesting, right? We got some articles from like the last week about TikTok, so we can better understand like what actually is happening. How are people using this right now? It's not gonna be pretty, but before we get there, let me take another quick break to talk about a sponsor.

• 1:14:17Cal Newport?

  The holidays are approaching, Halloween, Thanksgiving, Hanukkah, Christmas, all of them. You can get what you need to personalize your home for these holidays with Wayfair. We're talking, you know, decorations and inflatables, but also like stuff inside, like accent pillows. Or like maybe you're hosting some sort of big dinner for these various holidays. And you can make your holiday events a breeze with quality cookware that will wow any guest, et cetera. As longtime listeners know, my favorite Halloween to decorate for is Halloween. And Wayfair has been one of my go-to sources for the props and decorations and interior accents I need to make my grand schemes for Halloween haunts work. True story, I always look each year like, hey, what is, what are the new props that Wayfair has?

• 1:15:02Cal Newport?

  This is true, Jesse. They introduced this year a new prop that, man, I could have used last year. So, you know, I did a UFO themed haunt last year. And one of the hard things we had was actually trying to build a good, like, the crash UFO at the beginning scene of our narrative scenes. Well, Wayfair, I just discovered, they have a new UFO themed prop that's a nine-foot inflatable UFO. So it's nine feet tall, and the UFO's up top, and then there's like a abduction, you know, the light coming down from it. And then at the bottom, there's like someone down the bottom that's about to be pulled up into the UFO. Oh, would have been perfect. Would have been perfect. So, you know, again, Wayfair knows what they're doing. Now, here's the thing. Wayfair's huge selection of home items in every style makes it easy to find exactly what's right for you. And their prices are

• 1:15:47Cal Newport?

  always fair. Like, whatever budget you're looking for, whether it's like, hey, I just want like a, um, something on the lower end for use like in a, like a kid's party, or like a higher end thing that's going to be like an impressive piece of furniture that you'll have in your house for, for, you know, years to come. They have what you need at whatever budget point you're looking for. And with free and easy delivery, even for big stuff like tables or sofas, there's no limit to what you can get from Wayfair. So get organized, refreshed, and ready for the holidays for way less. Head to wayfair.com right now to shop all things home. That's W A Y F A I R dot com. Wayfair, every style, every home. I also want to talk about a, an app that I think is really smart. It's

• 1:16:29Cal Newport?

  called Headway. Headway is a daily app that keeps your brain sharp with key ideas from the world's best nonfiction. It has 1,800 summaries across more than 30 nonfiction categories, from productivities and happiness to money and beyond. With Headway, each session you do takes 15 minutes, making it easy to stay mentally active, whether you're commuting or taking a break, or it's like you're switching from one task over to another. So it's made for people who want to keep learning and evolving. Even on busy days, you can think of Headway as your personal growth guide, right? A way to take in useful information in an otherwise busy life, give you personalized recommendations that can expose you to new ideas that make a difference, and all done with

• 1:17:14Cal Newport?

  an award-winning design. So it's like very easy to use. It is the light to use this app. Headway has over 50 million downloads and 2 million monthly users. We kind of know it works. It's ranked number one in education on the US App Store and frequently featured as the app of the day by both App and Google. So a lot of people have already discovered Headway's power to help you grow some of the world's best ideas in just 15 minutes at a time. So go to makeheadway.com/deep to find out more and use the promo code DEEP.

• 1:17:53Cal Newport?

  That's makeheadway.com/deep and use the promo code DEEP. Grow your way with Headway. All right, Jesse, let's move on to our final segment. All right, God help us. I got three articles here all about TikTok that I haven't really looked at. I'm assuming, Jesse, these are all going to be uplifting and inspiring, maybe a little bit intellectually sophisticated, but will come away saying kudos to plaudits to the American ingenuity. It's gonna make us just feel really good about culture. That's what I'm assuming. It's gonna be something like TikTok helps schoolchildren understand the beauty of Jane Austen, leading to impromptu presentation of Emma that wows critics. Like, this is the type of thing I think we're gonna

• 1:18:36Cal Newport?

  find. So let's see what we got here. I'm gonna load up the first article here now. I'll put it on the screen for those who are watching instead of listening. TikTok child data protection inadequate, Canadian privacy officials say. Oh, God, we're not off to a good start here. All right, let's see here. TikTok's efforts to stop children using the app and protect their personal data have been inadequate, a Canadian investigation has found. Hundreds of thousands of children in the country use TikTok each year despite the firm saying it's not intended for people under the age of 13. Oh, God, the kids are on TikTok. So we have this like algorithmic, value-free curation, and we're showing it to kids under 13. I do have a solution to this problem. Don't let your kids have a phone.

• 1:19:18Cal Newport?

  Yeah, okay, TikTok's not preventing them from going on. You know what can prevent them from going on? Not letting an 11-year-old have access to the internet. That's nuts. That's the solution. All right, here we go. This one. Oh, God. I thought this one was gonna make us feel better about the human condition. All right, what do we got here? Why people on TikTok think the rapture is coming as the world is ending this week? Oh, man. This article's from yesterday, Jesse. Oh, my God. All right, I'm sure this is all gonna be positive. The rapture is here. That is according to worshippers on TikTok. People are convinced that the end of days will take place between today and tomorrow, tomorrow being the day we're recording this. So if you're listening to this, phew, crisis averted.

• 1:20:03Cal Newport?

  They have decided to prepare accordingly. The phenomenon stems from the evangelical Christian belief that Jesus Christ will return to earth and ascend to heaven with only his true believers. Yes, the rapture. So it's a South African pastor claimed on a YouTube video, you know, that Jesus told him that this is when the rapture was coming, but TikTok super amplified this. It's a corner of TikTok called Rapture Talk, where users are posting videos of themselves either discussing or preparing for the end of the world. Christians on TikTok staunchly believe that chaos will erupt on earth when they're swept up in the clouds of Jesus Christ, leaving earth, according to Christians on TikTok, is a glorious occasion and a great honor. Okay, a couple of points. Don't say Christians on TikTok. Say it's the people on Rapture Talk.

• 1:20:46Cal Newport?

  But this is a bigger point about what I was talking about. Why is there no, like, reasonable, human-curated news source would give a ton of print to just like this random clip from a random church? And be like, I don't know, some guy at some random church in South Africa said this, like, people say crazy stuff all the time. We better put this on the front page of our newspaper and put it on the news and be on the radio and the news. Like, people go, no, it's crazy. But TikTok, because it's just algorithms. The algorithm figures out, like, you know, this type of stuff does well with a lot of people, so it shows it to a lot of people. And then it gets into the trending thing. And once it's in the trending thing, it's going to show it to a lot of other people who might not have previously have shown their interest

• 1:21:28Cal Newport?

  in Rapture Talk stuff, and it discovers they are, and it takes off. That's algorithmic curation. It is crazy that this like particular little clip has gone wide, and it never would in a human curated concept, because how many people every day think, you know, have some prediction for when the earth is going to end. But this is what happens. We let algorithms do the curation. All right, we have one chance here, Jesse, to be uplifted by the possibilities of TikTok. Let's see here. Here we go. Teens facing criminal charges after friend dies during TikTok surfing stunt. Oh, God. One of the teenagers was standing on top of a table tied to a car when the crash took place. Yes, he is not good. All right. This is not

• 1:22:12Cal Newport?

  making me feel good about TikTok's impact on the world. Let's, let's hear a little bit about this. The incident took place in March when someone was driving at 35 miles per hour with her friend standing on the trunk and rear windshield of her car. The friend fell off. When you fall off things at 35 miles per hour on pavement, you can hit your head, and that's what happened, and the friend died. And then this has led to other, it looks like this has led to other incidents of this. Why? Because the algorithm used to do recommendations, you know, is capturing some sort of like fundamental human tendency to like, we like stunts that are outrageous and attention catching and potentially dangerous, and it catches our attention in a way that they get good watch time. And

• 1:22:56Cal Newport?

  so it promoted it in a way that like you would never see this. You would never see this. You can probably bring this back. You would never see this again on a television show, in a newspaper. You wouldn't see people dedicating their podcasts like week after week. How cool is this? Look at this person, this, these kids standing on cars. Like, that's crazy and it's dangerous. All of this is, I think, emphasizing the point that we got to in the beginning. When you're willing to concede or give over these issues of curation of information to a mindless algorithm that really has no values in place, it leads to dark places, it leads to weird places, it leads to people dying or weird news obsessing people or captures

• 1:23:38Cal Newport?

  kids because, you know, how are they not going to look at that? This thing is completely hacking their brain stem. So we end here, like, where do we end this, this look at the algorithms today? Algorithmic curation, and I got at this last week as well, is not a good innovation. It really is only good if you have like a large stock position in one of these companies. I promise you as a user of these tools, there are other things out there that can fill your life with interest and entertainment and funniness and new ideas. You do not, there's tons of sources of that, that are probably going to be better than watching kids standing on tables pulled by cars, leading them to fall and have, you know, life-ending brain injuries. There's better ways to be entertained. So there's

• 1:24:22Cal Newport?

  really no reason for these things to have such a big cultural position, other than the fact that like, they're very good at it and a small amount of people make a lot of money. So there we go. The algorithm is not some Frankenstein construction in which people put the wrong values into it, which we can fix if the right people run it. Algorithms, by definition, when it comes to recommendation systems, do not share our values, they do not know what values are, and to let them curate information on any sort of mass scale is a recipe for disaster. That is the solution to all these issues, not putting the right people in charge of algorithms, but allowing real people to say, I don't want algorithms making these decisions for me. Anyways, that's all the time we have for today. Thank you for listening. We'll be back next week with another episode. And until then, as always, stay deep.

• 1:25:12Cal Newport?

  Hi, it's Cal here. One more thing before you go. If you like the Deep Questions podcast, you will love my email newsletter, which you can sign up for at calnewport.com. Each week I send out a new essay about the theory or practice of living deeply. I've been writing this newsletter since 2007, and over 70,000 subscribers get it sent to their inboxes each week. So if you are serious about resisting the forces of distraction and shallowness that afflict our world, you gotta sign up for my newsletter at calnewport.com and get some deep wisdom delivered to your inbox each week.