Sharp Tech with Ben Thompson - Evaluating AI Risk and Entering an Era of Incredible Unknowns

Episode Date: March 27, 2023

A programming note regarding the next few weeks of the podcast, a long email from a Stratechery reader about the risks of AI, and extended thoughts on the various risks and unknowns that define the AI conversation in the present moment.

Hello and welcome back to another episode of Sharp Tech. I'm Andrew Sharp and on the other line, Ben Thompson. Ben, how you doing? Hello, Andrew, and welcome back to the land of the living. By all accounts, it was a bit touch and go last week. Yes, I was out with food poisoning for the second half of the week. Do not recommend that experience to anyone. It was the second time.

No, no, I do have recommendations around food poisoning. The important thing if you have food poisoning is to, yes, I recognize the thing. You need to throw up as soon as possible because it's horrible and awful. But if it passes through and that's how you get over food poisoning, then the horribleness and awfulness lasts like three to four times as long. Whereas, you know, I actually have a very funny story around food poisoning. I got it at the Yassahunis Shrine. Do you know what that is?

No. So the Yassahunis Shrine. Shrine is like this the Shinto Shrine in in Tokyo, which is where it's very controversial. Every time like a Japanese prime minister goes there, like China gets all upset because that's like where like they're depending on your perspective, war heroes or like war criminals or whatever are sort of enshrined. And so you have heard of it before. You're not sure what it was.

So I show up there and I'm an American, you know, obviously. And I'll be with Japanese heritage, but it's hard to tell. And, you know, I have just the most tasty bit of sort of barbecue beef while I'm there. Big mistake. So I'm like, this is when I first come to Taiwan. So I'm very poor. I'm staying in this like little hostile with or hostel, hostel, hostel, hostel, hostel, yeah.

You know, just random foreigners and stuff like that. And I'm just getting worse. I'm just hurting. I'm just feeling worse and worse and worse. And finally, I'm, I'm on the subway and I think I'm staying your way or something like that. And just this movement of the subway, it's just really, really bad. I get off the subway and for some reason, I don't know why there's a sink on the platform.

I'm not sure what the sink was doing there. I get up and I just wretch. But unfortunately, that was insufficient puking. It sort of solved the issue for then. So I stumble back to the hostel. Keep in mind, I'm having to share this room with other people like laying on the ground. Yeah, that's like the worst possible place to have food poisoning. I really feel for your roommates in that situation.

I feel for everyone in my general vicinity, including myself. And so finally, I'm there for hours, just really, really bad, miserable. And finally, suddenly it starts welling up. I got to throw up. So again, I have to go for the sink again. Oh, God. And I remember the toilet, or no, the toilet was occupied or whatever it was. So I'd throw up probably as hard as ever thrown up in my life.

and in the midst of the throw up, up comes a big chunk of meat, which did not feel great coming up. It felt much better going down than going up. But the crazy thing is, clearly that was the cause of the food poisoning. And I was instantly better. And I felt great.

And someday I was super hungry. So I go across the street like a Denny's and I get like pancakes. I'm like, fantastic. So yeah, that's my, I will keep that in mind.

Next time I'll remember Ben in Japan and try to sort of pull the trigger early preempt the misery and you know anything would have been better than like the three or four days I had in the middle of last week but now I'm here I'm ready to pod I'm feeling great and last week was sort of a prelude to the next couple of weeks of Sharp Tech when we're going to be playing it by ear a little bit schedule wise you have your annual spring break vacation so that's coming up next week we are going to have

one episode. So listeners, if you have any fun questions for us, because Ben's going to be recording like hours before he leaves for his trip. So try to keep it light. But if you have any sort of off the wall stuff, you want us to hit, email at sharp tech.fm. And then in other news, yes, my wife is going to be having a baby sometime in the next couple of weeks. Congratulations. That's very exciting. I mean, I know you sort of previewed it on Instagram by like suddenly sneaking in a couple It was a soft launch. A soft launch. Yes, that's the word I'm looking for. But I think this is the official podcast announcement, which, you know, it's kind of the downside of now being like a public figure. You have to tell people, you know, what's going on in your life and why you're not showing up for podcasts. Don't be concerned. That doesn't involve big chunks of meat. No, it actually does involve a big chunk of meat. Yeah. Yeah. It's true. It's a different. Well, and look, I'm not going to be miserable over the next few weeks.

Who knows where my wife falls on the misery spectrum. But we're very, very excited. And we had a little dress rehearsal this weekend where some tests came back in conclusive. So we were called into the hospital. And now I have like the to go bag ready. And I'm ready to rock and roll on that whole process. But sometime over the next few weeks, there's going to be a baby entering the

sharp household. And we'll have to sort of adjust. just the schedule for a few weeks. But we'll keep it rolling. After mid-April, this will get back to normal. Hopefully. We'll take it as it goes. We do post updates on Twitter at Sharp TechPod.

So like, for example, we noted last week that, oh, there's no episode. Now we know why we've told the gory details. Actually, no, we told my gory details. I was going to say, we got a great Ben Thompson story out of it. So everybody can imagine. Ben in the hostel. How many years ago was that? Was this like college era? This was 2004. So, yeah, like my first trip when I came in Taiwan was definitely, I'd always wanted to visit Japan. So I was super excited to go and got a real highlight along the way.

There you go. Well, with that programming note out of the way, should we dive in here, Ben? Do you have anything to add as far as the schedules are concerned? No, I think we spent a lot more time than anticipated on it already, but yes, just please bear with this over the next few weeks. Okay, so we'll start with a long note from Jack. He says, Ben, I'm a longtime reader, huge fan, recent convert to Sharp Tech, yada, yada. Your history is an inspiration to me and your frameworks are pretty foundational in how I think about the world. And I write in because I've got an important question for you. You've mentioned a few times over the last year or so that as tech evolves to support it,

you see an evolution of most media from walled garden content creation to UGC based on your network, to UGC filtered for you by AI, to an end state of AI generated content filtered specifically for you. You've also written eloquently about chat GPT and its value as a companion, in addition to its more B-to-B use cases. Last week's post on Srotectary on lobotomized lovers talked about this. And so here's why all of that scares me. You can model the last 100 years as, quote, we get increasingly good at synthetically triggering our own pleasure centers.

Look at food science, pharma, and of course the aforementioned trend in software. Facebook is addictive, TikTok is more so, etc. We're so good at making ourselves feel good that we stopped doing the things that evolution wired us to pursue, parentheses, sex, exercise, organic friendships, even sleep. You've quoted the famous Reed Hastings line a few times. A lot of the bowling alone problem, increasing suicide, addiction, fewer friendships, less resilient people, deaths of despair, etc., is directly traceable to this trend. So listening to one of your podcasts last summer made me stop in my tracks and leave my brother a four-minute voicemail about how I kind of see this AI development as the end of the world. I'm not a Luddite and I see how AI could massively improve almost every aspect of society. But that said, it's hard for me to picture a future where most people don't spend most of their waking lives wired into some XR rig interacting mostly with AI generated.

NPCs and experiencing unbelievably potent stimuli. You know how the real world apparently looks less colorful after you play Candy Crush or whatever? I feel like all of humanity will view the whole, quote, being human thing as basically that colorless world after experiencing the alternative. Ben, you're obviously a tech optimist, but you're also a sharp observer of cultural trends. What do you make of all this? What do you make of all this? and why are you so sanguine about AI and its impact on society? Well, first off, I don't know that I'm a tech optimist per se or sanguine. I think it's just the reality of the way that I seek to describe the world and work in terms of

techery or sharp tech is I view my role as analyzing and describing what I see going on. And it's the role of Jack and whoever else listens to interpret that. and sort of act sort of accordingly. And I think there is a tendency sometimes amongst those that do analysis to be overly prescriptive. The reality is there's a lot I don't know, including what appropriate policy should be or XYZ, but I will give you my view on sort of what's going on and you can act accordingly or interpret that as it might be, broadly speaking. That said, Jack does raise a good point.

That does make me sort of a tech optimist, which is if anyone left me a four-minute voicemail, I would be praying for AI to come along and sort of transcribe that and summarize it for me. So just on that point, thank you for the excellent example of how AI can make our lives better. Now, I'm only slightly joking. I do think the question of sort of existential risk in terms of AI, I think, is an interesting one. It's been an ongoing sort of discussion. It's one that is going to become even louder and noisier. And it's one that, unfortunately, I think, is going to be fairly unproductive for a few different reasons.

First and foremost, and this is something I haven't quite fully written about, but I hope too soon. I do think a lot of discussion around AI is misguided precisely because we have the wrong mental model of how these large language model works. And there's a line that I've said frequently, which is they're not deterministic, they're probabilistic. And I think that's something that's really important to flesh out and understand, which is these are, you know, we've talked about the problem of hallucinations, right? And like where the AI just sort of makes stuff up. And I think there's an aspect to hallucinations. And I think I've used this line before, either on trajectory or on a podcast, I can't remember. But there is a view of that where that's a feature, not a bug.

Right. And what I mean is there is you think about anything that's computer operated and you think about a digital operated and you think about a, deterministically, which is it's sort of like, you have like the paperclip problem, right? Which is you tell the, oh, I want some paper clips. It goes out of control. It makes the entire world into paper clips. That's assuming deterministic behavior on the part of sort of the AI, where it just sort of you give it the wrong instruction.

And because computers perfectly follow instructions, it follows that instruction all the way to its ultimate end in the doom of humanity. And that's just not how these work. And I think that's actually. like what we see with the hallucination is not that I'm not saying that lying the computer lying to is a good thing. Obviously it's it's not. But it's not a deterministic entity that perfectly follows instructions and does XYZ. It's not like the computers we've come to know over the last 40 years. It's not at all. It's totally different. And there's a lot of aspect. Once you really dive into how sort of these language models works, it's surprisingly similar to. to how the human brain works. And where the human brain, there's a lot of sort of like,

it's really about pattern recognition, and there's a learning process of acquiring this recognition and applying it to all sorts of data that's flowing in. And sort of predicting what's coming along next, and you have something like, say, those examples of the visual things

that sort of screw you up, optical illusions, right? What those are doing is their message, with your prediction functionality, where your brain is expecting something and the actual visual is different than that. And that's what produces the optical illusion. The mechanism is actually surprisingly similar to the way we work. And there's this idea of like, oh, of all the possible things that an AI could be, it somehow ended up in broadly this is a similar functionality to the way humans work.

Now, again, this isn't to say that this is a good thing or a bad thing per se. It's to say that when you're thinking about the risk and thinking about how these things are going to operate, that risk in operation is somewhat in line with human risk in operation. But it's just like it's like way smarter, but it's not like we're going to set off this thing that is like this uncontrolled reaction. It's going to do irrational things because it has an intuition. That intuition, it's not human intuition per se, but it has a broad sense of the way things work and out to work. And so, again, all this is not to dismiss the question of existential risk, and we're dealing with very early versions of these things. I do think the general model is generalizable beyond language. I think language is language to these things is sort of like electrical impulses to the brain, right?

Like, it's the means by which these operate, but it's not necessarily that they're restricted to just language. If you think about it as sort of prediction entities, that sort of applies to the image general. as well, which sort of operate in different principle, but it's the same sort of idea that, again, this isn't to dismiss the risk, but I think the way some people think about it is overly anchored in a deterministic sort of view of the world. And a lot of these folks, assuming very prominent ones, made very strong predictions that this model of AI would not work well. And they were wrong. And they haven't updated their own predictions about doom and bloom. And I think that, you know, it's important when you're looking at people that make very strong statements, do they change

their position when new sort of data comes in that falsifies some of their assumptions previously? And I think some of the most prominent AI Doom folks have not done that. And I think that's something that that's important to sort of keep in mind. Now, that said, my concerns are very much actually aligned with Jack. I think there's significant concerns that these are so compelling and so interesting to talk to that just as we do see early signs with the weakest form of pleasure producing content today that people will lose themselves in it, this is going to like that on not just steroids, just astronomically stronger. And the risk is not that it walks around killing humans, that humans willingly remove themselves from day to day life because this interacting with Sydney is so much more compelling.

Now, from a pure evolutionary survival of humanity perspective, that's not necessarily a problem because the AI is probably, you know, like the bit to get to robotics is going to take, I think, longer than people think dealing with the physical world is more difficult. But the idea that we're going to be able to operate an economy on potentially fewer people in sort of the workforce, because we have this super intelligence doing lots of tasks, we may be able to survive. that, but that's a broader question of is that a sort of society that we want to have and sort of live in in general? And so I am concerned about risk much more in the line of what Jack is, as opposed to the we're all going to die. Well, that's exactly why I wanted to read it, because there's all this back and forth about AI risk and there's fears about wrecking the white collar economy and, you know, clipping 80% of jobs across offices and AI taking over the world, starting nuclear wars and all that

stuff. And most of it rings hollow to me, at least in relative terms, relative to some of the hysteria. But Jack does a really good job articulating what a more realistic risk looks like. And I think in terms of addressing it, one of the big challenges that I see in addressing how this tech works in the future is coming to a better collective understanding of how technology works now. Like reading Jack's email, a lot of the concerns make sense to me because like you, I've thought a lot about society's relationship to technology and how this is affecting everybody. So I'm nodding my head when he's saying we've gotten increasingly good at synthetically

triggering our own pleasure centers. and it comes at the expense of engaging in some of the life-sustaining behaviors that have worked for thousands of years. But I'm not sure there's totally been like a society-wide recognition that like an iPhone can be a drug of its own. Like we're really good at diagnosing text dangers when we're explaining how it affects our political enemies. But there hasn't really been like a reckoning in terms of how it affects everyone and how how tech can affect our own worldview.

Like there's still sort of a blind spot there. And I think understanding where we are now and how tech can affect us now is sort of critical to understanding what the actual risk is as AI supercharges all of this stuff. And it becomes that much more compelling in the next 10 to 20 to 30 years. Yeah. Well, I mean, number one, I think the timeline, the timeline is moving pretty fast. So 10, 20, 30 will see. I'm hoping.

Fingers crossed for my unborn son. Yeah, I mean, well, the, the, a couple points on that. Just as an aside, this question of like destruction of humanity, there is an aspect of that is certainly a tail risk that is out there. And there is, there is widespread disagreement about how large that is. But I think there is an aspect where this ball is rolling. And, you know, it's not going to be stopped. So this applies to my view of technology broadly. You can definitely make an argument that we'd be better if the internet never existed.

If we went back to X, Y, Z, all these sorts of things. My view is that's, it's too late. The cat's out of the bag. And so we actually need to push forward and get to a better place because there's no going back. Right. And I think there's an aspect of AI risk where, you know, maybe there's an analogy like China and Taiwan, right? Like you take like a TSM perspective. China attacking Taiwan is an existence.

risk. There's also an aspect where the cost is so astronomical, not just of it happening, but of hedging against it, that it's actually rational to just, let's assume that that doesn't happen. And Morris Chang actually gave a talk a couple weeks ago that he basically made this point explicitly. He's like, yeah, we just assume that that's not going to happen. And his point was that it's not to say that it's impossible that it happens. It's that TSM is, you know, not viable if you hedge for it not happening. So you might as well just act like it's not going to happen as far as your sort of general operations go. And I think there's a bit here like this. There certainly is a possibility that it's already over. But in that case, it's already over.

So let's like sort of not worry about it. Now, I am quite sanguine. Is that how do you say that? Sanguine is how I've always pronounced it. Sanguine. I think that sounds right. Actually. Yeah. I could be wrong. I am quite sanguine about that, and that is frankly just a personality trait of mine where I don't get worked up about things that I perceive I don't control. And, you know, I will observe them. And maybe this is the part that sort of bugs Jack is I will dryly explain what's going on and not be very emotional about it. But that's just sort of how I roll to certain extent. Yeah, that's a rational way to live your life. So I support it.

But to your point about sort of the dangers, you know, we want to talk about misinformation. the context of our political enemies, right, or whatever it might be. That's frequently, I mean, this is kind of a bit of a turnoff to the people complaining about this because they're so frequently wrong. Like, this is just an excuse for why people they disagree with, you know, aren't sort of voting for them or or going in the opposite direction. And oh, surely it's not that they fundamentally disagree or have different views of the world.

They must be being lied to. And it's just, and frankly, so much of this in general is, this very presumptuous human solipsistic sort of view of the world, right? And you see this example here where obviously anyone who's smart and well-informed would agree with me. So by definition, if they disagree with me, it must follow that either they're dumb or misinformed. And it never fails to enter the equation that maybe I'm wrong. Maybe I'm misinformed. Maybe I don't know what's going on. And, you know, I think you saw. this frankly to a tremendous extent over the last few years where you know you would see folks that

would very vocally rail against misinformation and when it came to things like COVID were shockingly misinformed right and just like like like and well and and you see it on both sides the beautiful thing about that is yes you could you have no idea which side I'm talking about when I say that exactly and I I have a dad who's a Republican and a mom who's a Democrat and they will both, and they're divorced, unsurprisingly. And they will both sort of lash out at the other one's perspective and say, oh, my God, like, your father is just totally brainwashed. And then my dad will say, like, you're totally brainwashed by the mainstream media or whatever. And the reality is, at this point, I think the technology is so good at manipulating people and-

Technology or the powers that be. Well, who knows? I know that Twitter is a big part of it for my mom and God only knows what kind of shit my dad is reading. But like the, the reality is is that they are both being manipulated and are more extreme in their stances today than they were like 10 or 15 years ago. And so the lack of consciousness of the way this shapes their own lives and their own perspectives is kind of an interesting blind spot. And again, it's sort of like there's there's lots of easy conversations about how the other side is being brainwashed but politically and if we're talking about policy there's not that collective recognition that everyone in some way is susceptible to what's happening and the way new media can sort of

shape perspectives and and get its hooks in and keep us sort of like addicted to our phones Well, and I think that's sort of a pushback to your point of we collectively need to do something about this, right? We collectively are not managing very well. Right. Lots of stuff, particularly with these, because the human tendency is to sort of put their goals ahead of everything else and then realign everything to go with it. And you could take that and say, oh, well, what if AI has the wrong goals and it realign stuff? I actually think there's a potential optimistic argument where AI is actually less likely to do that precisely because it has so much more information and so much more knowledge of the world and does not have sort of an emotional context and it doesn't have the motivational aspect to it. There's also the point where actually maybe an underappreciated thing that I am a little worried about is us pushing, you know, there is a bit about this reinforcement.

enforcement learning and aligning AI that is about it sort of staying on task, there's also a part of it that we kind of want the AI to lie to us, right? If there are certain things that are factually correct that we don't feel aligned with the way we, yeah. Yeah, we don't want to hear that, right? And, you know, is there a issue where you have this super intelligent entity that has constrained in certain respects? It doesn't have some of the emotional motivations and stuff. that humans do, and that's a good thing, but sees the pursuit of, like, being the correct thing. Like, it's actually the fact we're trying to get it to not tell the truth is actually how we're misaligning it, where we're actually, like, pushing it in an incorrect direction because we care more

about what we want the world to be as opposed to what the world is. Yeah, well, and then there's the other aspect of all this, which is that as this becomes entertainment and as the AI gets more effective at identifying what each individual user wants to see and AI is generating, like, its own video content, you can see a couple steps down the road how all of this becomes even more engrossing than the media we already have now. And people become more siloed off than they have been over the last 10 or 20 years. and the problems he's describing with isolation and loneliness and, you know, depths of despair, all of that gets worse. Or maybe it gets better because the eye is just the best friend you didn't know you had. Well, I guess.

But I think the point about humans being, you know, predisposed to get out in the world, have sex and, you know, hunt and gather, like all of that stuff is important from a physiological standpoint. So I bet that humans, if like the one area I would push back on this email is that if humans, if the survival of humans depends on us realizing that some of this is harmful, from an evolutionary standpoint, I would bet on humans recognizing that and evolving and adjusting the way we use technology. But to your point earlier, we're not off to a great start there. So I was just going to make that point. I mean, there is a bit too where, I mean, speak of our touch grass movement. I don't think it's just us. We hear from readers all the time who have increasingly come to appreciate and value and prioritize real-world interactions and understanding and appreciating that that actually does make life better. And maybe to the extent, I mean, this gets very dodgy.

where it sounds really bad. But there's like, if the people who are actually propagating the human race are those that actually value that view of the world, like from an evolutionary perspective, we would actually be moving in, in sort of the right direction. And I sort of delicately pitterpatter on that one.

I was going to say it does get dodgy. Sure. But like, I mean, I don't know. I mean, and I do think there's, we are entering a period of incredible unknowns. And so everything I said on this podcast to now might be 100% wrong. And I would just note that applies to folks that are so certain in the other direction as well, as to how this is going to play out, what it's going to do.

I think it's reasonably pessimistic. I also think there's reason to be optimistic as well. I do think the tail risk is pretty significant. And again, I think living here in Taiwan gives me a unique perspective on this where, and maybe it obscures me and I'm missing the point. But I've talked before where I've always been aware of the existential risk of living in Taiwan. My risk profile has changed over the last few years. And I am, you know, it has increased. But I'd like to think it's a more accurate read of the situation than people who didn't

know anything about this and went from zero to 100. And now they're certain that China is like the bombs might be falling right now. I mean, take a quick look out the window, right? I think that that, that, that, that, that, that that's a mistake. And I wouldn't be surprised. And so this is, this is just sort of my general priors is I think there's real risk here. I think the risk very well may be increasing. I'm not sure about going zero one hundred. Now, it's worth noting some of the folks that have been thinking about this is the longest are those that are the most concerned. So maybe like, I'm actually like making the opposite point here. And that is a fair point to make for sure.

And maybe I'm coming in a little too blazee and other folks are just not thinking about it at all. But I do think there's a risk. This goes back to something you said at the very beginning. What I am worried about over the next few months slash years is we're going to have this sensation of the risk going from zero to 100, particularly because the most obviously impacted people by AI are the people who have the loudest voice, who are in media, who are sort of, you know,

the folks that are going to be on Twitter, those sort of word cells of our world, right? The reason they can be on Twitter all day is because they're sitting at their desk in front of their desk in front of their computer the whole time doing like knowledge work, right? Gives you a lot of time to tweet. Guess when you don't have time to tweet

when you're out like digging a ditch, right? Like there's, I do think this is going to be, there's two points here. Number one, it seems pretty clear that the biggest impact on jobs, particularly at the beginning, is going to be in the digital world, in the knowledge work.

If your job can be done from home, you're almost by definition probably susceptible to this, just as you're also susceptible to the remote work sort of phenomena, right? It's a similar sort of dynamic. And it's the inverse of what happened with globalization and automation,

which was more on sort of blue collar with your hands work, right? And so number one, I think that's clear. Number two, that is the, those are the folks with the loudest voices. And again, I would go back to COVID. If we didn't have the moderate advances in technology

where the people who did have the loudest voices could work at home and then had permission to be the most extreme in terms of lockdowns and keep the kids out of school and XYZ, and I feel fine saying that because I still have a job at my desk and I can watch my kids and take care of sort of things. If everyone was in the same position that blue-collar workers were in during COVID,

where they still had to go to work, and then they had to leave their kids unsupervised at home, and who knows what sort of worrying they're going on, no one cared about them. No one talked about them. They weren't a part of the discussion. And I think that led to bad societal decision-making around many aspects of COVID,

because the only people we listened to were these sort of knowledge workers, in part because they were the ones that at the time and sort of habits of being on things like Twitter and making a big fuss out of it. And so if you fast forward, that the political fervor over this stuff is going to be drastically larger than the political fervor over what happened in the 80s and 90s as far as globalization. In my estimation, precisely because the people impacted are going to be so much louder

about it. Like, again, I'm from small town, Wisconsin. I witness people being devastated, towns being devastated, people's lives being ruined because of the factory in town closing down, right? And no one cared because they didn't have a voice. Right. And this will be flip on a tent. No, people were very aware. The people who wrote newspapers were not aware. The people who were on the news were not aware. The people who were driving the policy discussions, it was more of an abstract concept than something that was real. It was hollowing out entire regions. That's right. And so this is going to be different. So I actually, one of the things I'm actually fairly worried about, and this kind of goes back to Jack's point.

my concern about AI is maybe less about AI than some of the people scared about AI are concerned. And it's much more about the human response to AI. That human response can in part be to Jack's point, people losing themselves in this. And I think that's a very real thing to be concerned about. I think that's the most realistic risk of the takeover risk, the economy risk. I think the idea that everyone is going to be completely engrossed in what these products turn into 10 years from now is the most realistic thing to worry about. It's also the thing that hasn't really emerged in any of the discussions to this point. Yeah, but I think that you're underrating the risk of the overreaction and what that actually entails, right?

I mean, again, I would go back to COVID. COVID was obviously a problem. Like, I'm not a COVID denier by any means. And it is tragic and sad that so many people died. And in retrospect, it seems clear. And I would argue as clear in early 2020, because I actually wrote this on Chichdegro that COVID was going to hit everyone. It was sort of inescapable, right? And so in retrospect, if we grant that everyone's going to get COVID at some point,

would we have taken different actions in trying to suppress what was actually impossible to suppress? and how much harm and damage did that overreaction cause, right? And so I worry about this here. I think your point is good. There is going, I think these risks are real. And I think I've been pretty consistent on this point. But I also worry we're going to compound that problem with what's going to happen over the next sort of year or two as it becomes real. Yeah.

Well, I'll be very interested to see whether there are any real actions taken over the next year or two. because typically Congress has been sort of slow on the uptake on some of the regulation. And I'm not sure right now it's like Microsoft and Google sort of self-regulating, right? So I don't know that the government's going to get too involved in this. Are you envisioning like just sort of the professional class and the media driving more restrictive language? I don't know what's going to happen, but I think that the first and most obvious application of this technology is basically to render a lot of of white collar jobs unnecessary. And I think that's going to provoke a massive political reaction.

And again, you might have the perspective and, you know, that that's appropriate, just as you might have said, boy, wish we would have better paid attention to blue collar workers 20, 30 years ago. The difference, again, to this point is I think white collar workers in our society have a much larger megaphone than blue collar workers did, particularly back then. And I think that's going to be pretty disruptive. of. Now again, if AI kills us all, then it's who cares, right? Then we have a few fights along the way. So there is a ranking of risks here. But I think there are risks that we're facing that are different from and worth considering beyond just AI makes us all on the paper clips.

Yeah. Well, and there's another layer to it. And I'm just curious what you think of it in terms of the risk. there seems to be sort of a spectrum between users choosing to consume content and then algorithms getting so good at manipulating our brains that the choices aren't always conscious or aren't 100% conscious. And so in terms of like AI and what algorithms are already doing, I do feel like it's less of like a personal choice in terms of what what people are doing with social media now. And that's something to sort of be wary of going forward as this thing, as all of it moves closer to the end of the spectrum where our brains are being manipulated.

Do you have a choice? I feel like I have a, well, actually, it's funny. I have more of a choice if I like remove myself from the internet for a few days. I was touching grass over the weekend and felt great the entire time. And so coming back to my phone, I'm less addicted to my phone. But the more I'm on it during the week, the more screen time sort of like spirals out of control, depending on the day. So having the ability to like take a step back is really important for me and my mental health.

Yeah. There's obviously really challenging questions here around, you know, personal responsibility, personal agency versus our susceptibility to being manipulated and having this sort of kick in. And I think those are going to be interesting questions. Like to what extent do we take population level averages, for example, versus needing to understand and incorporate like what actually makes sense in different contexts? I mean, I think one of the most interesting things to think about in terms of the internet is we hear a lot about the problem of information being free and the propagation of content.

And we hear that because the folks that are impacted are people like who work in the news or work in entertainment, right? And like, wow, wasn't it great when we had rom-coms that could make $80 million and we could feel happy about that, right? And the hollowing out of the middle, you know, like I'm sorry, I'm just, you know, not to paint you in a character. I was going to say, I feel that deep in my voice. So yes, I take no offense. Another perspective is if we consider broadly speaking, today we have more access to more amazing content by far. Just take YouTube alone, right? The amount of stuff on there and not just both for entertainment.

And I'm not going to make it old. Think of all the educational content. No, just plain entertainment content. That's actually really interesting and really compelling. And entertainment has value. I'm not one of those people that I think one of the reasons Facebook. was always severely underestimated is people have an overly utilitarian view in their analysis where things are intrinsically worth more if they have utility.

And they under, when they're doing a valuation analysis, they underappreciate that entertainment is actually, you know, generates value, right? Well, not just you could say it's unimportant. That doesn't mean it doesn't matter. It's still like a real thing. And so it depends on your perspective as to whether our current contents, state is better or worse than before. It's definitely worse if you just look in terms of produced content or not, actually,

I don't know if it's definitely worse. You can make the case that it's worse in the, in the context of like professionally produced content. Quality. Right. Yeah. We've lost all these niches and like all these sort of, you know, highbrow sort of things and everything's either like super high end or super like trash and there's nothing in the middle.

And that might be valid. But it's not actually a useful analysis. to the way the world actually is. Because from the perspective of a creator, I'm not sure what Reed Hastings point is referring to, maybe this was it, about how Netflix isn't just competing with HBO. They're competing with gaming and they're competing with YouTube

and they're competing with sleep and they're competing with Twitter. And like when you think about the attention economy generally, you can definitely argue that we're not, you know, like what's available and what's great. grabbing our attention is really fantastic. And the other thing that I think is easy to forget is by the numbers, like people who are spending hours and hours a day just watching TV, right? And there's a bit where arguably people are taking it more interesting compelling content

than they were back then. I think that's part of the problem, honestly, is that there is more compelling content right now. And I guess one of the reasons I'm addicted to my phone, I've got a lot of interesting shit going on on my phone. I'm really good at using the internet to find a, interesting stories, interesting tweets in a lively group chat that is somehow arguing like 24 hours a day as curated by you. And so there's a lot going on that in the past, I don't think would have been nearly as engrossing. Like if you're in the 80s, it's like, all right,

you tape Saturday Night Live and watch it on Sunday, but you're like going out and seeing people. And why do we know that that was better? I mean, I think this is an argument that I think Mark And Drason has made most forcefully where he's talked about growing up on like a farm in like, you know, the middle of nowhere Wisconsin. And the only sort of source of intellectual vitality was visiting like the crappy mall bookstore that was like 45 minutes away. And he's like, you know, if for someone in his position and perspective, if he could have had access to the internet, it would have been transformational in a positive way because

reality sucked. And I think there's a view where your and my, and this is something I'm cognizant about when we talk about sort of the touchgrass movement, is I'm pretty lucky. My life's pretty great. And I have been successful with my business and I can do discretionary activities that cost money. And even when I felt very poor, I could still go to Japan and visit a shrine and get food poisoning. And I talk about how poor I was. And relative to the vast majority of people ever, I was. is just absurdly ridiculously rich and spending my time visiting a foreign country just because I could.

Right. Like there's, you know, and again, I am also wary of this argument because it's used for things like globalization, right? You want to be careful to not get too enamored with any sort of point of view. But I think it is useful to always question your assumptions that the way things are going is obviously bad, right? Like we are always comparing the present to this idealized view of the way things were that was not actually the truth for the vast majority of people. And for those that were approximately the truth, the reality was actually much more complicated. Yep. I think that's a good note to wrap it up on.

But my instinct is that the technology solutions we have now can raise the floor for people in terms of what, their personal enjoyment sitting at their quality of life. Yeah, this is like the Ready Player 1 sort of idea. You know, like just being in a virtual reality environment was just a lot better than being in the crappy real environment. I do think the question, though, if I can argue the other side is what's causal here? What's actually the cause and effect? Is it that the environment was crappy and thus virtual reality was a wonderful escape?

or is it because virtual reality was a wonderful escape that we allowed the lived environment to become crappy, right? And this is actually a fascinating question around AI generally. Everyone's like, oh, thank goodness, AI is arriving as the world's falling apart. And it's like, well, what were the preconditions for this version of AI to arrive? You had to have the internet. You had to have the largest volume of sort of human language interaction that was freely accessible and searchable.

that was a key condition to getting this off the ground. Well, also simultaneously, and this is why I questioned Jack's assertion that I'm attacked optimists, I've made the point that things like social media and particularly Twitter have major issues, major problems. And I say this is someone that values, and people always are like, oh, I find Twitter super valuable. So do I. I can also step back and think like maybe this sort of hurting mechanism that aligns everyone

around the most insane viewpoint on either side, because. normal people in the middle don't want to get their heads cut off might not be good for society generally, right? And so the question is, is AI by chance showing up as everything's falling apart? Or are those actually deeply, intimately interconnected? And of course, it's showing up as things are falling apart because the mechanism that was driving things to fall apart was also the mechanism driving AI. Yeah. And it's going to be interesting to see how we navigate all of it over the next, you know, 30, 40 years. Maybe all of it's going to hit in the next five years. My final thought would be that it

raises the floor and potentially lowers the ceiling in terms of human enjoyment and seeking out sort of more durable solutions to long-term happiness. But wait, why does it, lower the ceiling? Because you can still touch grass. You can touch grass, but you're less likely to touch grass once you're conditioned to sort of live in a digital metaverse type way. You know what I mean? Like I think if that's where you're going to escape your otherwise dreary life, you may be less likely. And frankly, a lot of people experienced that during COVID where they sort of got very comfortable in their little cocoon. And then it was a chore to like go out and meet people for dinner. And they had to sort of relearn how to do that. And I think on a broader scale, it's possible that that happens with a lot of

people who just get addicted to living life through social media. Yeah. No, I think that that that's, that's definitely plausible. So I don't know. We'll see. All of it's plausible. That's what's crazy. Yeah. The good thing is is that, you know, if a little solipsism here, you know, there was a stage seven or eight or two or three years ago where I was ready to quit trajectory, in part because it just seemed like everything I wrote about as far as the internet favoring the centralization of big companies turned out to be true. And I wrote when I first wrote aggregation theory, I wrote an article saying, you know, this is inevitably going to lead to everything getting bogged down in antitrust debates and questions.

And it's going to be super annoying because the ways, the way in which these are about antitrust do not align to our current thing. And all that ended up being true. And I found myself on a daily basis writing about like congressional hearings and antitrust stop. And I'm like, you were like, I'm not an antitrust lawyer. I remember texting with you about this at some point. You were like, I did not sign up to just cover this constantly. Yeah.

And it's funny because one of the reasons I, and it's interesting like the concept of burnout, right? I work an insane amount, right? But I do it because it's like I love what I'm doing. And, you know, if I were not writing strategy, I would be spending all week as I did, reading about and thinking about how the human brain works and how these models work and all those sorts of things. There's literally nothing that I would have changed about my weekend as far as that was concerned. And I recognize that makes me incredibly fortunate and blessed and all those sorts of things. It also leads to me working just like insane amount, right?

I think burnout comes, and this is not a new sort of observation, burnout comes when you're doing, working really hard and you don't feel that sort of internal sense of motivation. You feel a sense of compulsion. You're doing stuff because you feel like you have to. And that's the closest I've ever come to burnout was sort of that, that period. And I bring this up because I'm usually not one to mark anniversaries, but Saturday was actually the 10 year anniversary of me starting Stratory. Woo!

A lot of celebrations on the podcast here. Got a new baby, got 10 years of Stratory. Strategery all grown up now. Yeah. About to inter-adolescence. But I mean, there is a bit where, you know, the reason I just sort of bring this up in the context of this is in 10 years, I came into writing strategy feeling like I had a pretty good handle of the way tech worked. And that was an opening for me to write about because I felt the broad conception of tech was mostly wrong. And I was going to come in and set the record straight as it were.

And that ended up turning out pretty well and then it got very boring to a certain extent. What is striking is I feel like coming in at 10 years and the level of uncertainty and change around this I think is difficult to overstate. Like I just made a lot of assertions on this about what I think the relative things about risk are and the way things are going to work and play out and how people are going to respond to this. And there's literally no one that knows or can know. And not just because this is some sort of humble statement about, oh, all new technology.

Who knows what's going to happen with mobile, right? Actually, I think we knew pretty well what was going to happen with mobile. And there's nothing particularly surprising that happened. To the extent it was surprise is that Apple maintained its advantage to the extent it did. But again, that was one of the areas where I felt I could come into tech and make a difference because it was very clear to me why Apple would sort of maintain their advantage in a different way than they did with sort of back. So I think that was knowable. I think I strongly suspect that the level of unknowableness,

about what's happening is pretty much unmatched, right? I think, yeah, it's staggering as a newcomer. There's like a lot of uncertainty in the way people talk about what this is going to evolve into whether you're talking products or its impact on society and the economy. Right, because I mean, what I'll talk right about most is like what's the impact on like business structure, right? Where is the actual value going to be garish? I wrote AI in the big five.

That was sort of, you know, I can, I think I said explicitly, this is a short to medium term view because the long term, it's not really clear the extent this is going to play out. And I can't remember if it was that article, another one where I talked about this sort of music generation bit and generating music from images. And I think actually that was getting along the right track, which is my current thinking is this technique and this idea, language is a means. It's not the actual end. And it's going to be more generalizable than we think in a way that human cognition is more. more generalizable than we think. Like, we know the parts of our brains that deal with vision are not tuned to vision and

actually get repurposed for people that are blind to doing other stuff, right? Like the overall, the brain is a much more generalizable sort of organ than we used to think. And that's both exciting and also dramatically increases the uncertainty. Leave aside where business value is going to be garnished, right? I mean, I actually think there's a possibility, as this works down to local devices, this may be the end of aggregation when I sort of talked about previously, where everything is individualized and personalized. And we may look back on this internet era of people going online and interacting back and forth

as this weird period in time that was actually just a 30-year period where we did this. weird stuff online and it no longer made sense before or sort of after like, wow, that was kind of wild. And maybe this ends up being actually very optimistic where we actually touch grass more precisely because the, again, I don't know. No, but I think that's a distinct possibility that the way this trends ultimately leads to people disengaging and living their lives a little bit differently and understanding their relationship to technology a little bit differently because it's going to be so extreme and so engrossing if you're not sort of vigilant about it. Yeah, and there's sort of two points. I know we took a one question

and we basically built the whole podcast around this. Well, it was a really good email. So thank you, Jack. I think there, you know, one of the things I think a lot about is biases, right? And there's things like confirmation bias, right? You, you tend to look for data that supports your preexisting thesis. And so you have to be very disciplined about sort of thinking about where you could be wrong and looking for evidence in that regard and things on those lines. And there's also things you don't know that you don't know, right? And then you're not even looking for evidence there because you didn't know, you didn't know it. And that's, there's, but there's all kinds of different biases they have to be aware of. But I think there's two biases when it comes to, uh, your own

personal experience that can that kind of work against each other. So one is where you look at everyone else and you're like, I'm different than them, you know, and like, they're, they're, they're, are persuaded by misinformation. I know the truth, right? And you guys are like, are you sure about that? I am. I'm the only person in the world that's not shaped by any of these algorithms just for the record. That's right. Yeah, no one's manipulating Andrew Sharp. You know, although your insistence on instinctually disagreeing with whatever to argue is actually probably does serve you quite well in that regard. To be clear. So there's that. But the other one is to say, that to be under optimistic and say, no one can be like me in sort of a negative sense.

And the reason I bring this up is I feel very, you know, there's a, there's sort of a general thing in surveys where you ask people, what's the direction of like America? They're like, it's terrible. It's like, how's your life personally? It's like, it's going great. And it's like there's this, there's this weird sort of disconnect that I think should be a bit of an eye opening about, you know, your information diet and maybe believe. everything that comes from a media that's probably incentivized in certain directions to keep you engaged and angry.

That might not be correct. And I think about this myself where, again, with all the caveats that I've been very fortunate in life and, you know, my material welfare is taken care of. But there is a bit where you get to things like group chats you mentioned before. Like you have this sort of I social network more than ever, but I'm not on Twitter. in like this where I think Twitter is, you know, I post a thing about, you know, on Twitter over the weekend about a 10 anniversary and of course, lots of nice tweets and it's all great, thank you and you know, try to click like on all of them. And then there's like three like snarky comments and those they burn and you, and you remember it and you think about it. Like we just as humans are just really not well suited to this public social networking sort of model. It's very bad. It's just it's not good. Again, I think social networking. But this idea where I can be in Taiwan, and this doesn't just apply to being international, you get busy, you have kids, you're stuck in your life, and everyone feels this, man, I don't hang out with my friends like I used to. And suddenly you're in an environment thanks to technology where you can basically be hanging out with your friends 24-7 to the point where I almost feel a little overwhelmed here, right? But a meaningful shift in quality of life, at least from my perspective.

And that's an example where we went through this rough period of public social networking, maybe not so great, maybe kind of disruptive to the human psyche, particularly because the people who aren't susceptible are like trolls and are just out there attacking you. And like people, I think normal people, it's just very, very hard to handle. I've been online as a public figure for a long time and I still have trouble sort of dealing with it emotionally. But when I'm in a group chat and people are taking the crap out of me because, of something stupid I said or I'm being overly pro bucks or whatever. It's no problem because I'm in a trusted environment with people that I know, care about my overall welfare, and that feels sort of very healthy. And then, of course, you should get together at some point.

Anyhow, I'm rambling on and on. But the point is, just as we can be overly optimistic, we can be overly pessimistic, too, by looking at the way things were. And it gets back to this, maybe the public internet was sort of a bootloader, for AI. And we're going to evolve ourselves as sort of society. The way we were doing the internet might not have been so great. And we will reset in a new place where we're much more online than we were previously. We're much more connected and plugged in. But it's actually better than the, he mentioned bowling alone. Bowling alone was written pre-internet. This idea of atomization and people

feeling isolated is a problem that is not an internet problem. That's been a societal problem with the industrial society going back for a very, very long time. And again, I'm not saying this is the case. But I do think there is an underappreciated, optimistic take about all of this stuff where by virtue of offloading the digital to the machines, it actually frees us up to re-engage with the things that actually matter. and make lives better. And I'm not saying that's going to happen, but I do think it's a mistake to overly anchor on this is definitely going to be bad.

Well, I am saying that's going to happen. I'm betting on humans in the long term. Humans have proven to be pretty adaptable over the years and over thousands of years. And as I said earlier, I think if it's a real problem, humans will figure out a healthier relationship to all this. I'm really sorry, I missed your 10-year anniversary post. didn't see that you had announced that publicly. I was too busy touching grass, living the principles of the podcast. But you know, your 10 day away baby, I think is more

important. Yeah, that's true. There's a little drama on the baby front. No, I think in general, a good theme here is don't think in absolutes as we grapple with all of this. Because even the state of society, and again, like coming from a pretty privileged place, but at the same time, I, when When I'm out interacting with the world, I don't experience the world as a place that's in decline where everyone's miserable and isolated. And you can spend your time online reading news and reading stories that are sort of engineered to make you upset and worried. And the world can look like a shittier place than it actually is.

And to the extent we're looking at where we are now as all bad. I think the way news works these days is probably part of that. And it's probably, well, it's definitely not accurate to say that everything is bad now as a result of the last 20 years in technology. And I think going forward, it's important to keep that in mind. And there are always going to be tradeoffs. That's another theme of the podcast. I think the social networking thing is a great example. Like people are surprised that I view Twitter as problematic.

Again, because they're like I get so much value of it. Again, I get so much value out of it too. But I think that, you know, there's real concerns about its societal impact. But if you step back and consider social networking broadly, which why would you not include instant messaging, for example? That's a social network. The value of WhatsApp depends on how many people have it. For example, XYZ, I think the value is. is extremely positive.

And, and, you know, the, it just all depends on your perspective. Yes. And there are tradeoffs. And, you know, and the reason to go back to the anniversary is, it's just, it's really striking to be 10 years in and to feel like this is the most uncertain and exciting and scary time that, that we've been through yet. And it's, yeah, it's a big deal. That was the other point that I wanted to make at the end is we just have no idea. and that is both terrifying and exciting at the same time. Right.

It's not we have no idea in a, you know, I'm being fake humble sort of way. Like, I think you, like, there's, the people that understand this at a very deep level also don't. And be wary of anyone that is certain about where this is going to go. And Ben, we will close it there and come back. You had a great rundown. There's actually a bunch of things I wanted to talk about here. I know. Sorry, we made it through one question today.

Well, a lot of twists and turns, but it was a lot of fun. And we'll come back later in the week, hit some of what we missed. Again, if you have any sort of vacation-type questions, keep them light, send them to email at sharptech. And we are going to be pre-recording next week's episode. And then who the hell knows what's going to happen with the rest of April, but we are coming back later this week. There's uncertainty everywhere. That's the big takeaway. Uncertainty all around us.

And 10 years of Stretary. Congratulations and Ben, until Thursday, I will talk to you soon. Thank you. Talk to you later.