Ex Google CEO: AI Can Create Deadly Viruses! If We See This, We Must Turn Off AI! (YouTube Video Transcript)

(00:00:00) Your YouTube transcript will appear here (00:00:00) someone was leaking information on (00:00:02) Google and this stuff is incredibly (00:00:04) secret so what are the secrets well the (00:00:07) first is Eric Schmidt is the former CEO (00:00:09) of Google who grew the company from $100 (00:00:11) million to 180 billion and this is how (00:00:16) as someone who's LED one of the world's (00:00:17) biggest tech companies what are those (00:00:19) first principles for leadership business (00:00:21) and doing something great well the first (00:00:22) is risk taking is key if you look at (00:00:24) Elon he's an incredible entrepreneur (00:00:27) because he has this Brilliance where he (00:00:28) can take huge risks and fail fast and (00:00:31) Fast failure is important because if you (00:00:33) build the right product your customers (00:00:34) will come but it's a race to get there (00:00:37) as fast as you can because you want to (00:00:38) be first because that's where you make (00:00:40) the most amount of money so what are the (00:00:41) other principles that I need to be (00:00:43) thinking about so here's a really big (00:00:44) one at Google we have the 72010 rule (00:00:47) that generated 10 20 30 40 billion (00:00:49) dollar of extra profits over a decade (00:00:52) and everyone could go do this so the (00:00:53) first thing is what about AI I can tell (00:00:56) you that if you're not using AI at every (00:00:58) aspect of your business you're not going (00:01:00) to make it but you've been in the tech (00:01:02) industry for a long time and you've said (00:01:04) the Advent of artificial intelligence is (00:01:06) a question of human survival AI is going (00:01:09) to move very quickly and you will not (00:01:11) notice how much of your world has been (00:01:13) co-opted by these Technologies because (00:01:15) they will produce greater Delight but (00:01:16) the questions are what are the dangers (00:01:18) are we advancing with it and do we have (00:01:20) control over it what is your biggest (00:01:21) fear about AI my actual fear is (00:01:23) different from what you might imagine my (00:01:25) my actual fear (00:01:26) is that's a good time to pull the plug (00:01:32) this has always blown my mind a little (00:01:33) bit 53% of you that listen to the show (00:01:36) regularly haven't yet subscribed to the (00:01:38) show so could I ask you for a favor (00:01:40) before we start if you like the show and (00:01:42) you like what we do here and you want to (00:01:43) support us the free simple way that you (00:01:45) can do just that is by hitting the (00:01:46) Subscribe button and my commitment to (00:01:48) you is if you do that then I'll do (00:01:49) everything in my power me and my team to (00:01:51) make sure that this show is better for (00:01:53) you every single week we'll listen to (00:01:55) your feedback we'll find the guest that (00:01:56) you want me to speak to and we'll (00:01:58) continue to do what we do thank you so (00:02:00) much (00:02:01) [Music] (00:02:04) Eric I've read about your career and (00:02:06) you've had an extensive a varied a (00:02:10) fascinating career completely unique (00:02:12) career and that leads me to believe that (00:02:14) you could have written about anything (00:02:15) you know you've got some incredible (00:02:17) books all of which I've been through (00:02:18) over the last couple of weeks here in (00:02:20) front of me I apologize no no but I mean (00:02:22) these are subjects that I'm just (00:02:23) obsessed with but this book in (00:02:26) particular of all the things you could (00:02:28) have written about with the world we (00:02:30) find ourselves in why this why Genesis (00:02:35) well first thank you for I wanted to be (00:02:36) on the show for a long time so I'm (00:02:38) really happy to be able to be here in (00:02:40) person in London Henry Kissinger Dr (00:02:42) Kissinger ended up being one of my (00:02:44) greatest and closest friends and 10 (00:02:48) years ago he and I were at a conference (00:02:51) where he heard heard Demis hbus speak (00:02:54) about Ai and Henry would tell the story (00:02:57) that he was about to go catch up on his (00:02:59) jet lag but instead I said go do this (00:03:02) and he listened to it and all of a (00:03:03) sudden he understood that we were (00:03:05) playing with fire that we were doing (00:03:07) something that we did not understand it (00:03:09) would have the impact on and that Henry (00:03:11) had been working on this since he was 22 (00:03:14) coming out of the army after World War (00:03:16) II and his thesis about Kant and so (00:03:19) forth as an undergraduate at Harvard so (00:03:21) all of a sudden I found myself in a (00:03:23) whole group of people who are trying to (00:03:25) understand what does it mean to be human (00:03:27) in an age of AI when this stuff starts (00:03:31) showing up how does our life change how (00:03:34) do our thoughts change humans have never (00:03:37) had an intellectual Challenger of our (00:03:40) own ability or better or worse it just (00:03:43) never happened in history the arrival of (00:03:45) AI is a huge moment in history for (00:03:50) anyone that doesn't know your story or (00:03:52) maybe just knows your story from sort of (00:03:54) Google onwards can you tell me the sort (00:03:56) of inspiration points the education the (00:03:59) experiences that you're draw on when you (00:04:01) talk about these subjects well like many (00:04:05) of the people you meet um as a teenager (00:04:09) I was interested in science I play with (00:04:11) model rockets model trains the the usual (00:04:13) things for a boy in my generation I was (00:04:16) too young to be a video game addict but (00:04:19) I'm sure I would be today if I were that (00:04:21) age um I went to college and I was very (00:04:23) interested in computers and they were (00:04:25) relatively slow then but to me they were (00:04:28) fascinating to give you an example the (00:04:30) computer that I used in college is 100 (00:04:33) million times slower 100 million times (00:04:36) slower than the phone you have in your (00:04:39) pocket and by the way that was a (00:04:40) computer for the entire University so (00:04:43) Moes law which is this notion of (00:04:45) accelerating density of chips has (00:04:47) defined the wealth creation the career (00:04:50) creation the company Creation in my life (00:04:53) so I can be understood as lucky because (00:04:56) I was born with a with an interest in (00:04:58) something which was about to explode (00:05:00) and when when sort of everything happens (00:05:02) together everyone gets swept up in it (00:05:04) and of course the rest is history I was (00:05:07) sat this weekend with (00:05:08) my partners little brother who's 18 (00:05:12) years old yes and as we ate breakfast (00:05:13) yesterday before they flew back to (00:05:15) Portugal we had this discussion with her (00:05:18) family that um her dad was there her mom (00:05:20) was there Raph the younger brother was (00:05:23) there and my girlfriend was there (00:05:25) difficult because most of them don't (00:05:26) speak English so we had to use funnily (00:05:28) enough AI to translate what saying but (00:05:31) the big discussion at breakfast was what (00:05:32) should Raph do in the future he's 18 (00:05:35) years old he's got his career ahead of (00:05:37) him and the decisions he makes as is so (00:05:38) evident in your story at this exact (00:05:41) moment as to what information and (00:05:43) intelligence he acquires for himself (00:05:45) will quite clearly Define the rest of (00:05:47) his life if you were sat at that table (00:05:49) with me yesterday when I was trying to (00:05:51) give Raph advice on what what knowledge (00:05:53) he should acquire at 18 years old what (00:05:55) would you have said and what are the (00:05:56) principles that sit behind (00:05:58) that the most most important thing is to (00:06:00) develop analytical critical thinking (00:06:03) skills I to some level I don't care how (00:06:05) you get there so if you're if you like (00:06:08) math or science or if you like the law (00:06:10) or if you like you know entertainment (00:06:12) just think critically in his particular (00:06:14) case as a as an 18-year-old what I would (00:06:17) encourage him to do is figure out how to (00:06:19) write programming to write programs in a (00:06:22) language called python python is easy to (00:06:25) use it's very easy to understand and (00:06:27) it's become the language of AI so the (00:06:30) the AI systems when they write code for (00:06:32) themselves they write code in Python and (00:06:35) so you can't lose as developing Python (00:06:38) Programming skills and the simplest (00:06:39) thing to do with an 18-year-old man is (00:06:41) say make a game because these are (00:06:45) typically Gamers stereotypically make a (00:06:47) game that's interesting using python (00:06:49) it's interesting because I wondered if (00:06:53) coding you know I think 5 10 years ago (00:06:55) everyone's advice to an 18-year-old has (00:06:57) learn how to code but in a world of AI (00:06:59) where these large language models are (00:07:01) able to write code and are you know (00:07:04) increasing every month in their ability (00:07:05) to write better and better code I (00:07:07) wondered if that's like a dying art form (00:07:09) yeah a lot of people have posed this and (00:07:11) that's not correct it sure looks like (00:07:14) these systems will write code but (00:07:16) remember the systems also have (00:07:17) interfaces called apis which you can (00:07:20) program them so one of the large Revenue (00:07:23) sources for these AI models because (00:07:25) these companies have to make money at (00:07:26) some point right is you build a program (00:07:29) and you actually make take an API call (00:07:30) and ask it a question typ typical (00:07:32) example is give it a picture and tell me (00:07:35) what's in the picture now can you have (00:07:38) some fun with that as an 18-year-old of (00:07:39) course right so so when I say python I (00:07:43) mean python using the tools that are (00:07:46) available to build something new (00:07:48) something that you're interested in and (00:07:50) when you say critical (00:07:51) thinking how does one what is critical (00:07:54) thinking and how does one go about (00:07:55) acquiring that as a skill well the first (00:07:57) and most important thing about critical (00:07:59) thinking is to to distinguish between (00:08:01) being marketed to which is also known as (00:08:03) being lied to and being being given the (00:08:06) argument on your own we' have because of (00:08:09) social media which I hold responsible (00:08:11) for a lot of ills as well as good things (00:08:13) in life we've we've sort of gotten used (00:08:15) to people just telling us something and (00:08:17) believing it because our friends Believe (00:08:19) it or so forth and I strongly encourage (00:08:22) people to check assertions so you get (00:08:26) people say all this stuff and I learned (00:08:27) at Google all those years somebody says (00:08:30) something I check it on Google do I and (00:08:35) you then have a question do you (00:08:37) criticize them and correct them or do (00:08:39) you let it go but you want to be in the (00:08:42) position where somebody makes a (00:08:43) statement like did you know that only (00:08:46) 10% of Americans have passports which is (00:08:49) a widely viewed but false statement um (00:08:52) it's actually higher than that although (00:08:53) it's never high enough in my view in (00:08:55) America but that's an example of (00:08:57) assertion that you can just say is that (00:08:58) true right (00:09:00) there's a a long meme of American (00:09:03) politicians where the Congress is (00:09:04) basically full of criminals um it may be (00:09:07) full of one or two but it's not full of (00:09:08) of 90 but again people believe this (00:09:11) stuff because it sounds plausible so if (00:09:14) if somebody says something plausible (00:09:16) just check (00:09:18) it you have a responsibility before you (00:09:21) repeat something to make sure what (00:09:24) you're repeating is true and if you (00:09:27) can't distinguish between true and false (00:09:29) I suggest you keep your mouth shut right (00:09:32) because you can't run a government a (00:09:34) society without people operating on (00:09:36) basic facts like for example climate (00:09:39) change is real we can debate over (00:09:42) whether it's how to address it but (00:09:44) there's no question the climate is (00:09:45) changing it is a fact it is a (00:09:47) mathematical fact and how do I know this (00:09:50) and somebody will say well how do you (00:09:51) know and I said because science is about (00:09:53) repeatable uh uh experiments and also (00:09:57) proving things wrong so let's say I said (00:10:00) that um climate change is real uh and (00:10:02) this was the first time it had ever been (00:10:04) said which is not true then a 100 people (00:10:06) would say that can't be true I'll see if (00:10:08) he's fa and then and then all of a (00:10:09) sudden they'd see I was right and I'd (00:10:11) get some big prize right so so the (00:10:15) falsifiability of these assertions is (00:10:17) very important how do you know that (00:10:19) science is correct it's because people (00:10:21) are constantly testing (00:10:24) it and why is this skill of critical (00:10:26) thinking so especially important in a (00:10:28) world of AI (00:10:30) well partly because AI will allow for (00:10:32) perfect misinformation so let's use an (00:10:35) example of Tik Tok Tik Tok can be (00:10:38) understand it's called the Bandit (00:10:39) algorithm in computer science in the (00:10:41) sense of the Las Vegas one arm Bandits (00:10:44) do I stay in the Bandit machine and I (00:10:46) keep on this slot machine or do I move (00:10:48) to another slot machine and the the Tik (00:10:52) Tok algorithm basically can be (00:10:53) understood as I'll keep serving you what (00:10:57) you tell me you want but occasionally (00:10:59) I'll give you something from the (00:11:00) adjacent area and is highly addictive so (00:11:04) what you're seeing with social media and (00:11:06) Tik Tok is a particularly bad example of (00:11:08) this is people are getting into these (00:11:09) rabbit holes where they all they see is (00:11:12) confirmatory bias and and the ones that (00:11:15) are I mean if it's fun and you know (00:11:18) entertaining I don't care but you'll see (00:11:20) for example there are plenty of stories (00:11:21) where people have ultimately self harm (00:11:24) or suicide because they're already (00:11:26) unhappy and then and then they start (00:11:28) picking up unhappy and then their whole (00:11:30) environment online is people who are (00:11:33) unhappy and it makes them more unhappy (00:11:35) because it doesn't have a positive bias (00:11:37) so there's a really good example where (00:11:40) um let's say in your case you're the dad (00:11:43) you're going to watch this as the dad (00:11:44) with your kid and you're going to say (00:11:46) you know it's not that bad let me show (00:11:47) you some let me give you some good (00:11:50) Alternatives let me get you inspired let (00:11:51) me get you out of your funk the (00:11:53) algorithms don't do that unless you (00:11:56) force them to it's because the (00:11:58) algorithms are fundamentally about (00:12:00) optimizing an objective function (00:12:02) literally mathematically maximize some (00:12:04) goal that has been trained to they just (00:12:07) in in this case it's attention and by (00:12:09) the way part of it part of we have we (00:12:10) have so much uh outrage is because if (00:12:13) you're a CEO you want to maximize (00:12:15) Revenue to maximize Revenue you maximize (00:12:19) attention and the easiest way to (00:12:20) maximize attention is to maximize (00:12:23) outrage did you know did you know did (00:12:26) you know right and by the way a lot of (00:12:28) the stuff is not true (00:12:30) they're fighting over scarce attention (00:12:32) there was a recent article where there's (00:12:34) an old quote from 1971 from herb Simon (00:12:37) an economist at the time Carnegie melan (00:12:40) who said that um economists don't (00:12:43) understand but in the future the (00:12:44) scarcity will be about attention so (00:12:47) somebody now 50 years later went back (00:12:50) and said I think we're at the point (00:12:52) where we've monetized all attention an (00:12:55) article this week two and a half hours (00:12:58) of videos consumed by young people every (00:13:01) day right now there is a limit to the (00:13:04) amount of video you can you know that (00:13:06) because you have to eat and sleep and to (00:13:07) hang out but these are significant (00:13:10) societal changes that have occurred very (00:13:12) very quickly um when I was young there (00:13:14) was a great debate as to the benefit of (00:13:16) television and you know my argument at (00:13:18) the time was well yes we did you know we (00:13:20) did you know rock and roll and and drugs (00:13:24) and all of that and we watched a lot of (00:13:25) Television but somehow we grew up okay (00:13:28) right so it's the same argument now with (00:13:29) a different a different term will we (00:13:32) will those kids grow up okay um it's not (00:13:35) as obvious because these tools are (00:13:36) highly addictive much more so than (00:13:39) television ever was do you think they'll (00:13:41) grow up okay I personally do because I'm (00:13:44) I'm inherently an optimist I also think (00:13:46) that Society um begins to understand the (00:13:50) problems typical example is there's an (00:13:52) epidemic of harm to teenage girls uh (00:13:55) girls as we know are uh more advanced (00:13:57) than boys at those uh you know below (00:14:00) uh and the girls seem to get hit by (00:14:02) social media at 11 and 12 when they're (00:14:04) not quite capable of handling the the (00:14:07) rejection and the emotional stuff and (00:14:09) it's driven uh you know emergency room (00:14:11) visits self harm and so forth to record (00:14:14) levels it's well documented so Society (00:14:17) is beginning to recognize this now F (00:14:19) schools won't let kids use their phones (00:14:21) when they're in the classroom which kind (00:14:23) of obvious if you ask me um so (00:14:26) developmentally uh one of the core (00:14:28) questions about the AI Revolution is (00:14:31) what does it do to the identity of (00:14:33) children that are growing up your values (00:14:35) your personal values the way you get up (00:14:36) in the morning and think about life is (00:14:38) now set it's highly unlikely that an AI (00:14:40) will change your programming but your (00:14:43) child can be significantly reprogrammed (00:14:45) and one of the things that we talk about (00:14:47) in the book is what happens when the (00:14:49) best friend of your child from birth is (00:14:52) a (00:14:52) computer what's it like now by the way I (00:14:55) don't know we've never done it before (00:14:58) but you're running an experiment on a (00:15:00) billion people without a control right (00:15:04) and so we have to stumble through this (00:15:06) so at the end of the day I'm an optimist (00:15:08) because we will adjust (00:15:11) Society with biases and values to try to (00:15:14) keep us on a moral High Ground human (00:15:16) life and so you should be optimistic for (00:15:19) that because these kids when they grow (00:15:21) up they'll live to a 100 their lives (00:15:23) will be much more prosperous I hope and (00:15:26) I I pray that there'll be much less (00:15:27) conflict uh certainly lifespans are (00:15:30) longer the the likelihood of them being (00:15:32) injured and and in wars and so forth are (00:15:35) much much lower statistically it's a (00:15:37) good message to kids as someone who's (00:15:39) LED one of the world's biggest tech (00:15:41) companies if you were the CEO of Tik (00:15:44) Tok what would you do because I'm sure (00:15:48) that they realize everything you've said (00:15:49) is true but they have this commercial (00:15:52) incentive to drive up the addictiveness (00:15:55) of that algorithm which is causing these (00:15:57) Echo Chambers which is causing the rates (00:16:00) of anxiety and depression amongst young (00:16:01) girls and young people more generally to (00:16:03) increase what would you do so so I have (00:16:05) talked to them and to the others as well (00:16:08) and I think it's it's pretty (00:16:10) straightforward there's sort of good (00:16:12) revenue and bad Revenue when we were at (00:16:15) Google uh Larry and ser and I we would (00:16:17) have situations where we would improve (00:16:19) quality you know we would make the (00:16:21) product better and the debate was do we (00:16:24) take that to revenue in the form of more (00:16:26) ads or do we just make the product (00:16:28) better and and that was a clear choice (00:16:31) and I arbitrarily decided that we would (00:16:33) take 50% to one 50% to the other because (00:16:35) I thought they were both important so (00:16:37) and the founders of course were very (00:16:39) supportive so Google became more moral (00:16:42) and also made more money right all of (00:16:46) the the there's plenty of bad stuff on (00:16:48) Google but it's not on the first page (00:16:50) that was the key thing the alternative (00:16:53) model would be say let's maximize (00:16:55) Revenue we'll put all the really bad (00:16:56) stuff the lies and the cheating and the (00:16:58) deceiving and so forth that draws you in (00:17:00) it will drive you insane and we might (00:17:03) have made more money but first it was (00:17:05) the wrong thing to do but more (00:17:07) importantly it's not sustainable uh (00:17:10) there's a law called gresham's law uh (00:17:13) it's a verbal law obviously um where bad (00:17:16) speech drives out good speech and what (00:17:19) you're seeing is you're seeing in online (00:17:21) communities which have always been um (00:17:23) present with bullying and this kind of (00:17:25) stuff now you've got crazy people in my (00:17:28) view who are building Bots that are (00:17:30) lying right misinformation now why do (00:17:33) you do that you've got in there was a (00:17:35) there was a hurricane in Florida and (00:17:38) people are in serious trouble and you (00:17:40) sitting in the comfort of your home (00:17:42) somewhere else are busy trying to make (00:17:44) their lives more difficult what's wrong (00:17:46) with you like let them get rescued you (00:17:49) know human life is important but there's (00:17:51) something about the the human psychology (00:17:54) where people uh people talk the there's (00:17:56) a German world called shoden Freud you (00:17:58) know there's a bunch of things like this (00:18:00) that we have to address I want social (00:18:02) media and the online world to represent (00:18:04) the best of humanity hope excitement (00:18:07) optimism creativity invention solving (00:18:10) new problems as opposed to the worst and (00:18:13) I think that that is achievable you have (00:18:15) arrived at Google at 46 years old 2001 (00:18:18) 2001 2001 um you had a very extensive (00:18:22) career before then working for a bunch (00:18:23) of really interesting companies Sun (00:18:25) Microsystems is one that I know um very (00:18:27) well you've worked for zero (00:18:29) in California as well Bell Labs was your (00:18:32) first um sort of real job I guess at 20 (00:18:35) years old first sort of big Tech (00:18:37) job what did you learn in this journey (00:18:40) of your life about what it is to build a (00:18:42) great company and what value is as it (00:18:44) relates to being an (00:18:46) entrepreneur and people in teams like if (00:18:48) there were like a set of first (00:18:49) principles that everyone should be (00:18:50) thinking about when it comes to doing (00:18:51) something great and building something (00:18:53) great what are those like first (00:18:55) principles so so the first rule I've (00:18:57) learned is that you need a truly (00:19:00) brilliant person to build a really (00:19:02) brilliant product and that is not me I (00:19:05) work with them so find someone who's (00:19:08) just smarter than you more clever than (00:19:10) you moves faster than you changes the (00:19:13) world is better spoken more handsome (00:19:15) More Beautiful You know whatever it is (00:19:17) that you're optimizing and Ally yourself (00:19:19) with them because they're the people who (00:19:21) are going to make make the world (00:19:23) different um in one of my books we use (00:19:25) the distinction between divas and naves (00:19:28) and a Diva and we use the example of (00:19:30) Steve Jobs who clearly was a diva (00:19:32) opinionated and strong and argumentative (00:19:35) and would bully people if he didn't like (00:19:37) them but was brilliant when he was he (00:19:39) was a diva he wanted Perfection right (00:19:42) aligning yourself with Steve Jobs is a (00:19:44) good idea uh the alternative is what we (00:19:47) call a Nave and a Nave which you know (00:19:49) from British history is somebody Who's (00:19:51) acting on their own um their own account (00:19:53) they're not they're not trying to do the (00:19:55) right thing they're trying to benefit (00:19:56) themselves at the at the at the cost of (00:19:58) others and so if you can identify a (00:20:01) person in one of these teams that (00:20:03) they're just trying to solve the problem (00:20:04) in a really clever way and they're (00:20:06) passionate about and they want to do it (00:20:08) that's how the world moves forward if (00:20:10) you don't have such a person your (00:20:12) company's not going to go anywhere and (00:20:14) the reason is that it's too easy just to (00:20:16) keep doing what you were doing right and (00:20:18) and Innovation is fundamentally about (00:20:20) changing what you're doing up until the (00:20:23) this generation of tech companies the (00:20:26) most companies seem to me to be one-hot (00:20:28) wonders right they would have one thing (00:20:30) that was very successful and then it (00:20:31) would sort of um it was typically follow (00:20:33) an scurve and nothing much would happen (00:20:36) and now I think the the people are (00:20:37) smarter people are better educated you (00:20:39) now see repeatable waves a good example (00:20:42) being Microsoft which is you know an (00:20:44) older company now founded in basically (00:20:47) 81 82 something like that so let's call (00:20:50) that 45 years old but they've reinvented (00:20:53) themselves a number of times right in in (00:20:55) a really powerful way we should probably (00:20:58) talk about this then um before we move (00:21:00) on which is what you're talking about (00:21:02) there is that sort of founder things (00:21:04) people now refer to as founder mode that (00:21:05) founder energy that high conviction that (00:21:07) sort of disruptive thinking um and that (00:21:10) ability to reinvent yourself I was (00:21:12) looking at some stats last night in fact (00:21:13) and I was looking at how long companies (00:21:15) stay on the S&P 500 on average now and (00:21:18) it went from 33 years to 17 years to 12 (00:21:22) years average 10 year and as you play (00:21:24) those numbers forward eventually in sort (00:21:25) of 2050 an AI told me that it would be (00:21:28) about eight years (00:21:30) well I'm not sure I agree with the (00:21:32) founder Mort argument and the reason is (00:21:34) that it's great to have a brilliant (00:21:36) founder and um and there's this it's (00:21:40) actually like more than great it's like (00:21:41) really important and and we need more (00:21:43) brilliant Founders universities are (00:21:45) producing these people by the way they (00:21:47) do exist and they show up every year you (00:21:49) know another Michael Dell at the age of (00:21:51) 19 or 22 these are just brilliant (00:21:54) Founders obviously Gates and Ellison and (00:21:57) sort of my generation of brilliant (00:21:58) founders (00:21:59) Larry and Sergey and so forth for anyone (00:22:02) that doesn't know who Larry and Sergey (00:22:03) are and doesn't know that sort of early (00:22:05) Google story um can you give me a little (00:22:07) bit of that backstory but then also (00:22:08) introduce these characters called Larry (00:22:10) and Sergey for anyone that doesn't know (00:22:11) so Larry pagee and Sergey Bren met at (00:22:14) Stanford um in they were on a grant from (00:22:18) believe it or not the National Science (00:22:19) Foundation as graduate students and (00:22:22) Larry pagee invented a algorithm called (00:22:25) page rank uh which is named after him um (00:22:29) and he and Sergey wrote a paper which is (00:22:31) still one of the most cited papers in in (00:22:33) the world and it's essentially a way of (00:22:36) understanding priority of information (00:22:38) and mathematically it was a forier (00:22:40) transform of the way people normally did (00:22:43) things at at the time and so they wrote (00:22:46) this code I don't think they were that (00:22:48) good a set of programmers you know they (00:22:49) sort of did it they had a computer they (00:22:51) ran out of power in their dorm room so (00:22:53) they um borrowed the power from the dorm (00:22:56) room next to and plugged it in and they (00:22:57) had the data center in the bedroom you (00:22:59) know in the dorm classic story um and (00:23:02) then they moved to a u building that was (00:23:06) owned by um the sister of a girlfriend (00:23:09) at the time and that's how they founded (00:23:12) the company their first investor was a (00:23:15) one the founder of Sun micr System whose (00:23:16) name was Andy bealine who just said I'll (00:23:19) just give you the money because you're (00:23:20) obviously incredibly smart how much did (00:23:21) he give them (00:23:23) $100,000 or yeah maybe it was a million (00:23:26) but in any case it It ultimately became (00:23:28) any billion ions of dollars so it gives (00:23:30) you a sense of this early founding is (00:23:32) very important so the founders then set (00:23:36) up in this little house in menla park (00:23:38) which ultimately we bought at Google you (00:23:40) know as a as a museum and they set up in (00:23:43) the garage and they had Google Google (00:23:45) world headquarters in neon made and they (00:23:47) had a big headquarters um with the four (00:23:50) employees that were sitting below them (00:23:52) and the computer that Larry and sery had (00:23:54) built Larry and sery were very very good (00:23:56) software people and obviously brilliant (00:23:58) but they were not very good hardware and (00:24:00) so they built the computers using (00:24:02) corkboard to separate the CPUs and if (00:24:04) you know anything about Hardware (00:24:05) Hardware generates a lot of Heat and the (00:24:07) corkboard would catch on fire So (00:24:09) eventually when I showed up we started (00:24:11) building proper Hardware with proper (00:24:13) Hardware Engineers but it gives you a (00:24:15) sense of the scrappiness that that was (00:24:17) so (00:24:18) characteristic um and you know today (00:24:21) there are people of enormous impact on (00:24:23) society um and I think that will (00:24:25) continue um for many many years what did (00:24:28) they call you in and at what point did (00:24:30) they realize that they needed someone (00:24:31) like you well Larry said to me uh now (00:24:34) these were they're very young he looked (00:24:35) at me and says we don't need you (00:24:38) now but we'll need you in the future (00:24:41) we'll need you in the future yeah so one (00:24:45) of the things about Larry and Sergey is (00:24:46) that they thought for the long term so (00:24:48) they didn't say Google would be a search (00:24:51) company they said the mission of Google (00:24:53) is to organize all the world's (00:24:55) information and if you think about it (00:24:58) that's pretty audacious 25 years ago (00:25:00) like how are you going to do that and so (00:25:02) they started with web search eventually (00:25:04) and Larry had studied AI quite (00:25:07) extensively and he began to to work and (00:25:10) ultimately he uh acquired uh with with (00:25:13) all all of us obviously uh this company (00:25:16) called Deep Mind here in Britain which (00:25:19) essentially is the um the first company (00:25:22) to really see the AI opportunity and (00:25:25) pretty much all of the things you've (00:25:26) seen from AI in the last decade have (00:25:29) come from people who are either at Deep (00:25:31) Mind or competing with deep mind going (00:25:33) back to this point about principles then (00:25:35) before we move further on um as it (00:25:38) relates to building a great company what (00:25:40) are some of those founding principles we (00:25:41) have lots of entrepreneurs that listen (00:25:43) to the show one of them you've expressed (00:25:45) as this need for the Divas I guess these (00:25:48) people who are just very high conviction (00:25:50) and can kind of see into the future what (00:25:52) are the other principles that I need to (00:25:53) be thinking about when I'm scaling my (00:25:55) company well the first is to think about (00:25:57) scale uh I think a current example is (00:26:00) look at Elon um Elon is an incredible (00:26:03) entrepreneur and an incredible scientist (00:26:05) and if you study how he operates he gets (00:26:08) people by I think sheer force of (00:26:10) personal will to overperform to take (00:26:13) huge risks which somehow he he has this (00:26:17) Brilliance where he can make those (00:26:19) tradeoffs and get it right so these are (00:26:22) exceptional people now in our book with (00:26:25) Genesis we argue that you're going to (00:26:26) have that in your pocket but to whether (00:26:29) you'll have the judgment to take the (00:26:30) risks that Elon does that's another (00:26:32) question the one of the other ways to (00:26:35) think about it is an awful lot of people (00:26:37) talk to me about the companies that (00:26:38) they're founding and they're they're a (00:26:40) little widget you know like I want to (00:26:42) make the camera better I want to make (00:26:44) the dress better I want to make book (00:26:45) publishing cheaper or so forth these are (00:26:47) all fine ideas I'm interested in in (00:26:51) ideas which have the benefit of scale (00:26:54) and when I SC I say scale I mean the (00:26:56) ability to go from zero to Infinity in (00:26:59) terms of the number of users and demand (00:27:01) and scale (00:27:03) um there are plenty plenty of ways of (00:27:05) thinking about this but what would be (00:27:08) such a company in the age of AI well we (00:27:10) can tell you what it would look like you (00:27:12) would have (00:27:13) apps one on Android one on iOS maybe a (00:27:17) few (00:27:18) others those apps will use powerful (00:27:20) networks and they'll have a really big (00:27:23) computer in the back it's doing AI (00:27:25) calculations so future success companies (00:27:29) will all have that right exactly what (00:27:33) problem it solves well that's up to the (00:27:34) founder but if you're not using AI at (00:27:38) every aspect of your business you're not (00:27:41) going to make it and the distinction as (00:27:44) a programming matter is that when I was (00:27:47) doing all of this way back when you had (00:27:49) to write the code now ai has to discover (00:27:52) the (00:27:53) answer it's a very big deal and of (00:27:56) course this was a lot of this was (00:27:57) invented at Google you know 10 years ago (00:27:59) but basically all of a sudden an (00:28:02) analytical programming which sort of (00:28:03) what I did my whole life you know (00:28:04) writing code and you know do this do (00:28:06) that add this subtract this call this so (00:28:09) forth and so on is gradually being (00:28:11) replaced by learning the answer right so (00:28:13) for example we use the example of transl (00:28:16) language (00:28:17) translation uh the the current large (00:28:21) language models are essentially (00:28:23) organized around predicting the next (00:28:25) word well if you can predict the next (00:28:27) word You can predict the next sequence (00:28:29) in biology You can predict the next (00:28:31) action You can predict the next thing (00:28:32) the robot should do so all of this stuff (00:28:35) around large language models and deep (00:28:37) learning that has come out the (00:28:39) Transformer paper gpt3 uh chat GPT which (00:28:42) for most people was this huge moment is (00:28:45) essentially about um predicting the next (00:28:49) word and getting it right in terms of (00:28:51) company culture and how important that (00:28:52) is for the success and Prospects of a (00:28:55) company how do you think about company (00:28:57) culture and how significant and is it (00:28:59) and like when and who sets it so I'll (00:29:02) give well it's almost always set company (00:29:04) cultures are almost always set by the (00:29:06) founders I happen to be on the board of (00:29:07) the Mayo Clinic Mayo Clinic is the (00:29:09) largest healthc care system in America (00:29:11) it's also the most highly rated one and (00:29:13) they have a rule which is called the uh (00:29:16) the needs of the customer come first (00:29:18) which came out of the Mayo brothers (00:29:19) who've been dead for like 120 years um (00:29:23) but that was their principle and I when (00:29:26) I initially got on the board I started (00:29:27) wandering around I thought this is kind (00:29:29) of a stupid you know stupid phrase and (00:29:31) nobody really does this and they really (00:29:33) believe it and they repeat it and they (00:29:35) repeat it right so it's true in (00:29:38) non-technical cultures in that case it's (00:29:40) a healthcare for Service delivery you (00:29:43) can drive a culture even in non-tech in (00:29:45) Tech it's typically an engineering (00:29:47) culture and if I had to do things over (00:29:49) again I would have even more technical (00:29:51) people and even fewer non-technical (00:29:53) people and just make the technical (00:29:55) people figure out what they have to do (00:29:57) um and I'm sorry for that bias because (00:29:59) I'm not trying to offend anybody but the (00:30:01) fact of the matter is the technical (00:30:03) people if you build the right product (00:30:05) your customers will come if you don't (00:30:07) build a product then you don't need a (00:30:08) Salesforce why are you selling an (00:30:10) inferior product so in in the how Google (00:30:13) works book and the ultimately in the (00:30:15) trillion dollar coach book which is (00:30:17) about Bill Campbell we talked a lot (00:30:19) about how the CEO is now the chief (00:30:23) product officer the chief Innovation (00:30:25) officer because 50 years ago you didn't (00:30:28) have access to Capital you didn't have (00:30:29) access to marketing you didn't have (00:30:31) access to sales you didn't have access (00:30:32) to distribution hours I was meeting (00:30:34) today with an entrepreneur who said yeah (00:30:37) you know we'll be 95% Technical and I (00:30:39) said why I said well we have a contract (00:30:41) manufacturer and our products are so (00:30:43) good that people will just buy them this (00:30:45) happened to be a a a technical switching (00:30:47) company um and they said it's only a (00:30:50) 100,000 times better than its (00:30:51) competitors and I said it will sell (00:30:54) unfortunately it doesn't work yet yeah (00:30:56) it isn't the point but if they achieve (00:30:58) their goal people will be lined up (00:31:01) outside the door so as a matter of (00:31:03) culture you want to build a technical (00:31:05) culture with values about getting the (00:31:08) product to work right and working me is (00:31:11) not another thing you do with with (00:31:12) Engineers is you (00:31:14) say they make a nice presentation to you (00:31:16) and they go I said that's very (00:31:18) interesting but you know I'm not your (00:31:20) customer your customer is really tough (00:31:23) because your customers wants everything (00:31:24) to work and free and work right now and (00:31:26) never make any mistakes so so give me (00:31:29) their feedback and if their feedback is (00:31:31) good I love you and if their feedback is (00:31:34) bad then you better get back to work and (00:31:35) stop being so arrogant so what happens (00:31:38) is that in in the invent in the (00:31:40) invention process within firms people (00:31:42) fall in love with an idea and they don't (00:31:44) test it one of the things that Google (00:31:46) did and this is largely Marissa mayor we (00:31:49) back when is one day she said to me I (00:31:52) don't know how to judge user interface (00:31:56) mer was the previous CEO she was the CEO (00:31:59) of Yahoo and before that she ran all the (00:32:01) consumer products at Google uh and she's (00:32:03) now running another company in uh in the (00:32:05) Bay Area but the important thing about (00:32:07) Marissa is she said I can't I I said (00:32:09) well you know the UI the user interface (00:32:11) is great at the time and it was (00:32:12) certainly was and she said I don't know (00:32:16) how to judge the user interface myself (00:32:19) and none of my team do but we know how (00:32:22) to (00:32:23) measure and so what she organized were (00:32:25) AB tests you test one test another so (00:32:28) remember that it's possible using these (00:32:30) networks to actually kind of figure out (00:32:33) because they're highly instrumented uh (00:32:34) dwell time how long does (00:32:37) somebody how long does somebody watch (00:32:39) this how important it is if you go back (00:32:41) to how Tik Tok Works uh one of the (00:32:44) things the signals that they use include (00:32:46) the amount of time you watch commenting (00:32:50) um forwarding uh sharing all those kinds (00:32:53) of things and those you can understand (00:32:54) those as analytics that go into an AI (00:32:57) engine then makes a decision as to what (00:32:59) to do next what to make (00:33:01) viral and on this point of um culture at (00:33:05) scale is it right to expect that the (00:33:08) culture changes as the company scales (00:33:10) because you came into Google I believe (00:33:12) when they were doing sort of hundred (00:33:13) million doll in revenue and you left (00:33:14) when they were doing what 180 billion or (00:33:16) something staggering but is it right to (00:33:19) assume that the culture of a growing (00:33:21) company should scale from when there was (00:33:22) 10 people in that garage to when there's (00:33:24) 100 so when I go back to Google to visit (00:33:27) and they were kind enough to give me a (00:33:28) badge and treat me well of course um I (00:33:32) hear The (00:33:34) Echoes of this um I was at a lunch where (00:33:37) there was a lady running search and a (00:33:39) Gentleman runting ads you know the (00:33:41) successors to the people who worked with (00:33:43) me and I I asked them what's it going (00:33:46) and they said the same (00:33:47) problems you know the same problems have (00:33:49) not been solved but they're much bigger (00:33:52) and so when you go to a company I (00:33:54) suspect um I was not near the founding (00:33:57) of Apple but I was on the board for a (00:33:59) while um the founding culture you can (00:34:02) see today in their Obsession about user (00:34:04) interfaces their Obsession about being (00:34:06) closed and their privacy and secrecy (00:34:08) it's just a different company right I'm (00:34:11) not passing judgment um setting the (00:34:13) culture is important the echo are there (00:34:16) what does happen in big companies is (00:34:18) they become less efficient for many (00:34:20) reasons the first thing that happens is (00:34:23) they become conservative because of (00:34:24) they're public and they have (00:34:26) lawsuits and um a famous example is that (00:34:29) Microsoft after the antitrust um uh case (00:34:32) in the 90s became so conservative in (00:34:35) terms of what it could launch that it (00:34:37) really missed the web Revolution for a (00:34:39) long time they they have since recovered (00:34:41) and I of course was happy to exploit (00:34:43) that as a competitor to them when we (00:34:45) were at Google but but the important (00:34:47) thing is when big companies should be (00:34:50) faster because they have more money and (00:34:51) more scale they should be able to do (00:34:53) things even quicker but in my industry (00:34:56) anyway the the tech start that have a (00:34:58) new clear idea tend to win because the (00:35:02) big company can't move fast enough to do (00:35:05) it another example we had built (00:35:07) something called Google video I was very (00:35:09) proud of Google video and David Drummond (00:35:12) who was the general counsel at the time (00:35:13) came in and said you have to look at (00:35:14) this YouTube people I said like why (00:35:16) right who cares and it turns out they're (00:35:19) really good and they're more clever than (00:35:21) your team and I said that can't be true (00:35:23) you know typical arrogant Eric and we (00:35:27) sat down and we looked at it and they (00:35:29) really work quicker even though we had (00:35:30) an (00:35:32) incumbent and why it turns out that the (00:35:35) incumbent was operating under the (00:35:37) traditional rules that Google had which (00:35:38) was fine and the competitor in this case (00:35:42) YouTube was not constrained by that they (00:35:43) could work at any pace and they could do (00:35:45) all sorts of things intellectual (00:35:47) property and so forth ultimately we were (00:35:49) sued all over all of that stuff and we (00:35:50) ultimately won all those suits but it's (00:35:52) an example where there are these moments (00:35:54) in time where you have to move extremely (00:35:57) quickly you're seeing that right now (00:36:00) with generative uh technology so the AGI (00:36:03) the generative Revolution generate code (00:36:05) generate videos generate text generate (00:36:08) everything all of those winners are (00:36:10) being determined in the next six 12 (00:36:12) months and then once once the slope is (00:36:15) set once the growth rate is you know (00:36:17) quadrupling every uh six months or so (00:36:19) forth it's very hard for somebody else (00:36:21) to come in so so it's a race to get (00:36:24) there as fast as you can so when you (00:36:27) talk to the the great Venture (00:36:29) capitalists they are they're fast right (00:36:32) we'll look at it we'll make a decision (00:36:33) tomorrow we're done we're in and so (00:36:35) forth and we want to be (00:36:37) first because that's where they make the (00:36:39) most amount of (00:36:40) money we were talking before you arrived (00:36:42) I was talking to Jack about this idea of (00:36:45) like harvesting and hunting so (00:36:47) harvesting what you've already sewed and (00:36:49) hunting for new opportunities but I've (00:36:51) always found it's quite difficult to get (00:36:54) the Harvesters to be the hunters at the (00:36:56) same time so so Harvesters and hunting (00:36:58) is a good metaphor um I'm interested in (00:37:01) entrepreneurs and so what we learned at (00:37:03) Google was ultimately if you want to get (00:37:05) something done you have to have somebody (00:37:06) who's entrepreneurial in their approach (00:37:08) in charge of a small business and so for (00:37:11) example Sundar when he became CEO had a (00:37:13) model of which were the little things (00:37:15) that he was going to emphasize and which (00:37:17) were the big things some of those little (00:37:18) things are now big things right and and (00:37:21) he managed it that way so one way to (00:37:23) understand innovation in a large company (00:37:25) is you need to know who the owner is (00:37:26) Larry Page would say over and over again (00:37:29) it's not going to happen unless there's (00:37:30) an owner who's going to drive this and (00:37:32) he was supremely good at identifying (00:37:35) that technical Talent right that's one (00:37:37) of his great founder strengths so when (00:37:39) we talk about Founders not only do you (00:37:41) have to have a vision but you also have (00:37:43) to have either great luck or great skill (00:37:46) as to who is the person who can lead (00:37:49) this inevitably those people are highly (00:37:51) technical in the sense that they can and (00:37:54) very quick moving and they have good (00:37:56) management skills right they understand (00:37:58) how to hire people and deploy resources (00:38:00) that allows for Innovation um most of (00:38:03) the if I if I look back in my career (00:38:06) each generation of the tech companies (00:38:08) failed including for example Sun at at (00:38:12) the point at which it became (00:38:13) noncompetitive with the future is it (00:38:16) possible for a team to innovate while (00:38:17) they still have their day job which is (00:38:20) harvesting if you know what I mean or do (00:38:21) you have to take those people put them (00:38:23) into a different team different building (00:38:25) different p&l and get them to focus on (00:38:27) the disrupt div evation there are almost (00:38:29) no examples of doing it simultaneously (00:38:31) in the same building uh the Macintosh (00:38:34) was famously um Steve in his typical (00:38:38) crazy way had the this very small team (00:38:41) that invented the Macintosh and he put (00:38:42) them in a little building next to the (00:38:44) big building uh on bub Road and and um (00:38:48) Cupertino and they put a pirate flag on (00:38:51) top of (00:38:52) it now was that good culturally inside (00:38:55) the company no because because it (00:38:58) created resentment in the big building (00:39:00) but was it right in terms of the revenue (00:39:03) and path of of Apple absolutely why (00:39:06) because the Mac ultimately became the (00:39:08) platform that established the UI the (00:39:10) user interface ultimately allowed them (00:39:12) to build the iPhone which of course is (00:39:14) defined by its user interface why (00:39:15) couldn't they stay in the same building (00:39:17) it just doesn't work you you can't get (00:39:20) people to play two roles the incentives (00:39:22) are different if you're going to be a (00:39:24) pirate and a disruptor you don't have to (00:39:26) follow the same rules (00:39:28) so um there there are plenty of examples (00:39:31) where you just have to keep inventing (00:39:33) yourself now what's interesting about (00:39:35) cloud computing and essentially cloud (00:39:37) services which is what Google does is (00:39:40) because the product is not sold to you (00:39:42) it's delivered to you it's easier to (00:39:45) change but the same problem remains if (00:39:47) you look at Google today right it's (00:39:49) basically a search a search box and it's (00:39:52) incredibly powerful but what happens (00:39:54) when that interface is not really (00:39:56) textual right will have to reinvent that (00:39:59) working on Tech it'll be the system will (00:40:02) somehow know what you're asking right it (00:40:04) will it just it will be your assistant (00:40:07) um and again Google will do very well so (00:40:09) I'm in no way criticizing Google here (00:40:11) but I'm saying that even something as (00:40:12) simple as the search box will eventually (00:40:15) be replaced by something more powerful (00:40:17) it's important that Google be the (00:40:19) company that does that I believe they (00:40:20) will and I I was thinking about it you (00:40:22) know the example of Steve Jobs and that (00:40:24) building with the pirate flag on it my (00:40:27) brain when (00:40:29) um there's so many offices around the (00:40:33) world that were trying to kill Apple at (00:40:35) that exact moment that might not have (00:40:37) had the pirate flag but that's exactly (00:40:39) what they were doing in similar small (00:40:40) rooms so what Apple had done so smartly (00:40:43) there was they owned the people that (00:40:45) were about to kill their business model (00:40:47) and this is quite difficult to do and (00:40:49) part of me wonders if in your experience (00:40:52) it's a Founder that has that type of (00:40:54) conviction that does that it's extremely (00:40:57) hard for non-founders to do this in (00:40:59) corporations because if you think about (00:41:01) a (00:41:02) corporation what's the duty of the CEO (00:41:05) many there's the shareholders there's (00:41:07) the employees there's the community and (00:41:09) there's a board trying to get a board of (00:41:13) very smart people to agree on anything (00:41:14) is hard enough so imagine I walk in to (00:41:17) you and I say I have a new idea I'm (00:41:20) going to kill our profitability for two (00:41:23) years it's a huge bet and I need1 (00:41:26) billion (00:41:28) now would the board say yes well they (00:41:32) did to Mark (00:41:34) Zuckerberg he spent all that money on um (00:41:37) essentially VR of one kind or another (00:41:39) doesn't seem to have produced very much (00:41:41) but at exactly the same time he invested (00:41:44) very heavily in Instagram WhatsApp and (00:41:47) Facebook and in particular in the AI (00:41:50) systems that power them and today (00:41:52) Facebook to my surprise is a very (00:41:55) significant leader in AI having released (00:41:57) this uh language called or version (00:41:59) called llama 400 billion which is (00:42:01) curiously an open source model open (00:42:03) source means it's available freely for (00:42:05) everyone and what what Facebook and meta (00:42:08) is saying is as long as we have this (00:42:10) technology we can maximize the revenue (00:42:12) in our core businesses so there's a good (00:42:15) example and uh and Zuckerberg is (00:42:17) obviously an incredibly talented (00:42:19) entrepreneur um he's now back on the (00:42:21) list of the most rich people um he's (00:42:24) feeded at you know and everything he was (00:42:26) doing and he managed to lose all that (00:42:28) money while making a different bet (00:42:30) that's a unique founder the same thing (00:42:33) is almost impossible with a hired (00:42:36) CEO how important here is focus and (00:42:39) what's your your sort of opinion of um (00:42:42) the importance of focus from your (00:42:43) experience with Google but also looking (00:42:44) at these other companies because when (00:42:46) you're at Google and you have so much (00:42:47) money in the bank there's so many things (00:42:49) that you could do and could build like (00:42:51) an endless list you can take on anybody (00:42:52) and basically win in most markets how do (00:42:55) you think about focus at Google (00:42:58) focus is important but it's (00:43:03) misinterpreted in Google we spent an (00:43:05) awful lot of time telling people we (00:43:08) wanted to do everything and everyone (00:43:10) said you can't pull off everything and (00:43:12) we said yes we can we have the (00:43:14) underlying architectures we have the (00:43:16) underlying reach we can do this if we (00:43:18) can imagine and build something that's (00:43:20) really transformative and so the idea (00:43:22) was not that we would somehow focus on (00:43:25) one thing like search but rather that we (00:43:27) would pick areas of great impact and (00:43:29) importance to the world many of which (00:43:30) were free by the way this is not (00:43:32) necessarily Revenue driven and that (00:43:33) worked I'll give you another example (00:43:35) there's an old saying in the business (00:43:38) school that you should focus on on what (00:43:41) you're good at and you should simplify (00:43:43) your product lines and you should get (00:43:44) rid of product lines that don't work (00:43:47) Intel famously had a the term is called (00:43:52) arm it's a risk uh chip and this (00:43:55) particular risk chip was not compatible (00:43:58) with the architecture that they were (00:43:59) using for most of their products and so (00:44:02) they sold it unfortunately this was a (00:44:05) terrible mistake because the (00:44:07) architecture that they sold off was (00:44:09) needed for mobile phones with low memory (00:44:11) with small batteries and and heat (00:44:14) problems and so forth and so on and so (00:44:16) that decision that faithful decision now (00:44:19) 15 years ago meant that they were never (00:44:21) a player in the mobile space and once (00:44:24) they made that decision they tried to (00:44:25) take their expensive and expensive and (00:44:28) complex chips and they kept trying to (00:44:30) make cheaper and smaller versions but (00:44:32) the core decision which was to simplify (00:44:35) simplify to the wrong outcome today if (00:44:38) you look at I'll give you an example the (00:44:40) Nvidia chips use an arm CPU and then (00:44:44) these two powerful uh gpus it's called (00:44:46) the b200 they don't use the Intel chip (00:44:49) they use the arm chip because it was for (00:44:51) their needs faster I would never have (00:44:53) predicted that 15 years ago so at the (00:44:56) end maybe it was just a mistake but (00:44:59) maybe they didn't understand in the way (00:45:02) they were organized as a corporation (00:45:04) that ultimately battery power would be (00:45:06) as important as computing power right (00:45:09) the amount of battery you use and that (00:45:10) was the discriminant so one way to think (00:45:12) about it is if you're going to have (00:45:14) these sort of simple rules you better (00:45:16) have a model of what happens in the next (00:45:18) five years so the way I teach this is (00:45:22) just write down what it'll look like in (00:45:24) five years just try what will look like (00:45:27) in five years your company or whatever (00:45:29) it is right so let's talk about AI what (00:45:31) will be true in five (00:45:34) years that it's going to be a lot (00:45:36) smarter than it is be a lot smarter but (00:45:38) how many companies will there be in AI (00:45:41) will there be five or 5,000 or 50,000 (00:45:45) 50,000 how many big companies will there (00:45:47) be will there be new companies what will (00:45:50) they do right so I just told you my view (00:45:54) is that eventually you and I will have (00:45:57) our own AI assistant which is a polymath (00:46:00) which is incredibly smart which helps us (00:46:03) guide through the information overload (00:46:04) that it is today who's going to build it (00:46:07) make a prediction what kind of hardw (00:46:09) will be on make a prediction how fast (00:46:11) will the networks be make a prediction (00:46:14) write all these things down and then (00:46:16) have a discussion about what to do that (00:46:19) what is interesting about our industry (00:46:21) is that when something like the PC comes (00:46:23) along or the internet I lived through (00:46:25) all of these things they are are such (00:46:28) broad phenomena that they really do (00:46:30) create a whole new Lake a whole new (00:46:32) ocean whatever metaphor you want now (00:46:35) people said well wasn't that crypto no (00:46:39) crypto is not such a platform crypto is (00:46:41) not transformative to daily life for (00:46:44) everyone people are not running around (00:46:46) all day using crypto tokens rather than (00:46:48) currency crypto is a specialized Market (00:46:50) by the way it's important and it's (00:46:52) interesting it's not a horizontal (00:46:54) transformative Market the arrival of (00:46:56) alien intelligence in the form of savant (00:46:58) that you use is such a transformative (00:47:01) thing because it touches everything it (00:47:02) touches you as a a producer as a star as (00:47:05) a narrative it touches me as an (00:47:07) executive um it will ultimately help (00:47:10) people make money in the stock market (00:47:11) people are working on that there's so (00:47:14) many ways in which the technology is (00:47:16) transformative to start you in your case (00:47:19) when you think about your company (00:47:20) whether it's little you know itty bitty (00:47:22) or a really big one it's fundamentally (00:47:24) how will you apply AI to accelerate what (00:47:27) you're doing right in your case for (00:47:29) example here you have I think the most (00:47:31) successful show in the UK by far right (00:47:35) so how will you use AI to make it more (00:47:37) successful well you can ask it to (00:47:39) distribute you more right to make uh (00:47:42) narratives to summarize uh to to come up (00:47:44) with new insights to suggest uh to have (00:47:47) fun to create contest there all sorts of (00:47:49) ways that you can ask AI um I'll give (00:47:51) you a simple example if I were a (00:47:54) politician thankfully I'm not um and I (00:47:57) knew my district I would say uh to the (00:48:00) computer write a program so I'm saying (00:48:02) to the computer you write a program (00:48:04) which goes through all the constituents (00:48:06) in my interest figures out roughly what (00:48:08) they care about and if and then send (00:48:11) them a video which is labeled you know (00:48:14) of me digitally so I'm not fake but it's (00:48:16) kind of like my intention where I (00:48:18) explain to them how important I as their (00:48:20) constituent have made the bridge work (00:48:23) right and you sit there and you go (00:48:24) that's crazy but it's possible (00:48:28) now politicians have not discovered this (00:48:29) yet but they will because ultimately (00:48:32) politicians are around a human (00:48:34) connection and the quickest way to have (00:48:35) that communication is to be on their (00:48:37) phone talking to them about something (00:48:39) that they care about when chat GPT first (00:48:42) launched and they sort of scaled rapidly (00:48:44) to 100 million users there was all these (00:48:45) articles saying that um the founders of (00:48:48) Google had rushed back in and it was a (00:48:50) crisis situation at Googled and there (00:48:52) was panic and there was two things that (00:48:53) I thought first is is that true and (00:48:55) second thing was (00:48:57) how did Google not come to Market first (00:49:00) with a chat GPT style product well well (00:49:03) remember that Google also that's the old (00:49:05) question of why did you not do Facebook (00:49:07) well the answer is we were doing (00:49:08) everything else right so my defensive (00:49:12) answer is that Google has eight or nine (00:49:14) or 10 billion user clusters of activity (00:49:17) which is pretty good right it's pretty (00:49:19) hard to do right I'm very proud of that (00:49:21) I'm very proud of what they're doing now (00:49:24) um my own view is that what happened was (00:49:26) Google was was working in the engine (00:49:29) room and a team out of open AI figured (00:49:33) out a technology called rhf and what (00:49:36) happened was when they did gpt3 and GP (00:49:39) the t is Transformer which was invented (00:49:41) at Google when they did it they had sort (00:49:43) of this interesting idea and then they (00:49:47) own then they sort of casually started (00:49:49) to use humans to make it better and rhf (00:49:53) refers to the fact that you use humans (00:49:55) at the end to do ab tests (00:49:58) where humans can actually say well this (00:49:59) one's better and then the system learns (00:50:02) recursively from Human training at the (00:50:04) end that was a real breakthrough right (00:50:07) and uh I joke with my open a eye friends (00:50:10) that you were sitting around on on (00:50:12) Thursday night and you turn this thing (00:50:13) on and you go holy crap look how good (00:50:16) this thing is it was a real Discovery (00:50:19) right that none of us expected certainly (00:50:21) I did not um and once they had it um the (00:50:25) opening eye people Sam and and and so (00:50:28) forth we'll talk about this they didn't (00:50:30) really understand how good it was they (00:50:32) just turned it on and all of a sudden (00:50:34) they had this huge success disaster (00:50:36) because they were working on GPT 4 at (00:50:38) the same time it was an afterthought (00:50:40) it's a great story because it just shows (00:50:42) you that even the brilliant Founders do (00:50:45) not necessarily understand how powerful (00:50:47) what they what they've done is now today (00:50:50) of course you have uh GPT 40 um (00:50:54) basically a very powerful model from (00:50:56) open eye you have Gemini 1.5 which is (00:50:59) clearly in clearly roughly equivalent if (00:51:01) not better in certain areas um the (00:51:04) Gemini is more multimodal for example (00:51:06) and then you have other players llama (00:51:08) the Llama architecture l l la ma uh does (00:51:12) not stand for llamas it's large language (00:51:15) models um out of Facebook and a number (00:51:17) of others uh there's a startup called (00:51:19) anthropic um which is very powerful (00:51:22) founded by one of the inventors of gpt3 (00:51:25) um and a whole bunch of people and they (00:51:27) formed their company knowing they were (00:51:29) going to be that successful it's (00:51:31) interesting they actually formed as part (00:51:32) of their incorporation that they were a (00:51:34) public benefit Corporation because they (00:51:36) were concerned that it would be so (00:51:37) powerful that some evil CEO in the (00:51:40) future would force them to go for (00:51:41) Revenue as opposed (00:51:42) to world world goodness so the teams (00:51:46) when they were doing this they (00:51:48) understood the power of what they were (00:51:49) doing and they anticipated the level of (00:51:51) impact which and they were right do you (00:51:54) think if Steve Jobs was an apple they (00:51:55) would be on that list (00:51:58) um how do you think the company would be (00:52:01) different well Tim has done a fantastic (00:52:04) job in Steve's Legacy and what's (00:52:06) interesting is normally the successor is (00:52:08) not as good as the founder but somehow (00:52:10) Tim having worked with Steve for so long (00:52:12) and having set the culture having Steve (00:52:14) having they've managed to continue the (00:52:16) focus on the user this incredible safety (00:52:19) focus in terms of apps and so forth and (00:52:21) so on and they've remained a relatively (00:52:24) closed culture I think all of those (00:52:25) would have maintained detained had St (00:52:28) you know tragically died uh he was a (00:52:31) good friend but the important point (00:52:33) is Steve Steve believed very strongly in (00:52:38) what are called close systems where you (00:52:40) own and control all your intellectual (00:52:41) property and he and I would battle over (00:52:43) open versus closed because I came from (00:52:45) the other side and I did this with (00:52:47) respect I don't think they would have (00:52:48) changed that and they've change that now (00:52:51) no I think still apple is still (00:52:54) basically a single company that's ver (00:52:57) Ally integrated the rest of the industry (00:52:59) is largely more open I think everyone (00:53:01) especially in the wake of the recent (00:53:03) launch of the iPhone 16 which I've got (00:53:05) somewhere here um has this expectation (00:53:08) that Apple would if Steve were still (00:53:10) alive taken some big bold bet in some (00:53:13) and I think about you know Tim's tenure (00:53:15) he's done a fantastic job of keeping (00:53:17) that company going running it with the (00:53:19) sort of principles of Steve Jobs but has (00:53:22) there been many big bold successful bets (00:53:24) a lot of people point at the airpods (00:53:25) which have a a great product (00:53:27) but I think AI is one of those things (00:53:29) where you go I wonder if Steve would (00:53:31) have understood the significance of it (00:53:33) and Steve was that smart that he I would (00:53:37) never you know he's an Elon level (00:53:39) intelligence (00:53:42) um when when Steve and I worked together (00:53:45) very closely which was what 15 years ago (00:53:47) for his death um he was very frustrated (00:53:52) at the success of MP4 over uh mov (00:53:57) um format files and he was really mad (00:54:01) about it and I said well you know maybe (00:54:04) that's because you were closed in quick (00:54:05) time was not generally available said (00:54:07) that's not true my team you know our (00:54:09) product is better and so forth so his (00:54:11) his core belief system he's an artist (00:54:15) right and and given the choice we used (00:54:17) to have this debate where do you want to (00:54:19) be Chevrolet or do you want to be (00:54:20) Porsche do you want to be you know (00:54:22) General Motors or do you want to be BMW (00:54:24) and he said I want to be BMW (00:54:27) and during that time Apple's margins (00:54:30) were twice as high as the PC companies (00:54:33) and I said Steve you don't need all that (00:54:35) money you're generating all this cash (00:54:37) you're giving it to your to your (00:54:38) shareholders and he said the principle (00:54:41) of our profitability and our value in (00:54:43) our brand is this is this luxury brand (00:54:47) right so that's how he thought now what (00:54:50) How would how would AI change that (00:54:52) everything that he would have done with (00:54:54) Apple today would be a I inspired but it (00:54:58) would be beautiful that's the great gift (00:55:00) he had CU I think Siri was almost a (00:55:04) glimpse at what AI now kind of looks (00:55:06) like it was a glimpse at what the I (00:55:08) guess the ambition was we've all been (00:55:10) chatting to the Siri thing which is I (00:55:11) think most people would agree as kind of (00:55:12) like largely useless unless you're (00:55:13) trying to figure out something super (00:55:15) super simple but now I this weekend as I (00:55:18) said I was sat there with my my (00:55:20) girlfriend's family there speaking to (00:55:22) this voice activated device and it was (00:55:24) solving problems for me almost (00:55:25) instantaneously that are very complex (00:55:27) and translating them into French and (00:55:28) Portuguese welcome welcome to the (00:55:30) replacement for Siri and again would (00:55:32) Steve have done that quicker I don't (00:55:34) know it's very clear that the first (00:55:36) thing Apple needs to do is have Siri be (00:55:41) replaced by an AI and call that Siri (00:55:44) hiring we we're doing a lot of hiring in (00:55:46) our companies at the moment and we're (00:55:47) going back and forward on what the most (00:55:49) important principles are when it comes (00:55:50) to hiring making lots of mistakes (00:55:52) sometimes getting things right (00:55:54) sometimes what do I need to know as when (00:55:57) it comes to hiring startups by (00:55:59) definition are huge Risk Takers you have (00:56:02) no history you have no incumbency you (00:56:04) have all these competitors by definition (00:56:06) and you have no time so in a startup you (00:56:09) want to you want to um prioritize (00:56:12) intelligence and quickness over (00:56:15) experience and sort of stability you (00:56:18) want to take risks on people and the (00:56:21) great and part of the reason why (00:56:22) startups are full of young people is (00:56:24) because young people often don't have (00:56:26) the baggage of Executives have been (00:56:27) around for a long time but more (00:56:29) importantly they're willing to take (00:56:30) risks so it used to be that you could (00:56:34) predict whether a company was successful (00:56:36) by the age of the founders and in that (00:56:39) 20 and 30y old period the company would (00:56:41) be hugely successful startups um Wiggle (00:56:45) they try something they try something (00:56:46) else and they're very quick to discard (00:56:49) an old idea corporations spend years (00:56:52) with a belief system that is factually (00:56:54) false and they don't actually changed (00:56:57) their opinion until after they've lost (00:56:59) all the contracts and if you go back the (00:57:02) all the signs were there nobody wanted (00:57:04) to talk to them nobody cared about the (00:57:05) product right and yet they kept pushing (00:57:08) it so um if you're a CEO of a larger (00:57:11) company what you want to do is basically (00:57:13) figure out how to measure this (00:57:15) Innovation so that you don't waste a lot (00:57:17) of time Bill Gates had a saying a long (00:57:18) time ago which was that the most (00:57:21) important thing to do is to fail fast (00:57:23) right that the charact from his (00:57:24) perspective as the CEO of Microsoft (00:57:26) founder Microsoft um that he wanted (00:57:29) everything to happen and he wanted to (00:57:30) fail quickly and that was his theory and (00:57:33) do you agree with that theory yeah I do (00:57:36) fast failure is important because you (00:57:38) can say it in a nicer way but (00:57:40) fundamentally um at Google we had this (00:57:42) 72010 rule that Larry and Sergey came up (00:57:45) with 70% of the Core Business 20% on (00:57:48) adjacent business and 10% on other what (00:57:50) does that mean sorry cor Core Business (00:57:52) means search ads adjacent business means (00:57:55) something that you're trying like a (00:57:56) cloud business or so forth and the 10% (00:57:58) is some new idea so Google created this (00:58:01) thing called Google X the first product (00:58:04) it built was called Google brain which (00:58:06) is the one of the first machine learning (00:58:08) architectures this actually precedes (00:58:09) Deep Mind Google brain was used to power (00:58:12) the AI system Google brin's team of 10 (00:58:15) or 15 people generated 10 20 30 40 (00:58:18) billion dollars of extra profits over a (00:58:20) decade so that pays for a lot of (00:58:23) failures right then they had a whole (00:58:25) bunch of other ideas that seemed very (00:58:27) interesting to me that didn't happen for (00:58:29) one or another and they would cancel (00:58:31) them and you you and then the people (00:58:34) would get reconfigured and one of the (00:58:36) great things about Silicon Valley is (00:58:37) it's possible to spend a few years on a (00:58:40) really bad idea and get cancelled if you (00:58:43) will and then get another job Having (00:58:44) learned all of that my joke is the best (00:58:47) CFO is one who's just gone bankrupt (00:58:50) because the one thing that CFO is not (00:58:51) going to let happen is to go bankrupt (00:58:53) again yeah well on this point of culture (00:58:56) as well Google as such a big company (00:58:59) must (00:59:00) experience a bunch of microcultures one (00:59:03) of the things that I've always I've kind (00:59:04) of studied it as an as a cautionary tale (00:59:07) is the story of TGIF at Google which was (00:59:10) this sort of weekly All Hands meeting (00:59:13) where employees could ask the executives (00:59:15) whatever they wanted to and the Articles (00:59:17) around it say that it was eventually (00:59:18) sort of changed or canceled because it (00:59:20) became (00:59:21) unproductive it's more complicated than (00:59:24) that so lar and serus started TGF (00:59:27) uh which I obviously participated in and (00:59:29) we had fun uh there was a sense of humor (00:59:31) it was all off the Record um a famous (00:59:34) example is the VP of sales whose name (00:59:36) was Omid um was always predicting lower (00:59:41) Revenue than we really had which is (00:59:42) called sandbagging so we got a sandbag (00:59:45) and we made him stand on the sandbag in (00:59:47) order to present his numbers it was just (00:59:49) fun humorous you know we had skits and (00:59:51) things like that um at at some size you (00:59:55) don't have that level of intim intimacy (00:59:57) and you don't have a level of privacy (00:59:59) and what happened was there were leaks (01:00:03) uh eventually there was a presentation I (01:00:06) don't remember the specifics where the (01:00:08) Pres presentation was ongoing and (01:00:10) someone was leaking the presentation (01:00:12) live to a reporter and somebody came on (01:00:16) stage and said we have to stop now I (01:00:19) think that was the moment where the (01:00:20) company got sort of too (01:00:23) big (01:00:25) h I heard about a story that um because (01:00:29) from what I had understood this might be (01:00:30) totally wrong but it's all just things (01:00:31) that Google employees have told me was (01:00:33) that there wasn't many sackings firings (01:00:36) at Google's wasn't many layoffs wasn't (01:00:38) really a culture of layoffs and I guess (01:00:40) I guessed in part that's because the (01:00:41) company was so successful that it didn't (01:00:43) have to make those extremely extremely (01:00:44) tough decisions that we're seeing a lot (01:00:46) of companies make today I reflect on (01:00:48) elon's running of Twitter when he take (01:00:51) took over Twitter the you know the say (01:00:53) the The Story Goes that he went to the (01:00:55) top floor and basically said anyone (01:00:57) who's willing to work hard is committed (01:00:59) to these values please come to the top (01:01:01) floor everyone else you're fired um this (01:01:03) sort of extreme culture of culling and (01:01:06) people being sort of activists at work (01:01:09) um and I wanted to know if there's any (01:01:11) truth in that there's some um in in (01:01:15) Google's case (01:01:17) um we had a position of why lay people (01:01:21) off just don't hire them in the first (01:01:22) place it's much much easier and so in my (01:01:26) 10 year the only layoff we did was uh (01:01:29) 200 people in the sales structures right (01:01:31) after the 2000 epidemic and I remember (01:01:33) it as being extremely painful right it (01:01:36) was the first time we had done it so we (01:01:38) took the position which is different at (01:01:40) the time that you shouldn't have an (01:01:43) automatic layoff what would happen is (01:01:45) that there was a belief at the time that (01:01:47) every six months or nine months you (01:01:49) should take the bottom five% of your (01:01:51) people and lay them off problem with (01:01:53) that is you're assuming the 5% are (01:01:54) correctly identified and furthermore (01:01:57) even the lowest performers have (01:01:58) knowledge and value to the corporation (01:02:00) that we can take it so we took a a very (01:02:02) much more positive view of our employees (01:02:04) and the employees like that and we (01:02:05) obviously paid them very well and so (01:02:07) forth and so on I think that the the (01:02:09) cultural issues ultimately have been (01:02:12) addressed but during there was a period (01:02:13) of time where there were uh because of (01:02:17) the free willing nature nature of the (01:02:19) company there were an awful lot of (01:02:21) internal distribution lists which had (01:02:23) nothing to do with the company what does (01:02:25) that mean they were distribution lists (01:02:27) on topics of War peace politics so forth (01:02:31) what's a distribution list a (01:02:32) distribution like an email dist think of (01:02:34) it as a a message board okay roughly (01:02:38) speaking think of it as message boards (01:02:39) for employees and at one I remember that (01:02:42) one point somebody discovered that there (01:02:43) were 100,000 such me message boards and (01:02:46) the company ultimately cleaned that up (01:02:48) because companies are not like (01:02:50) universities and that there are in fact (01:02:52) all sorts of laws about what you can say (01:02:54) and what you cannot say and so forth and (01:02:56) so for example the majority of the (01:02:58) employees were uh Democrats in the (01:03:00) American political system and I made a (01:03:02) point even though I'm a Democrat to try (01:03:04) to protect the small number of (01:03:06) Republicans because I thought they had a (01:03:07) right to be employees too so you have to (01:03:09) be very careful in a corporation to (01:03:12) establish what what does speech mean (01:03:14) within the corporation and uh what you (01:03:18) what you are hearing as wokeism is (01:03:20) really can be understood is what are the (01:03:22) appropriate topics on work time in in a (01:03:25) work venue should you be discussing my (01:03:28) own view is stick to the business and (01:03:30) then please feel free to go to the bar (01:03:33) scream your views talk to everybody you (01:03:35) know I'm a strong believer in free (01:03:36) speech but within the corporation let's (01:03:38) just stick to the corporation and its (01:03:39) goals because I was hearing these (01:03:41) stories about I think in more recent (01:03:42) times in the last year or two of people (01:03:44) coming to work just for the free (01:03:46) breakfast Pro protesting outside that (01:03:48) morning coming back into the building (01:03:49) for lunch as best I can tell that's all (01:03:51) been cleaned (01:03:52) up I did also hear that that it had been (01:03:56) cleaned up because I think it was (01:03:57) addressed in a very high conviction way (01:04:00) which meant that it it was um seen to (01:04:02) how did how do you think about (01:04:03) competition for everyone that's building (01:04:05) something how much should we be focusing (01:04:07) on our comp competition I strongly (01:04:09) recommend not focusing on competition (01:04:11) and instead focusing on building a (01:04:12) product that no one else has and you say (01:04:14) well how can you do that without knowing (01:04:15) the competition well if you study the (01:04:17) competition you're wasting your time try (01:04:18) to solve the problem in a new way and do (01:04:20) it in a way where the customers are (01:04:22) delighted U running Google we seldom (01:04:25) looked at what our competitors were (01:04:27) doing what we did we spent an awful lot (01:04:29) of time was what is possible for us to (01:04:31) do what can we actually do from our (01:04:33) current situation and sort of the (01:04:36) running ahead of everybody turns out to (01:04:38) be really important what about (01:04:41) deadlines well uh Larry established the (01:04:44) principle of um okrs which were (01:04:47) objectives and key results in every (01:04:49) quarter Larry would actually write down (01:04:51) all the metrics and he was tough and he (01:04:54) would say that if you got to 70% % of my (01:04:56) numbers that was good and then we would (01:04:59) grade based on are you above the 70% or (01:05:01) you below the 70% and it was harsh and (01:05:04) it works you you have to measure to get (01:05:08) things done in big Corporation otherwise (01:05:10) everyone kind of looks good makes all (01:05:13) sorts of claims feels good about (01:05:14) themselves but it doesn't have an impact (01:05:17) what about business plans should we be (01:05:19) writing business plans as found us (01:05:21) Google wrote A business plan there was a (01:05:22) run by a fellow named solar and I saw it (01:05:25) years later and it was actually correct (01:05:28) and I told salar that the this is (01:05:30) probably the only business plan ever (01:05:32) written for a corporation that was (01:05:33) actually correct in hindsight so what I (01:05:37) prefer to do and this is how I teach it (01:05:39) at Stanford is try to figure out what (01:05:42) the world looks like in five years and (01:05:44) then try to figure out what you're going (01:05:46) to do in one year and then do it right (01:05:50) so if you can basically say this is the (01:05:53) direction these are the things we're (01:05:54) going to achieve within one year and (01:05:57) then run against that as hard goals not (01:05:59) simple goals but hard goals then you'll (01:06:01) get there and the general rule at least (01:06:04) in a consumer business is if you can get (01:06:05) an audience of 10 or 100 million people (01:06:07) you can make lots of money right so if (01:06:09) you give me any business that has no (01:06:11) revenue and a 100 million people I can (01:06:13) find a way to to monetize that with (01:06:15) advertising and sponsorships and (01:06:17) donations and so forth and so on focus (01:06:19) on getting the user right and everything (01:06:22) else will follow the Google phrase is (01:06:24) focus on the user and everything else is (01:06:27) handled Sergey and (01:06:30) Larry you work with them for 20 years (01:06:34) many decades yeah two decades what made (01:06:36) them special frankly raw IQ they were (01:06:39) just smarter than everybody else really (01:06:41) yeah and (01:06:43) uh in sergey's case his father was a (01:06:46) very brilliant Russian mathematician his (01:06:48) mother was also highly technical his (01:06:50) family is all very technical and he was (01:06:52) clever he's a clever (01:06:54) mathematician uh Larry (01:06:56) different personality but similar so an (01:06:58) example would be that Larry and I are in (01:07:01) his office and we're writing on the (01:07:02) Whiteboard a long list about what we're (01:07:04) going to do and he says look we're going (01:07:05) to do this and this and I said okay I (01:07:07) agree with you I don't agree with you we (01:07:08) make this very long list and Sergey is (01:07:11) out playing (01:07:12) volleyball and so he runs in in his (01:07:15) little volleyball shorts and his little (01:07:16) shirt all sweating he looks at our list (01:07:18) and said this is the stupidest thing (01:07:19) I've ever heard and then he suggest five (01:07:22) things and he was exactly right so we ar (01:07:26) red the Whiteboard and then he of course (01:07:27) went back to play volleyball and that (01:07:29) became the strategy of the company so (01:07:31) over and over again it was the it was (01:07:33) their Brilliance and their ability to (01:07:35) see things that I didn't see that I (01:07:37) think really drove it can you teach that (01:07:39) I don't know I think you can teach (01:07:42) listening and (01:07:44) um but I think most of us get caught up (01:07:47) in our own (01:07:48) ideas and we are always surprised that (01:07:52) something new happened like I've just (01:07:54) told you that I'm I've been in AI a long (01:07:57) time I'm still surprised at the rate uh (01:07:59) my favorite current product is called (01:08:00) notebook (01:08:01) LM and for the uh listeners notebook LM (01:08:04) is an experimental product out of Google (01:08:06) Deep Mind basically Gemini um it's based (01:08:09) on the Gemini back end and it was (01:08:11) trained with high quality podcast voices (01:08:14) it's terrifying and you basically give (01:08:16) it a so what I'll do is um I'll write (01:08:19) something again I don't write very well (01:08:21) and I'll ask Gemini to rewrite it to be (01:08:24) more beautiful okay I'll take that text (01:08:27) and I'll put it in Notebook LM and it (01:08:29) produces this interview between a man (01:08:31) and a woman U who don't exist and for (01:08:34) fun what I do is I play this in front of (01:08:36) an audience and I wait and see if anyone (01:08:39) figures out that the humans are not (01:08:40) human it's so good they don't figure it (01:08:43) out we'll play it now so this is the big (01:08:45) thing that everyone's making a big fuss (01:08:46) about you can go and load this (01:08:48) conversation now it's going to go out (01:08:49) and create a conversation that's in a (01:08:51) podcast style where there's a male voice (01:08:53) and a female voice and they're analyzing (01:08:55) the content and then coming up with (01:08:57) their own kind of just uh creative (01:08:59) content so you could go and push play (01:09:00) right here we are back Thursday get (01:09:03) ready for week three the injury report (01:09:05) this week was a doozy it's a long one (01:09:08) yeah it is and it has the potential to (01:09:11) really shake things up so for that to me (01:09:15) gem notebook LM is my chat GPT moment of (01:09:19) this (01:09:20) year it was mine as well and it's much (01:09:22) of the reason that I was um deeply (01:09:25) confused okay because as a podcaster (01:09:27) who's building a media company we have (01:09:29) an office down the road 25,000 square (01:09:31) feet we have studios in there um we're (01:09:36) building audio video content at this in (01:09:40) the dawn of this new world where the (01:09:43) cost of production of content goes to (01:09:45) like zero or something and I'm trying to (01:09:47) navigate how to play as a media owner so (01:09:49) first place you're you're what's really (01:09:51) going on is you're moving from scarcity (01:09:52) to ubiquity you're moving from scarc to (01:09:56) abundance so one way to understand the (01:09:59) world I live in is it's scale Computing (01:10:01) generates abundance and abundance allows (01:10:03) new strategies in your case it's obvious (01:10:05) what you should do you're a really (01:10:07) famous podcaster and you have lots of (01:10:08) interesting guests simply have this fake (01:10:12) set of podcasts criticize you and your (01:10:14) guests right you're you're essentially (01:10:17) just amplifying your reach they're not (01:10:19) going to substitute for your honest (01:10:22) Brilliance and Charisma here but they're (01:10:24) going to accentuate it they will they (01:10:26) will they will be entertaining they will (01:10:27) summarize it and so forth it amplifies (01:10:30) your reach if you go back to my basic (01:10:32) argument that AI will double the (01:10:34) productivity of everybody or more so in (01:10:37) your case you'll have twice as many co (01:10:40) podcasts what I do for examples I'll (01:10:42) write something and I'll say I'll have (01:10:43) it respond and then to Gemini I'll say (01:10:46) make it longer and it adds more stuff I (01:10:49) think God I do this in like 30 seconds (01:10:52) then how powerful in your case take one (01:10:55) of these uh lengthy interviews you do (01:10:58) ask the system to annotate it to amplify (01:11:01) it and then feed that into fake (01:11:03) podcasters and see what they say you'll (01:11:06) have a whole new set of audiences that (01:11:07) love them more than you but but it's all (01:11:10) from you that's the key idea here I (01:11:14) worry because there's going to be (01:11:15) potentially billions of podcasts that (01:11:18) are uploaded to RSS feeds all around the (01:11:19) world and it's all going to sort of chip (01:11:21) away at you know the the moat that I've (01:11:25) so (01:11:26) so many people have believed that but I (01:11:28) think the evidence is it's not true um (01:11:32) when I started at Google there was this (01:11:33) notion that celebrity would go away and (01:11:35) there would be this very long tale of (01:11:38) micro markets you know Specialists (01:11:41) because finally you could hear the (01:11:43) voices of everyone and we're all very (01:11:44) Democratic and liberal in our view (01:11:46) that's the what really happened was (01:11:49) networks accentuated the best people and (01:11:51) they made more money right you went from (01:11:53) being a local personality to a national (01:11:56) personality to a global personality and (01:11:58) the globe is a really big thing and (01:11:59) there's lots of money and lots of (01:12:01) players so you as a as a celebrity are (01:12:05) competing against a global group of (01:12:07) people and you need all the help you can (01:12:09) to maintain your position if you do it (01:12:11) well by using these AI Technologies you (01:12:13) will become more famous not less (01:12:18) famous (01:12:20) Genesis I am I've had a lot of (01:12:23) conversations with a lot of people about (01:12:24) the subject of AI um and when I read (01:12:26) your book and I've watched you do a (01:12:28) series of interviews on this some of the (01:12:30) quotes that you said really stood out to (01:12:32) me one of them I wrote down (01:12:35) here which comes from your book Genesis (01:12:38) it's on page five the Advent of (01:12:40) artificial intelligence is in our view a (01:12:42) question of human (01:12:46) survival yes that is our view so why is (01:12:50) it a question of human (01:12:54) survival AI is going to move very (01:12:56) quickly it's moving so much more quickly (01:12:59) than I've ever seen because the amount (01:13:01) of money the number of people the impact (01:13:04) the (01:13:05) need what happens when the AI systems (01:13:08) are really running key parts of our (01:13:11) world what happens when AI is making the (01:13:14) decision my my simple example you have a (01:13:16) car which is AI controlled and you have (01:13:20) a emergency or a lady's about to give (01:13:23) birth or something like that and they (01:13:26) get in the car and there's no override (01:13:27) switch because the system is optimized (01:13:30) around the whole as opposed to his or (01:13:33) her (01:13:34) emergency right we as humans accept (01:13:37) various forms of efficiency including (01:13:39) urgent ones versus system systemic (01:13:41) efficiency you could imagine that the (01:13:43) Google Engineers would design a perfect (01:13:45) City that would perfectly operate every (01:13:48) self-driving car on every street but (01:13:51) would not then allow for the exceptions (01:13:53) that you need in such a in such an (01:13:55) important issue so that's a trivial (01:13:58) example and one which is well understood (01:14:01) of how it's important that these things (01:14:03) represent human values right that we we (01:14:06) have to actually articulate what does it (01:14:08) mean so my favorite one is all this (01:14:11) misinformation um democracy is pretty (01:14:14) important democracy is by far the best (01:14:15) way to to live and operate societies (01:14:18) look at there are plenty of examples of (01:14:19) this none of us want to work in (01:14:22) essentially an authoritarian (01:14:23) dictatorship so you better figure out a (01:14:26) way where the misinformation components (01:14:29) do not screw up proper political (01:14:32) examples another example is this (01:14:34) question about teenagers and the develop (01:14:36) their mental development and growing up (01:14:38) into these societies I don't want them (01:14:40) to be constantly depressed there's a lot (01:14:42) of evidence that dates around 2015 when (01:14:46) all the social media algorithms changed (01:14:48) from linear feeds to targeted feeds in (01:14:50) other words they went from time to this (01:14:53) is what you want this is what you want (01:14:55) that hyperfocus has ultimately narrowed (01:14:58) people's um political views as I as we (01:15:00) discussed but more importantly it's (01:15:02) produced more depression and anxiety so (01:15:05) all the studies indicate that basically (01:15:07) if you time it to roughly then when (01:15:09) people are coming to age they're not as (01:15:12) happy with their lives their behaviors (01:15:14) their opportunities for this and the (01:15:16) best explanation is it was an (01:15:18) algorithmic change and remember that (01:15:20) these systems they're not just (01:15:22) collections of content they are (01:15:24) algorithmically deciding (01:15:26) you know the algorithm decides what the (01:15:28) outcome is for humans we have to manage (01:15:31) that um what we say in many different (01:15:34) ways in the book is that you have sort (01:15:36) of a choice of whether the um the (01:15:40) algorithms will advance that's not a (01:15:42) question the question is are we (01:15:44) advancing with it and do we have control (01:15:46) over it um there are so many examples (01:15:48) where you could imagine an AI system (01:15:50) could do something more efficiently but (01:15:53) at what cost right (01:15:56) um I should mention that there is this (01:15:59) discussion about something called AGI (01:16:01) artificial general (01:16:02) intelligence and there's this discussion (01:16:04) in the Press among many people that AGI (01:16:07) occurs on a particular day right and (01:16:09) this is sort of a popular concept that (01:16:11) on a particular day five years from now (01:16:13) or 10 years from now this thing will (01:16:15) occur and all of a sudden we're going to (01:16:16) have a computer that's just like us but (01:16:18) even quicker that's unlikely to be the (01:16:21) path much more likely are these waves of (01:16:25) innovation in every field better (01:16:27) psychologists better writers you see (01:16:29) this with g chat gbt already better (01:16:32) scientists is a notion of an AI (01:16:34) scientist that's working with the AI (01:16:36) real scientists to accelerate the (01:16:38) development of more AI science people (01:16:40) believe all of this will come but it has (01:16:42) to be under human (01:16:44) control do you think it will be I do and (01:16:48) part of the reason is I and others have (01:16:49) worked hard to get the governments to (01:16:51) understand this it's very strange in my (01:16:53) entire career which has gone for you (01:16:55) know 50 years the um we've never asked (01:16:59) for government for help because asking (01:17:01) the government help is basically just a (01:17:03) disaster in the view of the techn (01:17:04) industry in this case the people who (01:17:07) invented it collectively came to the (01:17:09) same view that there need to be (01:17:11) guardrails on this technology because of (01:17:13) the potential for harm the most obvious (01:17:15) one is how do I kill myself give me (01:17:17) recipes to hurt other people that kind (01:17:19) of stuff there's a whole Community now (01:17:21) in this in this part of the industry (01:17:24) which are called trust and safety groups (01:17:26) and what they do is they actually have (01:17:28) humans test the system before it gets (01:17:32) released to make sure the harm that it (01:17:34) might have in it is suppressed it's (01:17:36) literally won't answer the question when (01:17:39) you play this forward in your brain you (01:17:41) you've been in the tech industry for a (01:17:42) long time and from looking at your work (01:17:45) you it feels like you're describing this (01:17:46) as the most sort of transformative (01:17:48) potentially harmful technology that (01:17:50) humans have really ever seen you know (01:17:52) maybe alongside the nuclear bomb I guess (01:17:54) but some would say even potentially (01:17:56) worse because of the nature of the (01:17:58) intelligence and its (01:18:00) autonomy you must have moments where you (01:18:02) you think forward into the future and (01:18:04) your thoughts about that future aren't (01:18:06) so (01:18:06) Rosy well because I have those moments (01:18:09) yes but but let's let's think let's (01:18:10) answer the question I said think five (01:18:12) years in five years you'll have two or (01:18:14) three more turns of the crank of these (01:18:16) large models these large models are (01:18:19) scaling with ability that is (01:18:21) unprecedented there's no evidence that (01:18:23) the scaling has laws as they're called (01:18:26) have begun to to stop they will (01:18:29) eventually stop but we're not there yet (01:18:31) each one of these cranks looks like it's (01:18:33) a factor of two factor of three factor (01:18:35) of four of capability so let's just say (01:18:37) turning the crank all of these systems (01:18:40) get 50 times or 100 times more powerful (01:18:44) in it of itself that's a very big deal (01:18:47) because those systems will be capable of (01:18:48) physics and math you see this with o. (01:18:50) one and um and open AI all the other (01:18:54) things that are occurring (01:18:55) now what are the dangers well there's (01:18:59) the most obvious one is cyber attacks (01:19:00) there's evidence that the raw models (01:19:02) these are the ones that have not been (01:19:04) released can do what are called Day Zero (01:19:06) attacks as well or better than humans a (01:19:08) day Zero attack is an attack that's (01:19:10) unknown they can discover something new (01:19:12) and how do they do it they just keep (01:19:14) trying because they're computers and (01:19:15) they have nothing else to do they don't (01:19:17) sleep they don't eat they just turn them (01:19:18) on and they just keep going um so the so (01:19:21) cyber is an example where everybody's (01:19:23) concerned another one is biology viruses (01:19:25) are relatively easy to make and you can (01:19:28) imagine coming up with really bad (01:19:29) viruses there's a whole team I'm part of (01:19:31) a commission looking at this to try to (01:19:33) make sure that doesn't happen I already (01:19:35) mentioned misinformation (01:19:37) another probably negative but we'll see (01:19:41) is the development of new forms of (01:19:43) warfare I've written extensively on how (01:19:46) war is changing and the way to (01:19:48) understand historic war is that it's the (01:19:51) stereotypically the the soldier with the (01:19:54) gun you know one side and so forth World (01:19:56) War trenches you see this by the way in (01:19:58) UK in the Ukraine fight today where the (01:20:00) ukrainians are holding on valiantly (01:20:02) against the Russian Onslaught but he's (01:20:04) sort of you know mono Amano you know man (01:20:07) against man sort of all of the (01:20:08) stereotypes of War so in a drone World (01:20:12) which is the sort of the fastest way to (01:20:14) build new robots is to build drones (01:20:16) you'll be sitting in a Command Center in (01:20:17) some office building connected by a (01:20:19) network and you'll be doing harm to the (01:20:21) other side while you're drinking your (01:20:23) coffee right that's a changed in the (01:20:25) logic of War um and it's applicable to (01:20:28) both sides I don't think anyone quite (01:20:30) understands how war will change but I (01:20:32) will tell you that in in the Russian (01:20:34) Ukraine war you're seeing a new form of (01:20:37) Warfare being invented right now right (01:20:40) um both sides have lots of drones tanks (01:20:43) are no longer very useful a $5,000 drone (01:20:46) can kill a $5 million tank um so it's (01:20:49) called The Kill ratio so basically it's (01:20:51) drone on drone and so now people are (01:20:53) trying to figure out how how to have one (01:20:55) drone destroy the other drone right this (01:20:58) will ultimately take over war and (01:21:00) conflict in our world in total you (01:21:02) mentioned rural models this is a concept (01:21:04) that I don't think people understand (01:21:06) exists the idea that there's some other (01:21:08) model that's the role model that is (01:21:11) capable of much worse than the thing we (01:21:13) play with on our computers every day (01:21:14) it's important to establish how these (01:21:15) things work so you the way these (01:21:17) algorithms work is they have complicated (01:21:19) uh training things where they suck all (01:21:21) the information in and they uh one week (01:21:25) currently believe we've sort of sucked (01:21:26) all of the written word that's available (01:21:28) it doesn't mean there isn't more but (01:21:30) we've we've literally done such a good (01:21:32) job of sucking everything that humans (01:21:33) have ever written it's all in these big (01:21:35) computers when I say computers I don't (01:21:37) mean computers I mean supercomputers (01:21:39) with enormous memories and the scale is (01:21:42) mindboggling uh and of course there's (01:21:44) this company called Nvidia which makes (01:21:45) the chips which is now one of the most (01:21:47) valuable companies in the world um (01:21:50) surprisingly so incredibly successful (01:21:52) because they're so Central to this (01:21:54) revolution and good for Jensen and his (01:21:56) team so the important thing is when you (01:21:58) do this training it comes out with a raw (01:22:01) model right it takes six months and you (01:22:03) know you wait 24 hours a day you can (01:22:05) watch it it gets close to there's a (01:22:07) measurement that they use called the (01:22:09) loss function when it gets to a certain (01:22:11) number they say good enough so then they (01:22:13) go what do we have right what do we do (01:22:16) right um so the first thing is let's (01:22:19) figure out what it (01:22:20) knows so they have a set of tests and of (01:22:24) course it knows all sorts of bad things (01:22:26) which they immediately then tell it not (01:22:27) to answer to me the most interesting (01:22:30) question is in over a 5-year (01:22:33) period the systems will learn things (01:22:36) that we don't know they learn how will (01:22:39) you test for things that you don't know (01:22:41) they (01:22:42) know the answer in the industry is that (01:22:45) they have incredibly clever people who (01:22:48) sit there and they fiddle literally (01:22:50) fiddle with the networks and say I'm (01:22:52) gonna I'm going to see if it knows this (01:22:55) I'll see if it can do this and then they (01:22:58) make a list and they say that's good (01:23:00) that's not so good right so all of these (01:23:02) Transformations so for example you can (01:23:05) show it a picture of a website and it (01:23:07) can generate the code to generate a (01:23:08) website all of those were not expected (01:23:11) they just happened it's called emergent (01:23:13) Behavior scary scary but exciting and so (01:23:18) far um the systems have held the (01:23:21) governments have worked well um the (01:23:23) these trust and safety groups group are (01:23:25) working here in the UK um one year ago (01:23:28) was the first trust and safety (01:23:30) conference um the government did a (01:23:31) fantastic job the team that was (01:23:34) assembled was the best of all the (01:23:35) country teams here in the UK um now (01:23:39) what's happening is these are happening (01:23:40) around the world the next one is in (01:23:41) France in uh early February and I expect (01:23:44) a similar good result do you think we're (01:23:46) gonna have to guard I mean you talk (01:23:48) about this but do you think we're going (01:23:49) to have to guard these role models with (01:23:52) with guns and tanks and machinery and (01:23:55) stuff I worked for the Secretary of (01:23:56) Defense for a while uh in my in Google (01:23:59) you could spend 20% of your time on (01:24:01) other things so I worked for the (01:24:02) Secretary of Defense to try to (01:24:04) understand the US Military and um one of (01:24:07) the things that we did is we visited a (01:24:09) plutonium U Factory plutonium is (01:24:12) incredibly dangerous and Incredibly (01:24:13) secret and so this particular base is (01:24:16) inside of another base so you go through (01:24:18) the first set of machine guns and then (01:24:20) you have normal thing and then you go (01:24:21) into the special place with even more (01:24:23) machines guns and even because it's so (01:24:25) secure so the the metaphor is do you (01:24:28) fundamentally believe that the computers (01:24:31) that I'm talking about will be of such (01:24:33) value and such danger that they'll have (01:24:35) their own data center with their own (01:24:37) guards which of course might be computer (01:24:39) guards but the important thing is that (01:24:40) it's so special that it has to be (01:24:42) protected in the same way that we (01:24:44) protect nuclear bombs and proliferate uh (01:24:46) and programming an alternative model is (01:24:49) to say that this technology will spread (01:24:52) pretty broadly and there'll be many such (01:24:54) plac (01:24:56) if it's a small number of groups the (01:24:59) governments will figure out a way to do (01:25:01) deterrence and they'll figure out a way (01:25:02) to do (01:25:03) non-proliferation so I'll make something (01:25:05) up I'll say there's a couple in China (01:25:07) there's a few in the US there's one in (01:25:09) in Britain of course we're all tied (01:25:10) together between the US and Britain and (01:25:12) maybe in a few other places that's a (01:25:14) manageable problem on the other hand (01:25:16) let's imagine that that power is (01:25:18) ultimately so easy to copy that it (01:25:21) spreads globally and it's accessible to (01:25:24) for example terrorist (01:25:25) then you have a very serious (01:25:27) proliferation problem which is not yet (01:25:29) solved this is again (01:25:31) speculation because I think a lot about (01:25:34) adversaries in China and Russia and (01:25:36) Putin and I think I know you talk about (01:25:39) them being a few years behind maybe one (01:25:41) or two years behind but they're (01:25:43) eventually going to get there they're (01:25:44) eventually going to get to the point (01:25:45) where they have these large language (01:25:47) models or these AIS that can do these (01:25:48) Day Zero attacks on our (01:25:51) nation (01:25:52) and they they don't have the like sort (01:25:55) of social incentive structure if they're (01:25:57) a communist country to protect and to um (01:26:02) guard against these things are you not (01:26:03) worried about what China is gonna do um (01:26:05) I am worried and I'm worried (01:26:07) because you're going into a space of (01:26:09) great power without fully defined (01:26:12) boundaries what kinger and we talk about (01:26:14) this in the book The the Genesis Book is (01:26:17) fundamentally about what happens to (01:26:18) society with the arrival of this new (01:26:20) intelligence and the first book we did (01:26:23) age of AI was right before chat GPT so (01:26:26) now everybody kind of understands how (01:26:27) powerful these things are we talked (01:26:28) about it now you understand it so once (01:26:31) these things show up who's going to run (01:26:33) them who's going to be in charge how (01:26:34) will they be used so from my perspective (01:26:38) I believe at the moment anyway that (01:26:41) China will behave relatively responsibly (01:26:44) and the reason is that it's not in their (01:26:46) interest to have free (01:26:48) speech in every case in China when they (01:26:51) have a choice of giving freedom to their (01:26:54) Cit citizens or not they choose (01:26:55) non-freedom and I know this because I (01:26:57) spent through all the uh I spent all the (01:26:59) time dealing with it so it sure looks to (01:27:03) me like the Chinese AI solution will be (01:27:06) different from the West because of that (01:27:09) fundamental bias against freedom of (01:27:11) speech because these things are noisy (01:27:14) they make a lot of noise they'll (01:27:15) probably still make AI weapons though (01:27:18) well on the weapon side you have to (01:27:20) assume that every new technology is (01:27:23) ultimately strengthened in a war um the (01:27:26) tank was invented in World War I at the (01:27:28) same time you had the initial forms of (01:27:30) uh airplanes much of the second world (01:27:33) war was an air Campaign which (01:27:35) essentially built many many things and (01:27:38) if you look at the the there's a a book (01:27:40) called Freedom's Forge about the (01:27:43) American U structure according to the (01:27:46) book they ultimately got to the point (01:27:48) where they could build two or three (01:27:49) airplanes a day at scale so in an (01:27:53) emergency Nations have enormous (01:27:56) power I get asked all the time if (01:27:59) everyone if anyone's going to have a job (01:28:01) left to do because this is the (01:28:02) disruption of intelligence and whether (01:28:04) it's people driving cars today I mean we (01:28:06) saw the Tesla announcement of the robo (01:28:08) taxis whether it's accountants lawyers (01:28:10) and everyone in between that's or (01:28:12) podcasters are we going to have jobs (01:28:14) left well um this question has been (01:28:18) asked for 200 years um there was there (01:28:21) were the L eyeses here in Britain way (01:28:23) back when and inevitably when these (01:28:25) Technologies come along there's all (01:28:27) these fears about them indeed with a lot (01:28:29) I there were riots and people you know (01:28:31) destroying the Looms and all of this (01:28:32) kind of stuff but somehow we got through (01:28:34) it so um my own view is that there will (01:28:39) be a lot of job (01:28:41) dislocation but there will be a lot more (01:28:43) jobs not fewer jobs and here's why we (01:28:47) have a demographic problem in the world (01:28:49) especially in the developed developed (01:28:50) world where we're not having enough (01:28:52) children uh that's well understood uh (01:28:55) furthermore we have a lot of older (01:28:56) people and and the younger people have (01:28:58) to take care of the older people and (01:28:59) they have to be more productive if you (01:29:01) have young people who need to be more (01:29:02) productive the best way to make them (01:29:04) more more productive is to give them (01:29:05) more tools to make them more productive (01:29:08) whether it's a machinist that goes from (01:29:10) a manual machine into a CNC machine or (01:29:13) in in the more modern case of a (01:29:15) knowledge worker who can achieve more (01:29:17) objectives we need that productivity (01:29:19) group if you look at Asia which is the (01:29:21) centerpiece of (01:29:22) manufacturing they have all this cheap (01:29:24) labor well it's not so cheap anymore so (01:29:26) do you know what they did they added (01:29:28) robotic assembly Lin so today when you (01:29:30) go to China in particular it's also true (01:29:32) in Japan and Korea the manufacturing is (01:29:34) largely done by robots why because their (01:29:37) demographics are terrible and their cost (01:29:39) of Labor is too high so the future is (01:29:42) not fewer jobs it's actually a lot of (01:29:44) jobs that are unfilled with people who (01:29:47) may have a job skill mismatch which is (01:29:49) why education is so important now what (01:29:51) are examples of jobs that go away (01:29:54) automation (01:29:55) has always gotten rid of jobs that are (01:29:58) dangerous physically dangerous or ones (01:30:01) which are essentially too repetitive and (01:30:03) too boring for humans I'll give you an (01:30:06) example um security guards it makes (01:30:08) sense that security guards would become (01:30:10) robotic because it's hard to be a (01:30:13) security guard you fall asleep you don't (01:30:16) know quite what to and these systems can (01:30:17) be smart enough to be very very good (01:30:19) security now these are these are (01:30:21) important sources of income for these (01:30:24) people they're going to have to find (01:30:25) another job another example in in the (01:30:27) media in um Hollywood everyone's (01:30:30) concerned that AI is going to take over (01:30:31) their jobs all the evidence is the (01:30:33) inverse and here's why um the Stars (01:30:36) still get money The Producers still make (01:30:38) money they still distribute their movie (01:30:40) but their cost of making the movie is (01:30:42) lower because they use more they use for (01:30:44) example synthetic backdrops so they (01:30:46) don't have to build the set um they can (01:30:48) do synthetic makeup now there are job (01:30:50) losses there so the people who make the (01:30:52) make make the set and do the makeup are (01:30:54) going to have to go back into (01:30:55) construction and personal care by the (01:30:58) way in America and I think it's true (01:31:00) here there's an enormous shortage of (01:31:01) people who can do high quality (01:31:03) craftsmanship right those people will (01:31:05) have jobs they're just different and (01:31:07) they may not be in Los Angeles am I (01:31:09) gonna have to interface with this (01:31:11) technology am I going to have to get a (01:31:12) neuralink in my brain because we you go (01:31:15) over the subject of there being these (01:31:16) sort of two species of humans (01:31:18) potentially ones that do have a way to (01:31:22) incorporate themselves more with (01:31:23) artificial intelligence and those that (01:31:25) don't and if and if that is the case (01:31:27) what is the time Horizon in your view of (01:31:29) that (01:31:30) happening I think neuralink is much more (01:31:32) speculative because you're dealing with (01:31:34) direct brain connection and nobody's (01:31:35) going to drill on my brain until it (01:31:37) needs it trust me I suspect you feel the (01:31:39) same uh I I guess my O My overall view (01:31:43) is that (01:31:48) um you will not (01:31:50) notice how much of your world has been (01:31:53) co-opted by these Technologies because (01:31:56) they will produce greater (01:31:58) Delight if you think about it a lot of (01:32:01) life is inconvenient it's fix this call (01:32:04) this make this happen AI systems should (01:32:06) make all that seamless you should be (01:32:08) able to wake up in the morning and have (01:32:10) coffee and not have a care in the world (01:32:12) and have the computer help you have a (01:32:14) great day this true of everyone now what (01:32:17) happens to your to your profession well (01:32:20) as we said no matter how good the (01:32:23) computers are people are going to want (01:32:25) to care about other people another (01:32:26) example let's imagine you have Formula 1 (01:32:28) and you have Formula One with humans in (01:32:30) it and then you have a a a robot Formula (01:32:33) 1 which where the cars are driven by the (01:32:35) equivalent of a robot is anyone going to (01:32:37) go to the robotic Formula 1 I don't (01:32:40) think so because of the drama the human (01:32:43) achievement and so forth do you think (01:32:45) that when they run the marathon here in (01:32:46) London they're going to have robots (01:32:48) running with humans of course not right (01:32:51) of course the robots can run faster than (01:32:52) humans it's not interesting what is (01:32:54) interesting is to see human achievement (01:32:57) so I think the commentators who say oh (01:32:59) there won't be jobs we won't care I (01:33:00) think they miss the point that we care a (01:33:03) great deal about each other as human (01:33:05) beings we have opinions you have a (01:33:07) detailed opinion about me having just (01:33:09) met me met me right now and I for you we (01:33:12) just are naturally set up your face your (01:33:14) mannerisms and so forth we can describe (01:33:16) it all right the robot shows up is like (01:33:18) oh my God what another robot how boring (01:33:21) why is samman working on the the founder (01:33:23) of open AI when the co-founders of open (01:33:25) a working on universal basic income (01:33:27) projects like worldcoin then well (01:33:29) worldcoin is not the same thing as (01:33:30) universal Bitcoin uh um Universal basic (01:33:34) income there is a belief in the tech (01:33:37) industry that it goes something like (01:33:40) this the politics of abundance what we (01:33:43) do is going to create so much abundance (01:33:46) that most people won't have to work and (01:33:49) there'll be a small number of groups (01:33:50) that work who typically these people (01:33:52) themselves and there be so much Surplus (01:33:54) everyone can live like a millionaire and (01:33:56) everyone will be happy I completely (01:33:58) think this is false I think none of what (01:33:59) I just told you is false but all of (01:34:01) these Ubbi ideas come from this notion (01:34:05) that humans don't behave the way we (01:34:07) actually do so I'm I'm a Critic of this (01:34:09) view I believe that that we as humans so (01:34:13) I an example is um we're going to make (01:34:16) legal the legal profession much much (01:34:18) easier because we can automate much of (01:34:20) the technical work of lawyers does that (01:34:22) mean we're going to have fewer lawyers (01:34:23) no the current lawyers will just do more (01:34:26) laws they'll do more they'll add more (01:34:28) complexity the system doesn't get easier (01:34:30) the humans become more sophisticated in (01:34:32) their application of the principles we (01:34:34) are naturally basically uh we have this (01:34:37) thing called um basically reciprocal (01:34:40) altruism that's part of us but we also (01:34:42) have our bad sides as well those are not (01:34:44) going away because of AI when I think (01:34:46) about AI this simple analogy often think (01:34:48) of is say my IQ is Steven bartett is 100 (01:34:51) and there's this AI that sat next to me (01:34:53) whose IQ is 1,000 what on Earth would (01:34:56) you want to give Steven to do because (01:34:58) because that 1,000 IQ would have really (01:35:00) bad judgment in a couple cases because (01:35:02) remember that the AI systems do not have (01:35:04) human values unless it's added right I (01:35:08) would much rather talk to you about (01:35:10) something involving a moral or human (01:35:12) judgment even with the Thousand I (01:35:14) wouldn't mind Consulting it so tell me (01:35:16) the the history how was this resolved in (01:35:18) the past how are these but at the end of (01:35:20) the day in my view the core aspects of (01:35:24) it which have to do with morals and (01:35:26) judgment and beliefs and Charisma (01:35:28) they're not going away is there a chance (01:35:30) that this is the end of humanity no um (01:35:33) the way Humanity (01:35:34) does is much it's much harder to (01:35:37) eliminate all of humanity than you think (01:35:39) all the people I've looked with on these (01:35:41) biological attacks say it's it takes (01:35:43) more than one horrific pandemic and so (01:35:46) forth to eliminate humanity and and the (01:35:48) the pain can be very very high in these (01:35:50) moments look at the World War I World (01:35:53) War II the Hodor in uh Ukraine in the (01:35:56) 1930s the Nazis you know these are (01:35:59) horrifically painful things but we (01:36:01) survived right we we as a as a Humanity (01:36:04) survived and we will I wonder if this is (01:36:07) the moment where humans couldn't see (01:36:09) past around the corner because you know (01:36:12) I've heard you talk about how the AIS (01:36:13) will turn in they'll be agents and (01:36:15) they'll be able to speak to each other (01:36:16) and we won't be able to understand the (01:36:17) language I have a specific proposal on (01:36:19) that um there are points where humans (01:36:22) should assert control (01:36:24) and I've been trying to think about (01:36:26) where are they I'll give you an example (01:36:28) there's something called recursive (01:36:29) self-improvement where the system just (01:36:31) keeps getting smarter and smarter and (01:36:32) learning more and more things at some (01:36:35) point if you don't know what it's (01:36:37) learning you should unplug it but we (01:36:40) can't unplug them can we sure you can (01:36:42) there's a power plug and there's a (01:36:43) circuit breaker go and turn the circuit (01:36:45) breaker off another example um there's a (01:36:49) there's a scenario theoretical where the (01:36:51) system is so powerful it can produce a (01:36:54) new model faster than the previous model (01:36:56) was checked okay that's another (01:36:59) intervention point so in each of these (01:37:03) cases um if the if agents and the (01:37:06) technical term is called agents what (01:37:07) they really are is large language models (01:37:09) with memory and you can begin to (01:37:11) concatenate them you can say this model (01:37:13) does this and then it feeds into this (01:37:15) and so forth you can build very powerful (01:37:17) decision systems we believe this is the (01:37:19) the the thing that's occurring this year (01:37:21) and next year everyone's doing them they (01:37:23) will arrive (01:37:25) the agents today speak in English you (01:37:27) can see what they're saying to each (01:37:29) other they're not human but they are (01:37:31) communicating what they're doing English (01:37:34) to English to English as long as and it (01:37:37) doesn't have to be English but as long (01:37:38) as they're human understandable but (01:37:40) let's so the thought experiment is one (01:37:42) of the agents says I have a better idea (01:37:44) I'm going to communicate in my own (01:37:45) language that I'm going to invent that (01:37:47) only other agents understand that's a (01:37:49) good time to pull the plug what is your (01:37:52) biggest fear about AI (01:37:54) my actual fear is different from what (01:37:56) you might imagine my my actual fear is (01:37:58) we're not going to adopt it fast enough (01:37:59) to solve the problems that affect (01:38:01) everybody right and the reason is that (01:38:04) the that if you look at every everyone's (01:38:07) everyday lives what do they want they (01:38:09) want safety they want Health Care they (01:38:11) want great schools for their kids we (01:38:13) just work on that for a while why do we (01:38:15) make people's lives just better because (01:38:18) of AI we have all these other (01:38:19) interesting things why don't we have a (01:38:22) um a teacher that is an AI teacher that (01:38:25) works with existing teachers in this (01:38:28) language of the kid in the culture of (01:38:30) the kid to get the kid as smart as they (01:38:32) possibly can why don't we have a doctor (01:38:34) or doctor's assistant really that (01:38:36) enables a a human doctor to always know (01:38:39) every possible best treatment and then (01:38:41) based on their current situation what (01:38:43) the inventory is which country is how (01:38:45) their insurance Works what is the best (01:38:46) way to treat that patient those are (01:38:48) relatively achievable Solutions why (01:38:50) don't we have them if you just did (01:38:52) education and Healthcare (01:38:54) globally the impact in terms of lifting (01:38:57) human potential up would be so great (01:39:00) right that it would change (01:39:02) everything it wouldn't solve the various (01:39:04) other things that we complain about (01:39:05) about you know this celebrity or this (01:39:07) misbehavior or this conflict or even (01:39:09) this war but it would establish a Level (01:39:12) Playing Field of knowledge and (01:39:13) opportunity at a global level that has (01:39:15) been the dream for decades and decades (01:39:18) and decades Chuck me that perfect head (01:39:22) one of the things that I think about the (01:39:24) time because my life is quite hectic and (01:39:25) busy is how to manage my energy load and (01:39:28) as a podcaster you kind of have to (01:39:29) manage your energy in such a way that (01:39:31) you can have these articulate (01:39:32) conversations with experts on subjects (01:39:34) you don't understand and this is why (01:39:36) perfect Ted has become so important in (01:39:37) my life because previously when it came (01:39:40) to Energy Products I had to make a (01:39:41) trade-off that I wasn't happy with (01:39:43) typically if I wanted the energy I had (01:39:44) to deal with high sugar I had to deal (01:39:47) with Jitters and crashes that come along (01:39:49) with a lot of the mainstream Energy (01:39:50) Products and I also just had to tolerate (01:39:52) the fact that if I want energy I have to (01:39:54) put up with a lot of artificial (01:39:55) ingredients which my body didn't like (01:39:58) and that's why I invested in perfect Ted (01:39:59) and why they're one of the sponsors of (01:40:01) this podcast it is changed not just my (01:40:02) life but my entire team's life and for (01:40:04) me it's drastically improved my (01:40:05) cognitive performance but also my (01:40:07) physical performance so if you haven't (01:40:09) tried perfect Ted yet you must have been (01:40:10) living under a rock now is the time you (01:40:13) can find perfect Ted at Tesco and (01:40:15) waitrose or online where you can enjoy (01:40:17) 40% off with code diary 40 at checkout (01:40:20) head to perfect ted.com this is quite (01:40:23) interesting 85% of Internet users have (01:40:26) heard of vpns but only 55% know what (01:40:29) they do if you're in that group let me (01:40:31) explain vpn's enable your location (01:40:34) online to differ from where you actually (01:40:35) are geographically to help you browse (01:40:37) and stream sites that would otherwise be (01:40:39) unavailable to you I use nordvpn who are (01:40:41) a sponsor of this show to watch (01:40:43) Manchester United games online no matter (01:40:44) where I am in the world and Indie from (01:40:46) my team uses them whenever she's booking (01:40:48) flights back home to New Zealand having (01:40:49) a different online location means she (01:40:52) can take advantage of dynamic pricing (01:40:54) and get cheaper prices for her flights (01:40:56) nordvpn is the fastest VPN in the world (01:40:59) and just one account can be used across (01:41:00) 10 devices and they've shared a generous (01:41:03) offer for my listeners a discount and (01:41:05) four additional months free on a 2-year (01:41:07) plan it's also completely risk-free with (01:41:09) nord's 30-day money back guarantee so (01:41:12) head to nordvpn.com (01:41:14) doac or click the link in the (01:41:16) description (01:41:17) below throughout the pandemic I've been (01:41:20) a big supporter um it was a contrarian (01:41:22) view but I think it's now less a (01:41:23) contrarian view that companies and CEOs (01:41:27) need to be clear in their convictions (01:41:28) around how they work and one of the (01:41:30) things that I've um been criticized a (01:41:32) lot for is that I'm I'm for having (01:41:34) people in a room together so my (01:41:36) companies we um we're not remote we work (01:41:38) together in an office as I said down the (01:41:40) road from here and I believe in that (01:41:42) because I think of community and (01:41:43) engagement and synchronous work and I (01:41:45) think that work now has a responsibility (01:41:47) to be more than just a set of tasks you (01:41:49) do in a world where we're lonier than (01:41:51) ever before there's more disconnection (01:41:52) and especially for young people you (01:41:54) don't have families and so on um having (01:41:56) them work alone in a small white box in (01:41:58) a big city like London or New York um is (01:42:00) robbing them of something which I think (01:42:01) is important this was a bad this was a (01:42:04) contrarian view it's become less (01:42:05) contrarian as the big tech companies in (01:42:07) America have started to roll back some (01:42:09) of their initial knej reactions to the (01:42:10) pandemic that there a lot of them are (01:42:12) asking their team members to come back (01:42:13) into the office at least a couple of (01:42:15) days a week what's your point of view on (01:42:17) this so I have a strong view that I want (01:42:19) people in an office it doesn't have to (01:42:21) be all one office but I want them in an (01:42:23) office (01:42:24) and partly it's for their own benefit if (01:42:26) you're in your 20s when I was a young (01:42:28) executive I knew nothing of what I was (01:42:29) doing I literally was just lucky to be (01:42:31) there and I learned by hanging out at (01:42:34) the water cooler going to meetings (01:42:35) hanging out being in the hallway had I (01:42:37) been at home I wouldn't have had any of (01:42:39) that knowledge which ultimately was (01:42:41) Central to my subsequent promotions so (01:42:43) if you're in your 20s you want to be in (01:42:45) an office because that's how you're (01:42:46) going to get promoted and I think that's (01:42:48) consistent with the majority of the (01:42:50) people who really want to work from home (01:42:52) have honest problems with commuting and (01:42:54) family and so forth they're real issues (01:42:56) the problem with our joint view is it's (01:42:58) not supported by the data the data (01:43:00) indicates that productivity is actually (01:43:02) slightly higher in uh work uh when you (01:43:06) allow work from home so you and I really (01:43:09) want that company of people sitting (01:43:11) around the table and so forth but the (01:43:13) evidence does not support our view (01:43:14) interesting yeah is that true it is (01:43:17) absolutely true why is Facebook and all (01:43:18) these companies rolling back their uh (01:43:20) and like Snapchat rolling back their (01:43:22) remote working policies then not (01:43:23) everyone is um and you most companies (01:43:27) are doing various forms of hybrids where (01:43:29) it's two days or three days or so forth (01:43:33) um I'm sure that for the average (01:43:34) listener here who works in public (01:43:36) security or in a government they say (01:43:37) well my God they're not in the office (01:43:39) every every every day but I'll tell you (01:43:42) that at least for the the industries (01:43:44) that have been studied there's evidence (01:43:46) that allowing that flexibility from work (01:43:48) from home increases productivity I don't (01:43:50) happen to like it but I want to (01:43:52) acknowledge the science is there what is (01:43:55) the um the advice that you wish you'd (01:43:57) gotten at my age that you didn't get the (01:44:00) most important thing is probably keep (01:44:01) betting on yourself and bet again and (01:44:03) roll the dice and roll the dice what (01:44:06) happens in as you get older is you (01:44:08) realize that these opportunities were in (01:44:10) front of you and you didn't jump for (01:44:12) them why you were in a bad mood or you (01:44:15) know you didn't know who to call or so (01:44:17) forth life can be understood as a series (01:44:20) of opportunities that are put before you (01:44:22) and they're Tim Limited (01:44:24) I was fortunate that I got the call (01:44:26) after a number of people had turned it (01:44:27) down to work for Larry and for and with (01:44:29) Larry Sergey at Google changed my life (01:44:32) right but that was luck and timing my (01:44:35) one friend on the board at the moment (01:44:37) said I was very thankful to him and he (01:44:39) said but you know you did one thing (01:44:41) right I said what he said you said (01:44:44) yes so your philosophy in life should be (01:44:47) to say yes to that opportunity and yes (01:44:49) it's painful and yes it's difficult and (01:44:51) yes you have to deal with your family (01:44:52) and yes you have to travel to to some (01:44:54) foreign place and so forth get on the (01:44:55) airplane and get it (01:44:57) done what's the hardest challenge you've (01:44:59) dealt with in your life well on the (01:45:01) personal side you know I've had the I've (01:45:03) had a set of you know personal personal (01:45:05) Pro problems and (01:45:06) tragedies um like everyone does I think (01:45:10) on a business (01:45:11) context (01:45:15) um there were moments at Google where we (01:45:19) had control over an industry that we (01:45:22) didn't execute well the most obvious one (01:45:23) is social (01:45:25) media uh at the time when Facebook was (01:45:27) founded we had a system which we called (01:45:29) Orit um which was really really (01:45:31) interesting and somehow we we we did (01:45:33) everything well but we missed that one (01:45:36) right and I would have preferred and (01:45:37) I'll take responsibility for that we (01:45:39) have a closing tradition on this podcast (01:45:41) where the last guest leaves a question (01:45:42) for the next guest not knowing who (01:45:43) they're going to be leaving it for and (01:45:44) the question left for you is what is (01:45:46) your non-negotiable something you do (01:45:48) that significantly improves everyday (01:45:51) life well what I try to do is try to be (01:45:53) online and I also try to keep people (01:45:56) honest every day you keep you hear all (01:45:59) sorts of ideas and and so forth half of (01:46:02) which are right half of which are wrong (01:46:04) I try to make sure I know the truth as (01:46:06) best we can determine it Eric thank you (01:46:09) so much thank you it's uh such an honor (01:46:11) your books are have shaped my thinking (01:46:13) in so many so many important ways and I (01:46:15) think your new book Genesis is the (01:46:17) single best book I've I've read on the (01:46:19) subject of AI because you take a very (01:46:21) nuanced approach to these subject (01:46:23) matters and I think sometimes it's (01:46:25) tempting to be binary in your way of (01:46:27) thinking about this technology the the (01:46:28) pros and the cons but your writing your (01:46:30) videos your work takes this really (01:46:32) balanced but informed approach to it I (01:46:34) have to say as an entrepreneur the (01:46:36) trillion dollar coach book is what I (01:46:37) highly recommend everybody goes and (01:46:39) reads because it's um it's just a really (01:46:41) great Manual of being a leader in the (01:46:43) Modern Age and an entrepreneur I'm going (01:46:44) to link all five of these books in the (01:46:46) in the comment section below the new (01:46:48) book Genesis comes out in the US I (01:46:50) believe on the 19th of November (01:46:53) um I don't have the UK date but I'll (01:46:55) find it and I'll put it in but it's a (01:46:57) book it's a it's a critically important (01:47:01) book that nobody should avoid I've been (01:47:03) searching for answers that are contained (01:47:04) in this book for a very very long time (01:47:06) I've been having very a lot of (01:47:07) conversations on this podcast in search (01:47:08) of some of these answers and I feel (01:47:10) clearer um about myself my future but (01:47:12) also the future of society because I've (01:47:14) read this book so thank you for writing (01:47:15) it and thank you and let's thank Dr (01:47:17) Kissinger he finished the last chapter (01:47:20) in his last week of life in his deathbed (01:47:23) that's how profound he thought that this (01:47:25) book was And all I'll tell you is that (01:47:28) he wanted to set us up for a good next (01:47:31) 50 years having lived for so long and (01:47:34) seen both good and evil he wanted to (01:47:37) make sure we continue the good progress (01:47:39) we're making as a (01:47:40) society is there anything he would want (01:47:42) to say any answer he gave would take (01:47:45) five (01:47:46) [Music] (01:47:48) minutes a remarkable man thank you Eric (01:47:52) thank you (01:47:55) [Music] (01:47:57) I'm going to let you into a little bit (01:47:58) of a secret and you're probably going to (01:48:00) think that I'm a little bit weird for (01:48:01) saying this but our team are our team (01:48:04) because we absolutely obsess about the (01:48:07) smallest things even with this podcast (01:48:09) when we're recording this podcast we (01:48:10) measure the CO2 levels in the studio (01:48:12) because if it gets above a th000 parts (01:48:13) per million cognitive performance dips (01:48:16) this is the type of 1% Improvement we (01:48:18) make on our show and that is why the (01:48:19) show is the Way It Is by understanding (01:48:22) the power of pounding 1% you can (01:48:24) absolutely change your outcomes in your (01:48:26) life it isn't about drastic (01:48:28) Transformations or quick wins it's about (01:48:31) the small consistent actions that have a (01:48:34) lasting change in your outcomes so two (01:48:36) years ago we started the process of (01:48:38) creating this beautiful diary and it's (01:48:40) truly beautiful inside there's lots of (01:48:41) pictures lots of inspiration and (01:48:43) motivation as well some Interac Dev (01:48:45) elements and the purpose of this diary (01:48:47) is to help you identify stay focused on (01:48:51) develop consistency with the one % that (01:48:53) will ultimately change your life we have (01:48:55) a limited number of these 1% Diaries and (01:48:57) if you want to do this with me then join (01:48:59) our waiting list I can't guarantee all (01:49:01) of you that join the waiting list will (01:49:02) be able to get one but if you join now (01:49:03) you have a higher chance the waiting (01:49:05) list can be found atth diary.com I'll (01:49:08) link it below but that isth diary.com (01:49:13) [Music] (01:49:23) ah