Lex Fridman (01:42:33) And there’s some aspect, even just inside business, where you don’t just make the customer happy, but you also have to think about where is this going to take humanity if you zoom out a bit?
Jeff Bezos (01:42:45) A hundred percent and you can feel your brain. Brains are plastic and you can feel your brain getting reprogrammed. I remember the first time this happened to me was when Tetris who’d first came on the scene. Anybody who’s been a game player has this experience where you close your eyes to lay down to go to sleep and you see all the little blocks moving and you’re kind of rotating them in your mind and you can just tell as you walk around the world that you have rewired your brain to play Tetris. But that happens with everything. I think we still have yet to see the full repercussions of this, I fear, but I think one of the things that we’ve done online and largely because of social media is we have trained our brains to be really good at processing super short form content. (01:43:52) Your podcast flies in the face of this. You do these long format things.
Lex Fridman (01:43:59) Books do too.
Jeff Bezos (01:44:00) And reading books is a long format thing and if something is convenient, we do more of it. We carry around in our pocket a phone, and one of the things that phone does for the most part is it is an attention shortening device because most of the things we do on our phone shorten our attention spans. And I’m not even going to say we know for sure that that’s bad, but I do think it’s happening. That’s one of the ways we’re co-evolving with that tool. But I think it’s important to spend some of your time and some of your life doing long attention span things.
Lex Fridman (01:44:41) Yeah, I think you’ve spoken about the value in your own life of focus, of singular focus on a thing for prolonged periods of time, and that’s certainly what books do and that’s certainly what that piece of technology does. But I bring all that up to ask you about another piece of technology, AI, that has the potential to have various trajectories to have an impact on human civilization. How do you think AI will change us?
Jeff Bezos (01:45:14) If you’re talking about generative AI, large language models, things like ChatGPT, and its soon successors, these are incredibly powerful technologies. To believe otherwise is to bury your head in the sand, soon to be even more powerful. It’s interesting to me that large language models in their current form are not inventions, they’re discoveries. The telescope was an invention, but looking through it at Jupiter, knowing that it had moons, was a discovery. My God, it has moons. And that’s what Galileo did. And so this is closer on that spectrum of invention. We know exactly what happens with a 787, it’s an engineered object. We designed it. We know how it behaves. We don’t want any surprises. Large language models are much more like discoveries. We’re constantly getting surprised by their capabilities. They’re not really engineered objects. (01:46:35) Then you have this debate about whether they’re going to be good for humanity or bad for humanity. Even specialized AI could be very bad for humanity. Just regular machine learning models can make certain weapons of war, that could be incredibly destructive and very powerful. And they’re not general AIs. They could just be very smart weapons. And so we have to think about all of those things. I’m very optimistic about this. So even in the face of all this uncertainty, my own view is that these powerful tools are much more likely to help us and save us even than they are to on balance hurt us and destroy us. I think we humans have a lot of ways of we can make ourselves go extinct. These things may help us not do that, so they may actually save us. So the people who are overly concerned, in my view, overly, it is a valid debate. I think that they may be missing part of the equation, which is how helpful they could be in making sure we don’t destroy ourselves. (01:48:07) I don’t know if you saw the movie Oppenheimer, but to me, first of all, I loved the movie and I thought the best part of the movie is this bureaucrat played by Robert Downey Jr, who some of the people I’ve talked to think that’s the most boring part of the movie. I thought it was the most fascinating because what’s going on here is you realize we have invented these awesome, destructive, powerful technologies called nuclear weapons and they’re managed and we humans, we’re not really capable of wielding those weapons. And that’s what he represented in that movie is here’s this guy, he wrongly thinks… he’s being so petty. He thinks that Oppenheimer said something bad to Einstein about him. They didn’t talk about him at all as you find out in the final scene of the movie. And yet he’s spent his career trying to be vengeful and petty. (01:49:19) And that’s the problem. We as a species are not really sophisticated enough and mature enough to handle these technologies. And by the way, before you get to general AI and the possibility of AI having agency and there’s a lot of things would have to happen, but there’s so much benefit that’s going to come from these technologies in the meantime, even before there are general AI in terms of better medicines and better tools to develop more technologies and so on. So I think it’s an incredible moment to be alive and to witness the transformations that are going to happen. How quickly will happen, no one knows. But over the next 10 years and 20 years, I think we’re going to see really remarkable advances. And I personally am very excited about it.
Lex Fridman (01:50:12) First of all, really interesting to say that it’s discoveries, that it’s true that we don’t know the limits of what’s possible with the current language models.
Jeff Bezos (01:50:24) We don’t.
Lex Fridman (01:50:24) And it could be a few tricks and hacks here and there that open doors to hold entire new possibilities.
Jeff Bezos (01:50:33) We do know that humans are doing something different from these models, in part because we’re so power efficient. The human brain does remarkable things and it does it on about 20 watts of power. And the AI techniques we use today use many kilowatts of power to do equivalent tasks. So there’s something interesting about the way the human brain does this. And also we don’t need as much data. So self-driving cars, they have to drive billions and billions of miles to try to learn how to drive. And your average 16-year-old figures it out with many fewer miles. So there are still some tricks, I think, that we have yet to learn. I don’t think we’ve learned the last trick. I don’t think it’s just a question of scaling things up. But what’s interesting is that just scaling things up, and I put just in quotes because it’s actually hard to scale things up, but just scaling things up also appears to pay huge dividends.
Lex Fridman (01:51:40) Yeah. And there’s some more nuanced aspect about human beings that’s interesting if it’s able to accomplish like being truly original and novel. Large language models, being able to come up with some truly new ideas. That’s one. And the other one is truth. It seems that large language models are very good at sounding like they’re saying a true thing, but they don’t require or often have a grounding in a mathematical truth, basically is a very good bullshitter. So if there’s not enough data in the training data about a particular topic, it’s just going to concoct accurate sounding narratives, which is a very fascinating problem to try to solve, how do you get language models to infer what is true or not to introspect?
Jeff Bezos (01:52:41) Yeah, they need to be taught to say, “I don’t know,” more often and I know several humans who could be taught that as well.
Lex Fridman (01:52:50) Sure. And then the other stuff, because you’re still a bit involved in the Amazon side with the AI things, the other open question is what kind of products are created from this?
Jeff Bezos (01:53:01) Oh, so many. We have Alexa and Echo and Alexa has hundreds of millions of installed base inputs. And so there’s Alexa everywhere. And guess what? Alexa is about to get a lot smarter. And so from a product point of view, that’s super exciting.
Lex Fridman (01:53:27) There’s so many opportunities there
Jeff Bezos (01:53:30) So many opportunities. Shopping assistant, all that stuff is amazing. And AWS, we’re building Titan, which is our foundational model. We’re also building Bedrock, which are corporate clients at AWS. Our enterprise clients, they want to be able to use these powerful models with their own corporate data without accidentally contributing their corporate data to that model. And so those are the tools we’re building for them with Bedrock. So there’s tremendous opportunity here.
Lex Fridman (01:54:03) Yeah, the security, the privacy, all those things are fascinating. Because so much value can be gained by training on private data, but you want to keep this secure. It’s a fascinating technical problem.
Jeff Bezos (01:54:13) Yes. This is a very challenging technical problem and it’s one that we’re making progress on and dedicated to solving for our customers.
Lex Fridman (01:54:21) Do you think there will be a day when humans and robots, maybe Alexa, have a romantic relationship like in the movie Her?
Jeff Bezos (01:54:29) Well, I think if you look at the-
Lex Fridman (01:54:31) Just brainstorming products here.
Jeff Bezos (01:54:32) … if you look at the spectrum of human variety and what people like, sexual variety, there are people who like everything. So the answer to your question has to be yes.
Lex Fridman (01:54:43) Okay. I guess I’m asking when-
Jeff Bezos (01:54:45) I don’t know how widespread that will be.
Lex Fridman (01:54:45) … All right.
Jeff Bezos (01:54:48) But it will happen. Productivity
Lex Fridman (01:54:49) I was just asking when for a friend, but it’s all right. Moving on. Next question. What’s a perfectly productive day in the life of Jeff Bezos? You’re one of the most productive humans in the world.
Jeff Bezos (01:55:03) Well, first of all, I get up in the morning and I putter. I have a coffee.
Lex Fridman (01:55:09) Can you define putter?
Jeff Bezos (01:55:11) I slowly move around. I’m not as productive as you might think I am. Because I do believe in wandering and I read my phone for a while. I read newspapers for a while. I chat with Laura and I drink my first coffee. So I move pretty slowly in the first couple of hours. I get up early just naturally, and then I exercise most days. Most days it’s not that hard for me. Some days it’s really hard and I do it anyway, I don’t want to, and it’s painful. And I’m like, “Why am I here?” And I don’t want to do any of this.
Lex Fridman (01:55:52) “Why am I here at the gym?”
Jeff Bezos (01:55:53) “Why am I here at the gym? Why don’t I do something else?” It’s not always easy.
Lex Fridman (01:55:59) What’s your social motivation in those moments?
Jeff Bezos (01:56:02) I know that I’ll feel better later if I do it. And so the real source of motivation, I can tell the days when I skip it, I’m not quite as alert. I don’t feel as good. And then there’s harder motivations. It’s longer term, you want to be healthy as you age. You want health span. Ideally, you want to be healthy and moving around when you’re 80 years old. And so there’s a lot of… But that kind of motivation is so far in the future, it can be very hard to work in the second. So thinking about the fact I’ll feel better in about four hours if I do it now, I’ll have more energy for the rest of my day and so on and so on.
Lex Fridman (01:56:42) What’s your exercise routine, just to linger on that? How much you curl? What are we talking about here? That’s all I do at the gym so I just…
Jeff Bezos (01:56:52) My routine on a good day, I do about half an hour of cardio and I do about forty-five minutes of weightlifting, resistance training of some kind, mostly weights. I have a trainer who I love who pushes me, which is really helpful. He’ll say, “Jeff, can we go up on that weight a little bit?” (01:57:18) And I’ll think about it and I’ll be like, “No, I don’t think so.” (01:57:23) And he’ll look at me and say, “Yeah, I think you can.” And of course he’s right.
Lex Fridman (01:57:31) Yeah, of course. Of course.
Jeff Bezos (01:57:32) So it’s helpful to have somebody push you a little bit.
Lex Fridman (01:57:34) But almost every day, you do that?
Jeff Bezos (01:57:37) Almost every day, I do a little bit of cardio and a little bit of weightlifting and I’d rotate. I do a pulling day and a pushing day and a leg day. It’s all pretty standard stuff.
Lex Fridman (01:57:48) So puttering, coffee, gym-
Jeff Bezos (01:57:49) Puttering, coffee, gym, and then work.
Lex Fridman (01:57:53) … work. But what’s work look like? What do the productive hours look like for you?
Jeff Bezos (01:57:59) So a couple years ago, I left as the CEO of Amazon, and I have never worked harder in my life. I am working so hard and I’m mostly enjoying it, but there are also some very painful days. Most of my time is spent on Blue Origin and I’m so deeply involved here now for the last couple of years. And in the big, I love it, and the small, there’s all the frustrations that come along with everything. We’re trying to get to rate manufacturing as we talked about. That’s super important. We’ll get there. We just hired a new CEO, a guy I’ve known for close to 15 years now, a guy named Dave Limp who I love. He’s amazing. So we’re super lucky to have Dave, and you’re going to see us move faster there.
(01:58:46) So my day of work, reading documents, having meetings, sometimes in person, sometimes over Zoom, depends on where I am. It’s all about the technology, it’s about the organization. I have architecture and technology meetings almost every day on various subsystems inside the vehicle, inside the engines. It’s super fun for me. My favorite part of it is the technology. My least favorite part of it is building organizations and so on. That’s important, but it’s also my least favorite part. So that’s why they call it work. You don’t always get to do what you want to do.
Lex Fridman (01:59:31) How do you achieve time where you can focus and truly think through problems?
Jeff Bezos (01:59:36) I do little thinking retreats. So this is not the only way, I can do that all day long. I’m very good at focusing. I don’t keep to a strict schedule. My meetings often go longer than I planned for them to because I believe in wandering. My perfect meeting starts with a crisp document. So the document should be written with such clarity that it’s like angels singing from on high. I like a crisp document and a messy meeting. And so the meeting is about asking questions that nobody knows the answer to and trying to wander your way to a solution. And when that happens just right, it makes all the other meetings worthwhile. It feels good. It has a kind of beauty to it. It has an aesthetic beauty to it, and you get real breakthroughs in meetings like that.
Lex Fridman (02:00:37) Can you actually describe the crisp document? This is one of the legendary aspects of Amazon, of the way you approach meetings is this, the six-page memo. Maybe first describe the process of running a meeting with memos.
Jeff Bezos (02:00:51) Meetings at Amazon and Blue Origin are unusual. When new people come in, like a new executive joins, they’re a little taken aback sometimes because the typical meeting, we’ll start with a six-page narratively structured memo and we do study hall. For 30 minutes, we sit there silently together in the meeting and read.
Lex Fridman (02:00:51) I love this.
Jeff Bezos (02:01:17) Take notes in the margins. And then we discuss. And the reason, by the way, we do study, you could say, I would like everybody to read these memos in advance, but the problem is people don’t have time to do that. And they end up coming to the meeting having only skimmed the memo or maybe not read it at all, and they’re trying to catch up. And they’re also bluffing like they were in college having pretended to do the reading.
Lex Fridman (02:01:42) Yeah. Exactly.
Jeff Bezos (02:01:43) It’s better just to carve out the time for people.
Lex Fridman (02:01:47) Yeah. And do it together.
Jeff Bezos (02:01:47) So now we’re all on the same page, we’ve all read the memo, and now we can have a really elevated discussion. And this is so much better from having a slideshow presentation, a PowerPoint presentation of some kind, where that has so many difficulties. But one of the problems is PowerPoint is really designed to persuade. It’s kind of a sales tool. And internally, the last thing you want to do is sell. Again, you’re truth seeking. You’re trying to find truth. And the other problem with PowerPoint is it’s easy for the author and hard for the audience. And a memo is the opposite. It’s hard to write a six-page memo. A good six-page memo might take two weeks to write. You have to write it, you have to rewrite it, you have to edit it, you have to talk to people about it. They have to poke holes in it for you. You write it again, it might take two weeks. So the author, it’s really a very difficult job, but for the audience it’s much better.
(02:02:45) So you can read a half hour, and there are little problems with PowerPoint presentations too. Senior executives interrupt with questions halfway through the presentation. That question’s going to be answered on the next slide, but you never got there. If you read the whole memo in advance… I often write lots of questions that I have in the margins of these memos, and then I go cross them all out because by the time I get to the end of the memo, they’ve been answered. That’s why I save all that time.
(02:03:11) You also get, if the person who’s preparing the memo, we talked earlier about group think and the fact that I go last in meetings and that you don’t want your ideas to pollute the meeting prematurely, the author of the memos has got to be very vulnerable. They’ve got to put all their thoughts out there and they’ve got to go first. But that’s great because it makes them really good. And you get to see their real ideas and you’re not trompling on them accidentally in a big PowerPoint presentation meeting.
Lex Fridman (02:03:50) What’s that feel like when you’ve authored a thing and then you’re sitting there and everybody’s reading your thing?
Jeff Bezos (02:03:54) I think it’s mostly terrifying.
Lex Fridman (02:03:57) Yeah. But maybe in a good way? Like a purifying?
Jeff Bezos (02:04:02) I think it’s terrifying in a productive way, but I think it’s emotionally, a very nerve-racking experience.
Lex Fridman (02:04:13) Is there a art, science to the writing of this six-page memo or just writing in general to you?
Jeff Bezos (02:04:20) It’s really got to be a real memo. So it means paragraphs have topic sentences. It’s verbs and nouns. That’s the other problem with PowerPoint presentations, they’re often just bullet points. And you can hide a lot of sloppy thinking behind bullet points. When you have to write in complete sentences with narrative structure, it’s really hard to hide sloppy thinking. So it forces the author to be at their best, and so they’re somebody’s really their best thinking. And then you don’t have to spend a lot of time trying to tease that thinking out of the person, and you’ve got it from the very beginning. So it really saves you time in the long run.
Lex Fridman (02:05:03) So that part is crisp, and then the rest is messy. Crisp document, messy meeting.
Jeff Bezos (02:05:07) Yeah, so you don’t want to pretend that the discussion should be crisp. Most meetings, you’re trying to solve a really hard problem. There’s a different kind of meeting, which we call weekly business reviews or business reviews that may be weekly or monthly or daily, whatever they are. But these business review meetings, that’s usually for incremental improvement. And you’re looking at a series of metrics, every time it’s the same metrics. Those meetings can be very efficient. They can start on time and end on time. Future of humanity
Lex Fridman (02:05:35) So we’re about to run out of time, which is a good time to ask about the 10,000-Year Clock.
Jeff Bezos (02:05:43) It’s funny.
Lex Fridman (02:05:44) Yes, that’s what I’m known for, is the humor. Okay. Can you explain what the 10,000-Year Clock is?
Jeff Bezos (02:05:53) Is? 10,000-Year Clock is a physical clock of monumental scale. It’s about 500 feet tall. It’s inside a mountain in west Texas at a chamber that’s about 12 feet in diameter and 500 feet tall. 10,000-Year Clock is an idea conceived by a brilliant guy named Danny Hillis way back in the ’80s. The idea is to build a clock as a symbol for long-term thinking. And you can kind of just very conceptually think of the 10,000-Year Clock as it ticks once a year, it chimes once every a hundred years, and the cuckoo comes out once every a thousand years. So it just sort of slows everything down. And it’s a completely mechanical clock. It is designed to last 10,000 years with no human intervention. So the material choices and everything else. It’s in a remote location, both to protect it, but also so that visitors have to make a pilgrimage.
(02:06:57) The idea is that over time, and this will take hundreds of years, but over time, it will take on the patina of age, and then it will become a symbol for long-term thinking that will actually hopefully get humans to extend their thinking horizons. And in my view, that’s really important as we have become, as a species, as a civilization, more powerful. We’re really affecting the planet now. We’re really affecting each other. We have weapons of mass destruction. We have all kinds of things where we can really hurt ourselves and the problems we create can be so large. The unintended consequences of some of our actions like climate change, putting carbon in the atmosphere is a perfect example. That’s an unintended consequence of the Industrial Revolution, got a lot of benefits from it, but we’ve also got this side effect that is very detrimental.
(02:07:56) We need to start training ourselves to think longer term. Long-term thinking is a giant lever. You can literally solve problems if you think long-term, that are impossible to solve if you think short-term. And we aren’t really good at thinking long-term. Five years is a tough timeframe for most institutions to think past. And we probably need to stretch that to 10 years and 15 years and 20 years and 25 years, and we’d do a better job for our children or our grandchildren if we could stretch those thinking horizons. And so the clock, in a way, it’s an art project, it’s a symbol. And if it ever has any power to influence people to think longer term, that won’t happen for hundreds of years, but we are going to build it now and let it accrue the patina of age.
Lex Fridman (02:08:52) Do you think humans will be here when the clock runs out here on earth?
Jeff Bezos (02:08:56) I think so. But the United States won’t exist. Whole civilizations rise and fall. 10,000 years is so long. No nation state has ever survived for anywhere close to 10,000 years.
Lex Fridman (02:09:12) And the increasing rate of progress makes that even fantastic.
Jeff Bezos (02:09:15) Even less likely so. Do I think humans will be here? Yes. How will we have changed ourselves and what will we be and so on and so on? I don’t know, but I think we’ll be here.
Lex Fridman (02:09:25) On that grand scale, a human life feels tiny. Do you ponder your own mortality? Are you afraid of death?
Jeff Bezos (02:09:32) No. I used to be afraid of death. I did. I remember as a young person being very scared of mortality, didn’t want to think about it, and so on. And as I’ve gotten older, I’m 59 now, as I’ve gotten older, somehow that fear has sort of gone away. I would like to stay alive for as long as possible, but I’m really more focused on health span. I want to be healthy. I want that square wave. I want to be healthy, healthy, healthy, and then gone. I don’t want the long decay. And I’m curious. I want to see how things turn out. I’d like to be here. I love my family and my close friends, and I’m curious about them, and I want to see. So I have a lot of reasons to stay around, but mortality doesn’t have that effect on me that it did maybe when I was in my twenties.