#217 Josh Wolfe: Human Advantage in the World of AI

0:00:02 Look at AI right now.
0:00:03 Billions of dollars being spent.
0:00:07 The flurry of all of this benefits us as consumers,
0:00:08 always and everywhere.
0:00:10 Just wait and they’ll compete and compete
0:00:12 and it will accrue to us as users.
0:00:15 How do we create an unfair advantage in a world of AI?
0:00:18 I think the limits to human intelligence
0:00:20 are rooted in our biology.
0:00:23 AI’s over time will understand us in many ways better
0:00:25 than we understand ourselves.
0:00:27 We can still be computers and chess, done.
0:00:29 We can still be them and go, done.
0:00:30 We can be them and video games, done.
0:00:32 Okay, but we still have creativity, done.
0:00:34 All these things have been trained
0:00:38 on the sum total of all human creation.
0:00:40 And now they’re being trained on the sum total
0:00:42 of human creation plus artificial creation.
0:00:46 I’m absolutely convinced that we are going to have machines
0:00:48 doing science 24/7.
0:00:50 I just think it’s gonna be part of the total overture
0:00:55 over of creation and I think it’s a beautiful thing.
0:00:57 (upbeat music)
0:01:00 (upbeat music)
0:01:11 Welcome to The Knowledge Project.
0:01:13 I’m your host, Shane Parish.
0:01:15 In a world where knowledge is power,
0:01:18 this podcast is your toolkit for mastering the best
0:01:20 what other people have already figured out.
0:01:22 If you want to take your learning to the next level,
0:01:24 consider joining our membership program
0:01:27 at fs.log/membership.
0:01:29 As a member, you’ll get my personal reflections
0:01:32 at the end of every episode.
0:01:35 Early access to episodes, no ads, including this,
0:01:38 exclusive content, hand edited transcripts,
0:01:39 and so much more.
0:01:42 Check out the link in the show notes for more.
0:01:44 While others ask what’s trending,
0:01:48 Josh Wolf asks, what seems impossible today?
0:01:51 He’s built a career betting on scientific breakthroughs
0:01:53 that most people don’t believe can happen.
0:01:55 As co-founder of Lux Capital,
0:01:57 he’s backed companies cleaning up nuclear waste
0:02:00 and building brain computer interfaces.
0:02:02 But here’s the contradiction.
0:02:03 Despite investing in technology
0:02:05 that could make humans obsolete,
0:02:09 Josh is profoundly optimistic about human potential.
0:02:12 His thinking challenges conventional wisdom.
0:02:15 While most see AI as automation and threats to humanity,
0:02:19 Josh sees it as a catalyst for human achievement.
0:02:22 In this conversation, we explore this paradox,
0:02:25 diving deep into how technological evolution
0:02:28 can amplify rather than diminish what makes us human.
0:02:30 From geopolitical power shifts
0:02:32 to the future of human creativity,
0:02:35 Josh reveals the exact frameworks he uses
0:02:36 for seeing what others miss
0:02:39 and betting on the seemingly impossible future.
0:02:42 It’s time to listen and learn.
0:02:48 – Start with what you’re obsessed with today.
0:02:49 What’s on your mind?
0:02:52 – Well, first and foremost, kids and family.
0:02:54 Trying to be a good dad, good husband.
0:02:57 Technologically obsessed with so many different things.
0:02:59 We were just in a partnership meeting.
0:03:01 And probably the most interesting thing at the moment
0:03:05 is thinking about the speed of certain technologies,
0:03:06 like the actual physical technologies
0:03:08 and where the bottlenecks are.
0:03:12 So in biology, you’ve got all kinds of reactions,
0:03:14 nature figured out, evolution figured out,
0:03:15 enzymes and catalysts and things
0:03:16 that can speed up reactions.
0:03:20 But you can move faster than the speed of biology.
0:03:22 Now you think about AI, total different field,
0:03:26 but the same sort of underlying philosophical principle.
0:03:31 If you’ve tried ChatGPT operator,
0:03:35 it can only move at the speed of the web.
0:03:37 And at that it’s sort of with latency a little bit slow.
0:03:40 So we’re thinking about what are the technologies
0:03:41 that can accelerate these things
0:03:43 that have these natural almost physics limits.
0:03:46 And even if those limits are biological or digital.
0:03:48 So that is something that at the moment I’m obsessed with
0:03:50 in part because I have ignorance about it.
0:03:51 And when I have ignorance about it,
0:03:53 my competitive spirit says,
0:03:54 how do I get smart about this?
0:03:56 How do I get some differentiated insight?
0:03:57 How do I know what everybody else thinks
0:04:00 and then find the sort of white space answer?
0:04:03 So that’s one big thing, obsessed with geopolitics.
0:04:08 It is the great game, everything from US-China competition
0:04:14 to Iran, Israel, access of religious conflict,
0:04:17 the Sahel and Maghreb in Africa,
0:04:19 which is not an area that a lot of people
0:04:21 talk about or think about that I believe
0:04:26 with low probability, super high magnitude import
0:04:28 is going to become the next Afghanistan,
0:04:30 that that region in the Sahel,
0:04:34 all these coups and cascades of these coups of failed states
0:04:36 where you have Russian mercenaries,
0:04:39 violent extremists, Chinese CCP coming in
0:04:42 with infrastructure and influence,
0:04:45 European colonial powers being kicked out
0:04:48 against this backdrop of brilliant scientists
0:04:51 and engineers and technologists that went to HBS or Stanford
0:04:53 and worked at Metta and Google and are going back
0:04:57 and particularly in like Ghana and Kenya and Nigeria
0:04:58 or building businesses.
0:05:03 So that continent is going to be a truly contested space
0:05:07 for future in progress and for utter chaos and terrorism.
0:05:09 So yeah, it’s a widespread of stuff
0:05:11 that I’m probably currently obsessed with.
0:05:12 – Let’s talk about all those things.
0:05:14 Let’s start with sort of diving in.
0:05:16 You mentioned chat, GBT operator.
0:05:16 – Yes.
0:05:18 – And the limitations sort of being,
0:05:19 we’re moving at the speed of the web.
0:05:21 At what point do you think that we’re,
0:05:24 those systems are all designed for humans?
0:05:26 At what point do they become designed for AI first
0:05:27 and then humans are using them
0:05:30 or do we just have two simultaneous interfaces?
0:05:31 – I think it’s going to be both.
0:05:34 I think that there’s always this like ignorance arbitrage
0:05:37 where somebody figures out that there’s an opportunity
0:05:39 to take advantage and improve a system
0:05:41 while people don’t understand it.
0:05:42 And so I think that there are people
0:05:46 that are probably going to launch redesign businesses
0:05:48 where they say we will optimize your web pages
0:05:50 just like people did for search engine optimization
0:05:53 when Google rose that you had to be more relevant
0:05:54 because Google was so important.
0:05:56 Google was influencing whether or not people would see you
0:05:57 or discover you.
0:06:00 And so if there are certain tasks like open table
0:06:02 or restaurants or shopping,
0:06:05 they are now going to start to shift their user interface
0:06:08 not just for the clicks of a human,
0:06:09 but for the clicks of an AI
0:06:11 that’s doing research on the best product.
0:06:14 And the way that those are going to negotiate
0:06:17 and in some cases influence or trick the judgment
0:06:20 and the reasoning of the AI is really interesting.
0:06:23 So I think that that is probably the next domain
0:06:25 where those things are going to get better, faster,
0:06:27 they’re going to have more compute,
0:06:29 but then people are going to be redesigning
0:06:31 some of these experiences, not for us
0:06:33 and the user interface of humans, but machine to machine.
0:06:35 And you’re already starting to see this
0:06:36 where there was one element of like
0:06:39 in instantiation of R1 communicating with another
0:06:41 in R1, but the language that they were commuting
0:06:43 was not English or Chinese or even traditional code.
0:06:45 It was like this weird almost alien language.
0:06:48 And so I think you can see a bunch of that.
0:06:49 There’s another adjacent theme
0:06:51 which I think is also really interesting,
0:06:54 which is AI portends the death of APIs.
0:06:58 So APIs allow Meta with their Metaglasses or Orion
0:07:03 to be able to communicate with Uber and Spotify
0:07:04 through the backend, through software.
0:07:07 But increasingly those things are complicated,
0:07:09 they’re hard to negotiate, there’s a lot of legal,
0:07:11 there’s API calls, there’s restrictions,
0:07:13 there’s privacy, there’s controls.
0:07:14 But if you’re one of these companies that are like,
0:07:16 I don’t want to go through all of that.
0:07:19 Can I just use AI to pretend that I’m a user?
0:07:20 And in fact, I had this experience
0:07:22 where I was using Operator
0:07:25 and I had this moral dilemma for a split second.
0:07:28 Do I click to confirm that I’m not a bot
0:07:29 because I had to take control over it
0:07:30 because the bot actually was trying
0:07:32 to do my research on behalf of me.
0:07:36 And so you see a world where APIs
0:07:39 that have been the plumbing of everything in SAS
0:07:41 and software and negotiating behind the scenes
0:07:44 may start to lose influence and power to AIs
0:07:47 that are able to basically just negotiate on the front end
0:07:48 as though they were a user.
0:07:50 So I think that that whole domain
0:07:54 is going to very rapidly evolve in like a quarter or two.
0:07:57 – You mentioned sort of the limitations
0:07:59 like biology moves at a certain speed.
0:08:01 There’s a couple of subset of questions here,
0:08:04 but one is where’s the limitation in AI growth right now?
0:08:07 It seems like we have energy as a key input.
0:08:10 We have compute as a key input.
0:08:13 And then we have data slash algorithms
0:08:15 as like the next key input.
0:08:16 What am I missing?
0:08:18 What’s the limitation on those?
0:08:20 – Well, start with conventional wisdom
0:08:21 which has heretofore been correct,
0:08:23 which is you need more compute,
0:08:26 you need more capital, scaling laws for AI,
0:08:29 just throw more GPUs and processors and money
0:08:31 and ultimately energy to support that
0:08:33 and you will get better and better breakthroughs.
0:08:35 The counter to that, which you’re seeing in some cases
0:08:37 with open source or people that have now trained
0:08:38 on some of the large models
0:08:41 is that there’s gonna be a shift towards model efficiency.
0:08:43 And so that’s number one,
0:08:44 that people are gonna figure out
0:08:46 how do we do these more efficiently with less compute.
0:08:49 Number two, which is a big sort of contrary thesis
0:08:53 that I have is that a significant portion of inference.
0:08:55 So, you know, if you break down training,
0:08:57 you still need large 100,000 ish clusters
0:08:59 of each 100 top performing chips.
0:09:00 It’s expensive.
0:09:03 Only the hyperscalers, heretofore have been able to do that.
0:09:04 You can do some training
0:09:06 if you’re using things like together compute
0:09:08 in some of our companies without having to do that yourself.
0:09:12 Sort of like going back to on-prem versus colo
0:09:15 versus cloud transition 15 years ago.
0:09:17 But I think that you’re gonna end up doing
0:09:20 a lot of inference on device.
0:09:22 Meaning instead of going to the cloud
0:09:25 and typing a query, like 30 to 50% of your inference
0:09:27 may be on an Apple or an Android device.
0:09:29 And if I had to bet today,
0:09:32 it’s Android because of the architecture over Apple.
0:09:33 But if Apple can do some smart things,
0:09:35 maybe they can catch up to this,
0:09:37 but some of the design choices and the closed aspects
0:09:40 which have been great for privacy as a feature
0:09:42 may hurt them in this wave.
0:09:44 And you could already see like perplexity
0:09:46 can actually be an assistant on my Android device.
0:09:49 I carry both so I can understand both operating systems.
0:09:51 You can’t do that yet on iOS and Apple.
0:09:52 But here’s the insight.
0:09:56 If 30 to 50% of my inference is my cached emails
0:09:59 and my chats and my WhatsApp
0:10:01 and all the stuff that I keep on my device,
0:10:04 my photos, my health data, my calendar,
0:10:08 then the memory players may play a much bigger role.
0:10:10 So now you have Samsung, SK Hynex, Micron
0:10:12 that are important here.
0:10:15 If I had to come up with somewhat pejorative analogy,
0:10:18 I think Samsung is going to be more like Intel.
0:10:20 I think that they’re just a little bit sclerotic
0:10:22 and bureaucratic and slow moving
0:10:24 and it’s gonna sort of ebb and decline.
0:10:27 I think Micron is a US company
0:10:30 which is going to be more constrained
0:10:33 by restrictions that are put on export control.
0:10:35 And I think SK being a Korean company
0:10:37 is gonna be able to skirt those
0:10:40 in the same way that Nvidia has with distribution
0:10:42 to China through Singapore and Indonesia and Malaysia,
0:10:43 which is now being investigated.
0:10:47 But the memory players are likely to be ascendant.
0:10:49 And so what Here2For was a bottleneck on compute
0:10:52 could shift attention, talent, money
0:10:55 into new architectures where memory plays a key role
0:10:58 on small models on device for inference.
0:11:00 – I’ve always thought of memory as like a commodity.
0:11:02 It doesn’t matter if I have a M-U chip
0:11:03 or a scan desk of Samsung.
0:11:05 – And that’s what people thought about CPUs back in the day
0:11:08 and then GPUs for just traditional video graphics.
0:11:11 And then the AI researchers came and said,
0:11:12 “Well, wait a second.
0:11:14 We can run these convolutional neural nets on these GPUs.”
0:11:17 And they reached into somebody else’s domain
0:11:19 of PlayStation and Xbox and pulled them in
0:11:22 and suddenly it lit up this phenomenon
0:11:25 that turned Nvidia from 15 billion
0:11:27 to two and a half or three trillion.
0:11:31 Do you think Nvidia has got a long runway?
0:11:34 – I think that their ability with that market capitalization
0:11:36 and that cash and the margin that they have,
0:11:38 they can reinvent and buy a lot of things.
0:11:41 So I think Jensen is a thoughtful capital allocator.
0:11:44 I think he’s benefited and caught lightning in a bottle
0:11:45 over the past 10 years.
0:11:47 And now just particularly the past six,
0:11:49 I would not count Nvidia out.
0:11:52 Now, what’s the upside from two and a half or three trillion
0:11:53 to five trillion or 10 trillion?
0:11:56 Or do they go back down to, I have no idea.
0:11:58 So that’s more of a fundamental valuation
0:11:59 based on the speculation.
0:12:01 But if you have 90, 95% margins
0:12:04 and your chips are $30,000,
0:12:07 could you shrink them down and sell for $3,000
0:12:09 and take small margins but get more volume?
0:12:12 And this was some of the debate that I think happened
0:12:14 with the release of DeepSeq where you had Satya
0:12:16 basically talking about Jeven’s paradox
0:12:19 that any one thing might get more efficient
0:12:21 but the result is not that you have less demand.
0:12:24 In aggregate, you have much more demand.
0:12:26 The classic example of this is like refrigerators.
0:12:29 Single refrigerator back in the day was an energy hog.
0:12:32 All of a sudden you make these things more efficient
0:12:33 and what happens, it becomes much cheaper.
0:12:37 So if something’s cheaper, you’re gonna buy more of it
0:12:39 and they shrunk refrigerators down.
0:12:41 And so now everybody had one in their garage
0:12:44 and in their office and in their basement.
0:12:46 And so the aggregate demand for electricity
0:12:49 and then for all the components and coils and refrigerant
0:12:50 went up, not down.
0:12:52 Same thing with bandwidth.
0:12:56 If you have a 56K Bod modem, which was like my first modem,
0:12:58 you know, dial up internet and all that kind of stuff
0:13:01 on CompuServe and, you know, then you go to a T3
0:13:03 and fiber optic, you know, at the speed of light,
0:13:06 it is way more efficient but your usage now is just huge.
0:13:09 You’re streaming 4K videos and watching Netflix
0:13:11 and trading on Bloomberg.
0:13:14 Whereas if you actually want to decrease use,
0:13:17 the non-obvious thing that you would wanna do
0:13:18 is actually slow down speed.
0:13:20 I mean, put me on a 56K Bod modem today
0:13:22 and just like pull out my hair and never use it.
0:13:23 You couldn’t even use Gmail on that.
0:13:24 – Right, exactly. – Exactly.
0:13:26 Where do you think intelligence
0:13:28 is the limiting factor of progress?
0:13:30 – Human intelligence?
0:13:34 I think that the great thing about us is that we are,
0:13:36 I don’t know, 60 or 70% predictable
0:13:41 in that so many of the foibles and virtues and vices
0:13:45 of man and woman are Shakespearean.
0:13:46 You know, there are hundreds of, I mean,
0:13:48 tens of thousands of years old from modern evolution,
0:13:52 but with that we are still irrational.
0:13:53 You know, Danny Kahneman was a friend.
0:13:56 Danny passed last year, just an amazing guy,
0:13:58 but he could document all the heuristics
0:14:01 and all the biases which you study and write about.
0:14:04 And he’s like, I’m still a victim of them.
0:14:06 Even knowing them doesn’t really insulate me
0:14:08 from falling to them.
0:14:09 It’s just like an optical illusion.
0:14:11 You can know it’s an optical illusion,
0:14:13 but you still see it and you still fall for it.
0:14:17 So I think the limits to human intelligence
0:14:19 are rooted in our biology.
0:14:22 And we have all of these embodied intelligence
0:14:26 that have sort of been externalized in calculators
0:14:29 and in computers that help us to overcome that.
0:14:32 And I certainly feel today that a significant portion
0:14:35 of my day that might be spent with a colleague
0:14:36 riffing on something.
0:14:39 And sometimes that’s like great for a muse
0:14:41 or tapping into information or intelligence
0:14:43 or some tidbit or a piece of gossip
0:14:45 or an experience that they had.
0:14:47 And that’s why like the diversity of cognitive diversity
0:14:49 is really important.
0:14:51 I’m complimenting that with all day chats
0:14:54 with perplexity and Claude and open AI
0:14:58 and any number of LLMs that might hallucinate
0:15:01 just like a friend and might be wrong about something,
0:15:03 but might give me some profound insight.
0:15:05 – I think what I’m interested in is
0:15:08 at what point will machines like if intelligence
0:15:10 is the limiting factor on progress
0:15:12 in certain domains or areas.
0:15:15 It strikes me that in the near future,
0:15:19 the machines will be able to surpass human intelligence.
0:15:20 – For sure.
0:15:21 – And if that’s the case,
0:15:25 then those areas are rife for either disruption
0:15:26 or rapid progress.
0:15:30 – Well, I think, you know what Peter Drucker never imagined
0:15:32 when he was talking about knowledge workers
0:15:33 in the shift from like blue collar workers
0:15:35 to white collar workers was that machines
0:15:38 would actually most threaten those professions.
0:15:42 And so you take some of the most elite professions, doctors.
0:15:46 The ability to do a differential diagnosis today.
0:15:47 I take my medical records as soon as I get them
0:15:50 from top doctors and I still put them into LLMs
0:15:52 to see what did they miss?
0:15:55 And sometimes it unearths some interesting correlations.
0:15:58 Is there a scientific paper from the past 10 years
0:16:00 that might have bearing on this particular finding
0:16:01 or something in a blood test?
0:16:03 So that is really interesting.
0:16:07 Lawyers, languages, code, multi-billion dollar lawsuits
0:16:09 sometimes come down to the single placement
0:16:11 and interpretation by a human of a word.
0:16:13 One of the things that Danny Kahneman recognized
0:16:16 and he published this in his last book Noise
0:16:19 was that the same case, the same information,
0:16:21 the same facts presented to the same judge
0:16:23 at different times of day
0:16:25 or to presented to different judges.
0:16:26 They are not objective.
0:16:29 And he actually thought that for justice and fairness
0:16:31 that you would want the human intelligence
0:16:34 applied to these situations with their biases
0:16:36 to be either complimented or replaced by AIs
0:16:38 that had a consistency and a fidelity
0:16:40 in how they made decisions.
0:16:42 So those are all high paying jobs
0:16:44 with lots of training, lots of time
0:16:47 to gain the experience, the reasoning, the intelligence.
0:16:50 And many of those things are at risk.
0:16:54 Government itself, the ability to legislate, make decisions.
0:16:55 You wanna be able to capture
0:16:56 and express the will of the people.
0:16:59 But increasingly we have social media
0:17:00 that’s able to do that, it could be corrupted
0:17:02 but there’s mechanisms to figure out
0:17:05 how do you really surface what does the populace care about?
0:17:07 And some in Europe have tried to do these things.
0:17:11 The key thing is always like what’s the incentive
0:17:13 and what’s the vector where somebody can come in
0:17:15 and corrupt these things?
0:17:18 But interestingly, I actually think that the people
0:17:21 whose jobs are like most protected in this new domain
0:17:23 are blue collar workers.
0:17:26 A robot today can’t really fully serve a meal
0:17:28 and they cannot effectively even though
0:17:32 every humanoid robot tries to do folding laundry
0:17:36 and there’s still basic jobs that are not low paying
0:17:38 but they’re arguably safer than ever.
0:17:40 And this ties into immigration and technology
0:17:44 and human replacement, but people that are doing maintenance,
0:17:46 people that are plumbers, many of these things
0:17:49 are standardized systems, but it’s like the old joke
0:17:52 about the plumber that comes in and he comes in
0:17:55 and he taps a few things and suddenly the pipes are fixed
0:17:57 and he says how much is it, it’s 1,000 bucks.
0:18:00 1,000 bucks for that, he’s like it was a $5 party.
0:18:03 He’s like yeah, the par was $5 but 995 was knowing
0:18:05 where to tap and where to put it.
0:18:07 And so I think that there’s still a lot
0:18:10 of this like tacit knowledge and craft and maintenance
0:18:13 that is gonna be protected against the rise of the machines
0:18:15 that are gonna replace most of the white collar
0:18:16 intelligence and knowledge workers.
0:18:19 – Do you believe, I think it was Zaku who came out
0:18:22 and said by mid this year, 2025,
0:18:26 that AI would be as good as a mid-level engineer.
0:18:27 – Encoding.
0:18:28 – Encoding.
0:18:29 – Yeah, for sure.
0:18:30 – What are the implications of that?
0:18:34 Walk me through like the next 18 months if that’s true.
0:18:36 – Well, again, if you take this in the frame
0:18:38 of Jevons Paradox, then a lot more people are going
0:18:41 to be able to code in ways that they never have.
0:18:44 And in fact, I think it was like Andre Caparthi
0:18:46 who was on Twitter a day or two ago
0:18:49 talking about who he himself as a coder
0:18:51 was basically just talking in natural language
0:18:52 and having the AI, and I forget which one
0:18:55 he was particularly using, generate code.
0:18:58 And then if it tweaks something and he wanted to change
0:19:01 a design and make a particular sidebar
0:19:02 a little bit thicker or thinner,
0:19:04 he would just say make the sidebar and it was able
0:19:05 to do that.
0:19:08 So I think the accessibility for people who never coded,
0:19:11 never programmed to be able to come up with an idea
0:19:12 and say, “Oh, I wish there was an application
0:19:14 “that could do X, Y and Z to be able to quickly do that.”
0:19:15 Is great.
0:19:17 For the big companies who employ many coders
0:19:20 that are competing at an ever faster speed.
0:19:21 You know, you have somebody like Mark Benioff
0:19:23 who’s saying that they’re not hiring any more coders
0:19:25 at the same time that he’s still talking about
0:19:27 the primacy of SAS, which is this weird contradiction.
0:19:30 But I would suspect that maybe you lose 10 to 30%
0:19:33 of the people that you normally would have hired,
0:19:35 but the people that are there are still like
0:19:38 these 10 X coders and now they have a machine
0:19:40 that’s helping them be like 20 to 100 X.
0:19:42 Do you think margins go up then
0:19:43 for a lot of these companies?
0:19:44 I don’t know.
0:19:46 I always feel like margins are always fleeting in the sense
0:19:48 because it’s like a fallow safe composition.
0:19:50 One company stands up a little bit higher
0:19:52 and then everybody else is on their tippy toes.
0:19:54 So I think it just changes the game,
0:19:57 but I don’t think that you have some like permanent margin.
0:19:59 The only time you get really large margins
0:20:02 is when you truly have like a monotonic NVIDIA today.
0:20:03 Until there’s an alternative,
0:20:06 whether in architecture or algorithms or in something else,
0:20:07 you know, they’ve got dominant margins
0:20:10 because they can charge super high prices
0:20:12 because there is no alternative.
0:20:15 So when you have that, there is no alternative,
0:20:19 but in many domains, given enough time,
0:20:21 there’s an alternative and then margins just resettle
0:20:24 and look at cars, you know, the average margin on cars.
0:20:27 Cars today are 10,000 times better
0:20:29 by every measure of fuel efficiency,
0:20:33 comfort, air conditioning, satellite radio,
0:20:35 but those margins never persisted
0:20:37 as being like permanently high.
0:20:38 I always come back to sort of buff it
0:20:40 in the ’60s with the loom, right?
0:20:42 I always relate everything to the loom
0:20:43 ’cause everybody was coming to him
0:20:44 when he first took control and they’re like,
0:20:45 “Oh, we got this new technology.”
0:20:47 And he’s like, “Yeah, but all the benefits
0:20:49 “gonna go to the customer, it’s not gonna go to me.”
0:20:52 I mean, look at AI right now.
0:20:54 Billions of dollars being spent.
0:20:56 All the foundation models, all the competition,
0:20:57 the second that deep seat comes out,
0:21:00 it suddenly accelerated the internal strategic decisions
0:21:02 from open AI of when are we gonna release models?
0:21:06 And so the flurry of all of this benefits us as consumers,
0:21:07 always and everywhere.
0:21:11 And so we were looking at internally
0:21:13 installing some new AI system
0:21:16 to surface all of our disparate documents.
0:21:17 And there’s a bunch of these.
0:21:22 And our Gmail, our Slack, our Google Docs,
0:21:23 our PDFs, our legal agreements
0:21:25 and just have a repository with permissions
0:21:28 and all this, and it’s expensive
0:21:29 in part because going back to that
0:21:31 Univerance arbitrage, somebody could charge us
0:21:33 a lot of money to do that implementation.
0:21:35 And my default was just wait.
0:21:36 Why don’t we wait six months?
0:21:37 Because this is gonna be available
0:21:40 from all of the major LLM providers today
0:21:42 that want to get the enterprise accounts.
0:21:44 And let’s just wait and they’ll compete and compete
0:21:46 and it will accrue to us as users.
0:21:47 – Talk to me about all these models.
0:21:50 People are spending hundreds of billions,
0:21:52 if not trillions of dollars around the globe,
0:21:54 competing on a model.
0:21:56 Do you think that’s the basis of competition
0:21:57 or how does that play out?
0:22:00 And then you have Zaku’s trying to open source it
0:22:03 and he spent, I don’t know what 60 to a hundred billion
0:22:07 probably by the end of June this year, open sourcing it.
0:22:08 So he’s basically like,
0:22:09 I wouldn’t say he’s doing it for free,
0:22:11 but what’s the strategy there?
0:22:14 – First you have just straight head-to-head competition,
0:22:15 you know, Anthropic and Claude
0:22:18 and Chajapiti and opening AI and others.
0:22:20 Then you have sovereign models.
0:22:21 So there are countries that are saying
0:22:24 we don’t want to be beholden to the US or to China.
0:22:27 We find a company in Japan called Sakana.
0:22:27 This is one of the lead authors
0:22:30 from the Google Transformer and a guy, David Ha
0:22:32 and just incredible team.
0:22:34 And they are actually trying to do
0:22:37 these super efficient novel architecture.
0:22:37 So they’re not trying to treat
0:22:39 in these multi-hundred thousand clusters.
0:22:40 Their latest model,
0:22:42 which was based on this evolutionary technique
0:22:45 was like eight GPUs, which was wild.
0:22:46 So that’s one trend.
0:22:48 But on the strategic question for Zaku,
0:22:50 I actually think that he’s probably playing
0:22:52 at the smartest of everybody, which is,
0:22:53 and he’s been open about this.
0:22:56 We’re going to open source the models with Lama
0:22:58 and we’re going to let people develop on them.
0:23:01 Why? Because the real value is going to be
0:23:04 in the repository of data.
0:23:06 Longitudinal data, deep data.
0:23:07 If you go back 10, 15 years,
0:23:09 like the number one thing in tech was big data,
0:23:10 big data, big data, okay?
0:23:12 Well, now if you actually have big data,
0:23:14 you want to use whatever models are out there
0:23:16 to run on your proprietary silo of data.
0:23:18 So the people that I think are going to be advantaged,
0:23:19 Meta, why?
0:23:21 They’ve got all my WhatsApp messages.
0:23:23 Apple doesn’t, Meta does.
0:23:25 They’ve got all my Instagram likes and preferences
0:23:27 and every detail of how long I spend
0:23:29 and linger on something and what I post
0:23:30 and all of that content.
0:23:32 My Facebook, which I don’t really use anymore
0:23:34 other than when Instagram, you know, cross posts to it.
0:23:36 But that is super valuable.
0:23:39 And they care about that in part
0:23:43 because Zaku needs to route around both Apple and Google.
0:23:44 He does not have a device.
0:23:47 I mean, you’ve got Oculus and MetaQuest and whatnot,
0:23:48 but that’s not the one.
0:23:51 This Orion with the neural band
0:23:53 from the company control labs that we funded,
0:23:55 which was for the non-invasive brain machine interface
0:23:57 to be able to use free gestures,
0:23:59 which is an absolute directional arrow of progress, right?
0:24:02 Disintermediating the control surfaces you have,
0:24:03 remote controls and all that kind of stuff
0:24:05 and just being able to gesture, map a device
0:24:08 to your human body is absolutely the trend.
0:24:10 But he’s thinking about how do I route around these devices
0:24:12 and how do I have a long repository
0:24:13 of everybody’s information
0:24:15 and use the best model that’s out there.
0:24:16 And the great thing about open source
0:24:18 is it’ll continue to improve over time.
0:24:20 So I think that that’s a winning strategy.
0:24:22 I think the people that are continuing
0:24:23 to develop ever better models
0:24:25 unless they have proprietary data
0:24:27 are gonna be sort of screwed.
0:24:28 Bloomberg should do really well.
0:24:31 I mean, the huge amount of proprietary information,
0:24:33 all the acquisitions that they’ve done over time,
0:24:36 being able to normalize merger data
0:24:37 and historic information
0:24:40 and the longitude of price information
0:24:42 and correlations between different asset classes,
0:24:46 being able to run AI on top of that is like a quant’s dream.
0:24:50 So I think that people that have hospital systems
0:24:52 arguably some governments have used efficiently,
0:24:55 but anybody that has a proprietary source of information,
0:24:57 clinical trials, failed experiments
0:24:59 inside of pharma companies,
0:25:02 being able to do that is the real gold.
0:25:05 And the large language models are effectively like,
0:25:07 over time, I think going to trend towards zero
0:25:09 in a commodity excavator of that data.
0:25:11 So the mode is really in the data.
0:25:12 I think so.
0:25:14 Because everything will be sort of comparable
0:25:15 running on top of that.
0:25:17 The data sitting by itself is like an oil well
0:25:20 that isn’t mined, you know, we’re not gas finding,
0:25:21 that is in fact.
0:25:23 So it needs to be extracted.
0:25:25 And I think that most likely open source,
0:25:27 but in some cases enterprise partnerships
0:25:29 between anthropic or open AI
0:25:32 with some of these siloed data sets
0:25:34 will unleash a lot of value.
0:25:35 – So aside from meta,
0:25:38 what counterintuitive sort of public companies
0:25:42 would you say have like really interesting data sources?
0:25:43 – Ah, that’s a good question.
0:25:45 I haven’t really spent a lot of time on that
0:25:49 to figure out who’s got crazy amounts of proprietary data.
0:25:50 Pharma would be a good one
0:25:52 because obviously they’re, you know,
0:25:54 tracking both their successful
0:25:56 but their unsuccessful clinical trials.
0:25:59 There’s a lot of information in the unsuccessful data,
0:26:01 like the things that fail that you can learn from.
0:26:02 You could argue that Tesla, of course,
0:26:04 who I’m very publicly critical of,
0:26:05 like if they truly are collecting
0:26:09 a ton of road user data from,
0:26:11 you know, every Tesla that’s being driven,
0:26:12 that would be valuable.
0:26:14 Anything where there’s a collection,
0:26:17 a set of sensors, a repository of information
0:26:19 that is owned by them.
0:26:21 Anything that we’ve signed off on that, you know,
0:26:25 your data is free for us to use, like meta.
0:26:26 – You think of Tesla,
0:26:28 they should have the best mapping software in the world.
0:26:32 They literally like drive millions of miles every day.
0:26:33 They can update everything.
0:26:34 They can locate police.
0:26:36 They can locate speed cameras.
0:26:37 They can get real time traffic.
0:26:38 – Weather patterns, yeah, totally.
0:26:39 – Yeah.
0:26:41 – But okay, the flip side of that though, right?
0:26:45 Taking sort of like the opposite view for a moment, Netflix.
0:26:46 Netflix has all of our viewing data.
0:26:47 They know what we like.
0:26:48 They know what you like.
0:26:50 They can make a perfect set of channels for you.
0:26:53 And the recommendations are reasonably good approximations
0:26:55 of adjacencies to things that you liked,
0:26:57 but they haven’t been successful,
0:27:01 nor has the human algorithm at say HBO in the past,
0:27:04 of like perfectly creating the next show
0:27:06 that you really want to see.
0:27:09 And what’s interesting about that is,
0:27:11 they’ve put a lot of money into this,
0:27:14 but it hasn’t yielded the recipe maker
0:27:16 for like the next perfect show.
0:27:18 And oftentimes the thing that you want to see
0:27:20 is almost something that’s orthogonal
0:27:22 from what you’ve been watching.
0:27:27 Like I heard Anthony Mackie, who’s in the latest Marvel movie
0:27:29 talking about an expectation that he’ll be,
0:27:30 somebody was like, how long do you think
0:27:32 you’ll be doing Marvel movies?
0:27:34 And he’s like, oh, I think probably like the next 10 years.
0:27:37 And I’m like, probably two or three,
0:27:38 because people are just bored of this stuff after a while.
0:27:41 Like nobody wants to see another,
0:27:42 I don’t want to see another Marvel movie.
0:27:45 I like the adjacency of like The Boys,
0:27:47 which was like the dark superhero kind of movie.
0:27:50 And I think trying to find groups
0:27:52 that have proprietary data
0:27:56 that have some predictive value,
0:27:58 the most value probably for society,
0:28:00 I don’t know if it’ll be entirely captured by companies,
0:28:03 is just all the scientific information that we have.
0:28:04 Because I’m absolutely convinced
0:28:09 that we are going to have machines doing science 24/7.
0:28:10 – Well, so talk to me a little bit about,
0:28:12 I want to come back to Tesla in a sec,
0:28:13 but let’s go down the science that,
0:28:16 why has nobody sort of taken every study published
0:28:19 in a domain say Alzheimer’s research,
0:28:22 popped it into GPT and be like, where are we wrong?
0:28:25 What studies have been fabricated or proven not true
0:28:27 that we’re investing research in, right?
0:28:29 Like, ’cause studies get built on studies and studies.
0:28:32 And so if something from the 80s came out
0:28:33 and it’s like completely false,
0:28:37 we’ve probably spent $20 billion down this rabbit hole.
0:28:39 And what’s the next most likely thing to work?
0:28:41 Is anybody doing that?
0:28:43 – I have to imagine they are because deep research
0:28:46 came out today or in the past 24 hours from OpenAI,
0:28:48 which is sort of their model with a better engine,
0:28:50 so to speak, than Google’s deep research,
0:28:51 which itself was impressive,
0:28:53 both because of its ability to search many sources
0:28:55 and then the ability to sort of,
0:28:57 I think it was either there or through Nopakellum
0:28:58 to conjure the podcast,
0:29:01 which at first was a static presentation
0:29:03 of near human quality voice,
0:29:04 but now you can interrupt it like a radio call,
0:29:05 which is super cool.
0:29:10 But you can say go through the past 15 years of PNAS papers,
0:29:14 or science and nature papers around this particular topic
0:29:16 and find correlations between papers
0:29:18 that do not cite each other
0:29:20 or tell me any spurious correlations.
0:29:21 And the beauty of all of that
0:29:24 on sort of the information or informatics side
0:29:27 is eventually you will have a materials
0:29:29 and methods output of that,
0:29:32 that you can feed into something like bench-ling
0:29:35 or some of the automated lab players
0:29:37 to actually say like run the experiment.
0:29:40 So I’m absolutely convinced, like high-certitude,
0:29:42 I don’t know exactly which company will do it.
0:29:44 We’ve invested in some, they haven’t worked,
0:29:46 we’ll invest in more, hopefully they will.
0:29:49 But this directional hour of progress of the idea
0:29:54 by analogy of machines doing science 24/7 automated
0:29:55 is going to happen.
0:29:56 I’ll give you one or two analogies.
0:30:00 If you were a musician back in the day,
0:30:01 you know, if you and I were starting a band,
0:30:04 we would have to go and get studio time here in New York City
0:30:05 or Electric Ladyland or whatever,
0:30:07 you bring your instruments, okay,
0:30:09 maybe you could rent the instruments there
0:30:12 and then GarageBand and Logic and Pro Tools pops up.
0:30:15 And now we don’t have to be in the same physical space.
0:30:17 My instrument is virtualized.
0:30:20 I can create a temporal sequence of notes.
0:30:22 I can layer them in, you could play drums,
0:30:23 you could do vocals, blah, blah, blah, okay.
0:30:25 Science is the same thing.
0:30:26 I can be on the beach in the Bahamas
0:30:28 and conjure a hypothesis
0:30:31 and use one of the AIs to test the hypothesis
0:30:34 and look at past literature searches,
0:30:36 see freedom to operate, see if there’s white space
0:30:39 and then tell one of these cloud labs
0:30:41 that is literally like sending something to AWS
0:30:44 back in the day and say, run this experiment.
0:30:46 And the beauty of this is the robot will do the,
0:30:48 well, here’s the beauty, the virtue of the vice.
0:30:49 The robot will do the experiment
0:30:51 and it should do it perfectly because it’s digital
0:30:54 and it’s high fidelity.
0:30:56 The vice is so much scientific breakthrough
0:30:59 has often happened because of serendipitous screw ups.
0:31:02 And so you want almost to engineer
0:31:04 like a temperature on an AI model,
0:31:06 a little bit of stochastic randomness
0:31:08 so that the machine can sort of screw it up
0:31:09 to see what might happen because, you know,
0:31:13 penicillin and Viagra and rubber and Vulcan,
0:31:15 all these things happen by like random processes
0:31:17 and then post-fact, we’re like, huh, that’s funny.
0:31:19 And then, you know, you run with it.
0:31:22 But then the machine will say, here’s the results.
0:31:23 And it will then reverse prompt you
0:31:25 and say, do you want to run the experiment again
0:31:27 but changing the titration of this
0:31:28 to 10 milliliters instead of five?
0:31:30 And you just click a button from the beach
0:31:31 and you’re like, yes, and the robots run it.
0:31:35 Whoever ends up creating and building that,
0:31:36 I think is going to make a fortune.
0:31:37 – Well, you don’t even have to decide.
0:31:39 The robots could decide, yes, right?
0:31:40 – Totally.
0:31:41 – And you’re sort of out of the loop
0:31:42 and it just outputs science.
0:31:43 – Totally.
0:31:45 – That would be so interesting.
0:31:46 Before we get to that point,
0:31:47 we’ll probably get to the point
0:31:49 where models make themselves better.
0:31:51 Is that the point where it really starts
0:31:54 to go like parabolic almost?
0:31:55 – I don’t know.
0:31:58 I definitely see that models can improve
0:32:01 ’cause you can even argue like deep seeks R1
0:32:03 is a model that was improving upon outputs
0:32:06 from chat GPT and so on.
0:32:08 So I definitely think that there will be
0:32:10 this recursive improvement,
0:32:14 but you’re still going back to being rate limited by time
0:32:17 and biological or chemical reactions.
0:32:20 You still need to instantiate this
0:32:21 into a physical experiment.
0:32:23 And so you can model and simulate all you want,
0:32:26 but then you actually have to like do the thing
0:32:27 and make the compound.
0:32:31 And so those still take steps and organic chemistry
0:32:33 and there’s like 20 reaction steps
0:32:36 and people optimize to like reduce them down to six.
0:32:38 And you still need the physical reagents
0:32:41 and the right temperature and the experimental design.
0:32:43 So I still think that that’s gonna be the bottleneck,
0:32:46 but for sure like the ideation and experimental design
0:32:49 is gonna, that’s just gonna absolutely explode.
0:32:52 And then you’ll have these automated labs
0:32:53 with lots of different instruments
0:32:55 where robots will be able to take out
0:32:57 a sample from a centrifuge and put it into the next thing.
0:33:00 And like you don’t need humans to do that
0:33:02 any more than you need humans to assemble
0:33:04 sophisticated iPhones.
0:33:05 I mean, we still have very cheap labor
0:33:06 and Foxconn factories in China
0:33:08 and Vietnam and elsewhere now doing that,
0:33:11 but there’s no reason for that over time.
0:33:12 Isn’t that low hanging fruit though?
0:33:15 Just like look at all the work that we’ve done so far
0:33:17 and tell us where we’re on the right track
0:33:19 and where we’re sort of like we’re going astray.
0:33:21 And like there’s nothing preventing that
0:33:22 from happening today.
0:33:24 I mean, it’s very like David Doi chains.
0:33:25 Like if it obeys the laws of physics,
0:33:27 that should be possible, there’s nothing about this
0:33:29 that is like totally speculative and fantastical
0:33:31 that it doesn’t obey the laws of physics.
0:33:33 – The other project that I wanted to,
0:33:35 I was thinking about sort of doing
0:33:38 is just calling like a prior art.org or something
0:33:41 and having AI read through all the patents
0:33:44 and make the next adjacent sort of like patent like
0:33:46 improvement and then just publish it
0:33:47 ’cause then there’s prior art.
0:33:48 – Right.
0:33:50 Well, yeah, there will for sure be AI patent trolls
0:33:52 if you put it negatively.
0:33:53 – This would dissuade that.
0:33:55 I mean, in a sense, it would sort of be making
0:33:59 prior art for as much as you can 24/7.
0:34:01 – There are companies that do this
0:34:02 where they have creative patent filers
0:34:04 for continuations in part so that they can keep
0:34:06 sort of the life of this going.
0:34:10 But the rise of agentic AI,
0:34:13 you can have some crazy idea, some brain fart it,
0:34:15 you know, 9/30, 10 o’clock a night.
0:34:17 And you just had, you know, cocktail with friends.
0:34:20 You’re like, oh, like imagine, you know, if that exists.
0:34:22 Well, do the research, does this exist?
0:34:25 And if it doesn’t exist, can you, you know, write a patent
0:34:27 and sketch a diagram for me and file it
0:34:29 and starting incorporate a company.
0:34:31 Now, all those things have to go at the speed
0:34:33 of like certain processes,
0:34:35 but all of that could be done overnight
0:34:36 where you literally wake up
0:34:39 and there are multiple agents working on your behalf
0:34:42 that have filed a patent, created a design,
0:34:45 incorporated a company, possibly even put it out
0:34:48 to some group and raised money for it overnight
0:34:49 that have opted into it.
0:34:52 And I don’t know if there was like successful enough
0:34:53 and it hit a bunch of criteria.
0:34:57 I might allocate some capital into an account
0:35:01 to allow a robot AI to actually receive pitches,
0:35:03 respond to it and create a small portfolio
0:35:04 to allocate as an experiment.
0:35:07 So you can see this whole thing is just like a human idea
0:35:10 or maybe one inspired by interactions with an AI
0:35:13 that by the time you wake up, you have a company started
0:35:17 and the basis for people to actually do work.
0:35:20 Like that in 10 or 20 years, people look back
0:35:23 and be like, how did we not see that coming?
0:35:24 And look at all the jobs that are being created
0:35:27 because every single person is now creating
0:35:29 and has like six virtual companies.
0:35:31 – The future is gonna be wild.
0:35:32 – Yeah.
0:35:34 – Talk to me, let’s go back to Tesla for a second.
0:35:36 Why the hate on Tesla?
0:35:40 – Let me say, I think Elon is amazing at certain things.
0:35:45 Elon is arguably the greatest storyteller fundraiser
0:35:50 inspiration for anybody in the past, maybe in all time.
0:35:53 Truly, I think his relationship with the truth
0:35:55 has been questionable.
0:35:58 And so in Tesla particularly, I think there was a time
0:36:01 where the short sellers started to identify things
0:36:03 not because they just hated the company
0:36:04 or hated the future or any of this.
0:36:06 And he was able to very shrewdly weaponize
0:36:08 the us versus them.
0:36:10 They’re trying to kill us, right?
0:36:11 Most short sellers that I know
0:36:13 happen to be very disaffected people
0:36:15 and they have a chip on their shoulder.
0:36:19 And to me, the motivating force
0:36:21 and the incentive for them is not that
0:36:25 they just want to make money, but they want to be right.
0:36:27 And they want to be right because they’ve identified
0:36:30 somebody that they think is intentionally doing wrong.
0:36:32 It’s the same thing as an investigative journalist.
0:36:36 It’s the same thing as a opposition research
0:36:37 for a politician.
0:36:38 It’s the same thing for somebody
0:36:43 that is trying to debunk a Sunday preacher charlatan
0:36:46 that basically is almost intellectually competitive
0:36:49 to say you are trying to pull the wool
0:36:52 over these people’s eyes and I know what you’re doing
0:36:54 and I’m going to call you out on it.
0:36:58 And so for me with Tesla, I think that they got away
0:37:00 with accounting fraud on warranty reserves
0:37:02 and a whole bunch of other things.
0:37:04 I think there was a lot of presided digitization
0:37:07 and magic of look over here while we’re doing this.
0:37:10 And today it doesn’t matter because they got away with it.
0:37:14 But I think that there was not the same kind of honesty
0:37:18 that I would ascribe to Jeff Bezos in how he built Amazon
0:37:21 and raised a few hundred million dollars of equity.
0:37:22 You could look at stock-based compensation.
0:37:24 You could look at debt as capitalization,
0:37:29 but created this monster that is profitable cashflow positive
0:37:31 and never raised another dollar of equity.
0:37:33 And Elon raised north of $50 billion,
0:37:38 took out $50 billion, treated it like an ATM,
0:37:40 said I’m never selling a share and sold lots of shares.
0:37:43 And then whether he had to do it to buy Twitter or whatever.
0:37:46 I just, I don’t feel it was done as honestly
0:37:48 as other entrepreneurs that I greatly admire.
0:37:51 Now that said, SpaceX, I have no issue with.
0:37:53 I think SpaceX is an extraordinary company.
0:37:57 I think it’s an incredibly important American company.
0:37:59 I think without it, we would be at a massive disadvantage.
0:38:01 I think it is truly a national treasure,
0:38:03 run by Gwynne Shotwell, incredible engineers.
0:38:05 We’ve backed a bunch of these engineers
0:38:07 that have come out from Tom Mueller
0:38:10 to that I just, I think the world of.
0:38:13 I’ve just, yeah, I’ve been much more critical
0:38:15 about Elon’s relationship with the truth
0:38:17 as it came to Tesla and in many ways,
0:38:19 I felt like the whole thing was unnecessary.
0:38:21 – Do you think those are one-time things
0:38:23 or they’re systemic and they crop up
0:38:24 every few months or something?
0:38:25 – In his personality?
0:38:26 – Yeah.
0:38:27 – He’s past the stratosphere now.
0:38:32 Like it’s, you know, he’s proximate to power in ways
0:38:34 that people can’t compete with.
0:38:39 If you’re an investor in or you’re Sam Altman in open AI,
0:38:40 you’re not only worried about competition,
0:38:42 you’re worrying about a personal grudge
0:38:44 from somebody who has the ear of the president
0:38:48 that can weaponize all kinds of systems of power
0:38:51 from the DOJ to the FTC to the FBI.
0:38:54 And I would be very nervous
0:38:56 being an adversary with that kind of power.
0:38:57 So.
0:38:58 – Altman came out and said he didn’t think
0:39:01 Elon would use that power against him.
0:39:03 – Which is a nice and smart thing
0:39:06 and a necessary thing to say publicly.
0:39:07 And I think that Elon has even said like,
0:39:09 I won’t use that, but.
0:39:11 – What’s the saying, power corrupts?
0:39:12 – Yeah, an absolute power corrupts, absolutely.
0:39:15 But I don’t know, like you’re in a position
0:39:20 of power at DOJ and OMB and Office of Personnel
0:39:23 and are you, and you have influence
0:39:26 and you can shut some of these things down.
0:39:27 Does Elon love the SEC?
0:39:31 You know, he’s been pretty vocal about that institution.
0:39:35 Does he love the National Highway Transit Safety?
0:39:37 So if these things are gutted, you know,
0:39:39 I think you’ve got more free reign
0:39:41 to shut down criticism.
0:39:43 You know, you’ve got similarly the best entrepreneurs
0:39:47 when short sellers are like, you know, saying something,
0:39:48 they don’t want to ban short selling.
0:39:50 They don’t want short sellers to be arrested.
0:39:52 They just prove them wrong.
0:39:53 And so.
0:39:55 – My favorite story about that was Brad Jacobs
0:39:59 and his, the short report came out at the stop draw.
0:40:04 Precipitously, he borrowed $2 billion, bought back shares.
0:40:06 – Yeah, this is big skin in the game, right?
0:40:07 – Right, totally.
0:40:09 – We’re gonna double down and go forward.
0:40:12 I think he turned that two into 10 or eight or something.
0:40:13 Like it was just crazy.
0:40:15 – So for me, I don’t know, 13 or so,
0:40:19 like after I was bar mitzvah, I became atheist.
0:40:22 And I just wouldn’t, I would see like these preachers
0:40:24 exploiting people.
0:40:26 It just like irked me and it irked me in this.
0:40:28 It wasn’t in some, I reflected on this over the years.
0:40:32 It wasn’t in some virtuous holier than now kind of thing.
0:40:33 It was intellectually competitive.
0:40:35 It’s like, I see what you’re doing.
0:40:38 You’re running a con and I wanna call it out.
0:40:41 And it wasn’t rooted in like self virtue
0:40:43 of like pursuit of truth.
0:40:45 The real thing when I like thought about it is,
0:40:48 no, I wanna show that you’re cheating people.
0:40:50 – Short sellers are necessary to a well-functioning market,
0:40:51 right? – I think so.
0:40:52 – We need to hear both sides of the story
0:40:54 and make our own judgments and decisions.
0:40:56 What point do you think that computers
0:40:58 are gonna really make most of the investing decisions?
0:41:02 – Well, you could argue today they are,
0:41:04 not because they’re doing reasoning and analysis
0:41:06 and fundamental work, but because the structure
0:41:11 of the market is so dominated by passive indexation.
0:41:13 And that is effectively an algorithm.
0:41:18 And that algorithm says $1 in buy, $1 out sell.
0:41:21 And in both cases, indiscriminately.
0:41:24 And so you just have a flood of money
0:41:27 that goes into the market and these indices buy everything.
0:41:29 And it becomes this massive market cap weighted,
0:41:32 accelerant and then people say sell
0:41:33 and then the money just comes out.
0:41:36 So the past, I don’t know, 10 plus years
0:41:37 where this has really become the case
0:41:40 with Fidelity and BlackRock and State Tree
0:41:42 and others that the ETFs, which were well-intentioned,
0:41:44 you know, you go and listen to Buffett back in the day.
0:41:46 It’s like, just put it in the market, right?
0:41:49 It’s hard for active managers to out-compete.
0:41:52 Definitely the case for the past 10 or 15 years.
0:41:55 But I do think that we will see a return
0:41:59 to active managers that are able to discriminate
0:42:01 true fundamentals in part because I think
0:42:03 that the cost of capital is just going to rise
0:42:06 and all the funny money of the past 10 years
0:42:07 is going to wash out.
0:42:09 – Two questions, two rabbit holes.
0:42:09 I want to go down here.
0:42:11 One, at what point do you think active managers
0:42:14 and analysts is replaced by AI in the same way
0:42:17 that Zuck is saying an engineer at Metta
0:42:19 is going to be replaced by AI?
0:42:23 – Already, there are AIs that can not only go through Qs
0:42:27 and Ks, 10 Qs and 10 Ks and can listen
0:42:30 to quarterly earnings reports and CEOs
0:42:33 that are talking at conferences or on podcasts
0:42:35 and can get an emotional sentiment,
0:42:37 can see where they’re varying their language
0:42:40 in ways that only the subtlest of analysts in the past
0:42:42 or portfolio manager could do.
0:42:44 And I think the most valuable thing that AI is going to do
0:42:47 when you ask it questions and it comes up with the answers
0:42:49 and assuming those answers are accurate
0:42:51 and cross-correlated and double-checked,
0:42:55 they actually say, here’s the five questions you didn’t ask.
0:42:58 And so that is going to unleash real insight.
0:43:00 Now, there is still this human aspect
0:43:03 of being able to look at somebody and decide,
0:43:05 do I trust them or not?
0:43:10 And I think that the best analysts are able to say,
0:43:12 very Buffett-like or Joel Greenblatt-like,
0:43:14 is this a good business?
0:43:15 And there’s ways to measure that,
0:43:17 like a fundamentally good business,
0:43:18 even if an idiot was running it,
0:43:21 and then do I think it’s had a good price
0:43:23 and is therefore my expected return going to be high
0:43:25 and do I trust the people that are running it
0:43:27 because ultimately I am allocating capital to them
0:43:29 in the same way that somebody allocates capital to us.
0:43:31 I like the virtue of our private markets
0:43:34 because I am less, we’re still beholden,
0:43:37 but far less beholden to a day-to-day market,
0:43:42 Mr. Market, fluctuation of manic depressive positivity
0:43:44 or pessimism.
0:43:46 We have 10-year locked funds.
0:43:48 We’re able to make long-term bets.
0:43:50 It’s arguably this great source of time arbitrage
0:43:52 when everybody else is looking and discounting back a year
0:43:54 or 18 months or two years.
0:43:57 But you think about the three main sources of edge
0:43:58 and we’ve talked about this in the past,
0:44:01 but informational, analytical, and behavioral.
0:44:04 Informational advantage used to exist a long time ago,
0:44:08 regulations like Reg FD and avoidance of insider trading
0:44:12 and information that tried to equalize the playing field
0:44:15 in addition to a huge influx of really brilliant people
0:44:19 that are able to use cutting edge data tools.
0:44:21 Having an information advantage is really hard.
0:44:23 Having an analytic advantage where AI can play a role
0:44:26 in that, let’s assume we all have the same information
0:44:28 and I don’t have any better intel or information.
0:44:30 Like for example, just inside on the information advantage,
0:44:33 there was a hedge fund, which I won’t name,
0:44:36 which was very cleverly going and actually buying stuff
0:44:38 online from Adobe.
0:44:41 Every time they did, they got a piece of legal
0:44:45 inside information, which is Adobe’s web URL,
0:44:49 actually, when you made a purchase for Creative Cloud,
0:44:52 would tell you 4,723 or whatever.
0:44:54 And then like six hours later, they would go,
0:44:56 and it was like 4,000,000, and so they could infer
0:45:00 and extrapolate what the sales were based on this
0:45:02 because when they bought it six days later
0:45:04 or six hours later or whatever they saw
0:45:07 where they were in the queue and you can sort of extrapolate.
0:45:10 That was legal to do, but once that signal leaks out
0:45:11 and somebody’s like, oh, that’s a clever way
0:45:15 to figure that out, then it rapidly erodes.
0:45:17 So informational advantage, hard analytical advantage,
0:45:20 brilliant people combined with brilliant technology,
0:45:22 really hard, and then it goes into this last one,
0:45:25 which is the behavioral, and that, to me,
0:45:27 is the persistent thing.
0:45:30 Now, AIs over time will understand us in many ways better
0:45:32 than we understand ourselves.
0:45:35 Google already arguably knows more about you
0:45:37 than the closest people in your life based on the things
0:45:40 that you search for, search in private for,
0:45:41 and AI will as well.
0:45:44 And already there’s an eerie moment that I appreciate
0:45:47 because I’ve given myself over to the information gods.
0:45:51 There’s a required energy to try to maintain privacy,
0:45:53 and I just feel like it’s not worth it.
0:45:55 We can talk about privacy ’cause most people basically
0:45:57 just wanna keep private, their sex life,
0:46:00 their bathroom time, and how much money they have,
0:46:03 unless you are super rich or super poor.
0:46:05 Because if you’re super rich, you broadcast.
0:46:08 You’re on the Forbes 100, 400.
0:46:09 You’re showing the house you just bought,
0:46:11 the art you just bought, you’re signaling your wealth.
0:46:14 And if you’re broke, then you’re really poor.
0:46:15 There’s people that are on Twitter like,
0:46:17 I’m dead ass broke, I have no money,
0:46:19 and they literally, it’s the people that are in the middle
0:46:21 that are middle class, but want people to think
0:46:23 that they’re upper class, or people that, you know.
0:46:26 And so everything else for privacy,
0:46:27 I think is like out the window.
0:46:30 But the reason I was saying this is,
0:46:32 I’ve given myself over to the information gods,
0:46:35 and when I go to ChatGPT, because it has memory,
0:46:36 and it’s constantly updating it,
0:46:38 there are times where I remember this thing,
0:46:39 and I’m like, how did you know that?
0:46:41 And I forgot that it was like from a search three months ago
0:46:43 where I mentioned something about my kids,
0:46:44 and a place that I like to vacation.
0:46:46 And part of me actually appreciates
0:46:48 that that repository is compounding,
0:46:51 but it does sort of scare you
0:46:54 because I remember the things that we talked about.
0:46:56 And then if you were like, oh, like I heard you went to that,
0:46:57 I’d be like, well, who’d you hear that from?
0:47:00 But now if I asked the AI, who’d you hear that from?
0:47:01 It would be like, you, you told me that.
0:47:04 – You told me, wait ’til it gets on your device.
0:47:05 – Exactly.
0:47:07 – And then, okay, well, let’s go back up this rabbit hole
0:47:11 a little bit here, back to passive indexation.
0:47:13 So the rise of this is really post 2010, right?
0:47:16 The mass rise of passive indexing.
0:47:20 We’ve never seen, well, we did during COVID,
0:47:22 but there was so much money thrown into the system.
0:47:25 What do you think the second, third order effects of this
0:47:28 will be, especially in terms of volatility
0:47:30 or something unforeseen?
0:47:34 – Well, you saw this a little bit with just how quickly
0:47:38 the market reacted with a single largest one day loss
0:47:41 of five, six, 700 billion dollars within video,
0:47:43 just because of the fear over deep seek.
0:47:47 The fear was a cascading, traditional information cascade.
0:47:49 Oh my gosh, what does this mean for the expectations
0:47:53 we have about demand for compute and CAPEX
0:47:54 and the expenditures that people are gonna have?
0:47:57 Do we need to rethink this mental model?
0:48:00 And so I think that there’ll be things like that,
0:48:03 that whipsaw and shock people.
0:48:05 – Did they become reflexive at some point?
0:48:08 Like maybe at 500 million or 500 billion,
0:48:12 it didn’t, but had that hit 700 or 800 or a trillion,
0:48:16 like does it then start the auto selling and the,
0:48:17 – That’s a good question.
0:48:18 I don’t know.
0:48:21 Obviously systems where there are significant leverage
0:48:23 in the system are ones that are most prone to that,
0:48:25 these sort of Minsky moments where, you know,
0:48:26 things are going fine, going fine,
0:48:28 and then suddenly they just collapse.
0:48:30 And usually that’s where you have a lot of leverage
0:48:31 in the system.
0:48:32 And sometimes it’s hidden leverage.
0:48:34 I don’t know other than some of the two extra,
0:48:39 three X lever ETFs that that’s the case in traditional market.
0:48:42 But where does this go from here?
0:48:45 – I’m not sure on the passive active piece,
0:48:50 what will break this other than if you were to have
0:48:53 widespread news reports of a handful of active managers
0:48:56 that are suddenly beating the market, you know, decisively,
0:48:59 and they’re pointing to the structure of the market.
0:49:02 And so there’s a rebalancing where people start
0:49:04 to shift out of these things.
0:49:05 On the volatility piece,
0:49:08 what’s interesting is over the past five, six, seven years,
0:49:10 maybe five especially,
0:49:14 a lot of LPs, allocators have gone into private credit,
0:49:16 have gone into private equity,
0:49:20 in part because they are mechanisms of muting volatility
0:49:22 and the vicissitudes of the market
0:49:24 because you don’t have daily market to market.
0:49:28 And so there’s been a little bit of this perverse incentive,
0:49:30 but as Buffett says, you know,
0:49:32 what the wise do in the beginning, the fool does in the end,
0:49:33 and then these things get overdone.
0:49:37 And so I’m actually worried about some of those asset classes
0:49:39 where private credit in particular,
0:49:43 I think was wise to do a few years ago and now is overdone.
0:49:45 You have another phenomenon,
0:49:47 which is every major sophisticated,
0:49:51 large private equity firm, Apollo, KKR, Carlisle, et cetera,
0:49:53 are all starting to think about,
0:49:56 or are actively thinking about both permanent capital
0:49:58 in the form of insurance vehicles like Apollo
0:50:01 and accessing retail in a huge way.
0:50:04 That many people see retail being the next wave of this.
0:50:05 – And when you say retail,
0:50:07 you actually just mean normal day-to-day consumers.
0:50:10 – Individual investors that might’ve been on Robinhood
0:50:13 and could never hear before access Apollo or Carlisle,
0:50:16 but in aggregate, you’re talking about
0:50:18 trillions of dollars of investor money.
0:50:21 And so I think that they’re gonna be tapped,
0:50:22 they’re gonna be into these vehicles.
0:50:27 That will present new interesting financial vehicles
0:50:29 because you’re gonna have to find ways
0:50:31 to give people liquidity for these things.
0:50:34 And so I’ve heard about some interesting things,
0:50:36 actually from a friend, Mike Green,
0:50:39 who I think is a really smart practitioner
0:50:42 and student of markets.
0:50:45 He was one of the earliest to this passive active piece.
0:50:46 He was one of the earliest to understanding
0:50:49 the mechanisms behind the scenes for the SPAC movement.
0:50:53 He’s early now to this idea of Uniswap,
0:50:58 which has a certain mechanism that provides liquidity
0:51:01 by having, let’s say like 80 or 90% in treasuries
0:51:03 and 10% in some underlying.
0:51:07 And you’re able to swap out some illiquid thing
0:51:09 for effectively some liquid pool up to some point,
0:51:11 there’s some repricing.
0:51:14 And he’s been thinking about
0:51:17 something that like Apollo is doing with State Street
0:51:19 and that this portends a movement
0:51:22 into these almost artificial,
0:51:23 like if I would have talked about ETF 20 years ago,
0:51:25 people would be like, “What is it?
0:51:26 Don’t we have mutual funds already?”
0:51:28 And they’re like, “No, but they’re gonna go super low fee
0:51:31 and you’ll be able to trade them on a daily basis.”
0:51:34 There’s something here to watch about the flood
0:51:38 of retail money that will go into illiquid alts,
0:51:39 private equity in particular,
0:51:42 and new vehicles that are formed
0:51:45 to be able to provide liquidity because of that.
0:51:48 And I think that that’s both gonna be really interesting
0:51:49 and potentially to your point,
0:51:53 creates something that sets up some massive blow up.
0:51:55 – The other thing that I wanna come back to
0:51:58 in the rabbit hole here is you mentioned
0:52:01 persistent advantage is behavioral.
0:52:02 – Yes.
0:52:04 – Talk about that in the context of humans
0:52:06 and how do we create an unfair advantage
0:52:08 in a world of AI for humans?
0:52:10 Like what are the ways that we can,
0:52:12 like your kids, like how are you teaching them
0:52:15 to navigate this in a way that gives them–
0:52:16 – In advantage.
0:52:18 – An advantage, a behavioral advantage.
0:52:21 – So you go back some years and it was like,
0:52:26 okay, we can still be computers in chess, done.
0:52:27 We can still beat them and go, done.
0:52:28 We can beat them in video games, done.
0:52:31 Okay, but we still have creativity, done.
0:52:34 Now, of course, human creativity is not dead,
0:52:37 but every day I am doing something creative on AIs
0:52:37 that I can’t do.
0:52:40 I cannot paint, I cannot draw, I cannot conjure.
0:52:43 I like taking photographs and I like the composition of that.
0:52:46 But I can engineer prompts,
0:52:48 which itself is an act of creativity
0:52:52 and get the most inspiring muses and results.
0:52:54 And I can take works of art that I like
0:52:56 and put them in and ask it to describe it
0:52:59 and do it six times, particularly in like mid-journey,
0:53:04 and recreate from the prompt some alternative of it.
0:53:05 There was even an artwork that I loved,
0:53:09 which was this mixed mash of superheroes
0:53:11 that looked like it was put through a blender.
0:53:12 And I was gonna buy the artist’s work
0:53:15 and then I couldn’t describe it to Lauren, my wife.
0:53:18 And so I put it into mid-journey and described it.
0:53:19 And then I just clicked a button
0:53:21 and it made four versions of it.
0:53:23 I guess 16 versions ’cause each one was four.
0:53:25 And it was insane.
0:53:26 And I was like, why am I gonna buy this?
0:53:30 ‘Cause I just recreated it and I felt morally bad
0:53:31 ’cause I wasn’t copying.
0:53:33 But I had a perfect description of the style.
0:53:36 And so I thought that that was pretty wild.
0:53:40 So these are tools that I think kids should be using.
0:53:40 They should be learning.
0:53:42 It’s just like a language.
0:53:45 And I think they need to be versed with it,
0:53:49 in part to understand the domains to avoid
0:53:54 because they’re gonna be not void of emotional
0:53:58 or aesthetic or moral value, but you’re not gonna make money.
0:54:01 And so my wife and I debate this all the time about dance.
0:54:03 Dance is amazing.
0:54:09 We started dating and she took me to this dance thing.
0:54:12 I was really not into dance at all at Parsons Theater.
0:54:15 It was this guy, David Parsons at the Joyce.
0:54:18 And he has this performance called “Caught.”
0:54:20 And me being into science and technology,
0:54:21 I’m watching a bunch of people dance.
0:54:22 I’m not really into it.
0:54:27 All of a sudden, it’s a weird ambient sound,
0:54:29 very electronic, very tron-like.
0:54:33 And there’s a guy in white pants and that’s it.
0:54:35 And then a strobe light goes on.
0:54:37 Anyway, and the strobe starts flashing.
0:54:42 And all you see as a viewer is this person floating in the air.
0:54:45 But behind the scenes, what they’re really doing is
0:54:47 jumping, jumping, jumping, jumping perfectly timed
0:54:49 to the choreography of the strobe.
0:54:51 So you see them caught,
0:54:54 but they’re doing this crazy kinetic athletic thing.
0:54:56 And I was just like, that’s super cool, right?
0:54:58 Everybody should see this, right?
0:54:59 And it’s just like inspiring.
0:55:02 Now that said, we had this debate afterwards
0:55:04 because she’s like, you know, these dancers make no money.
0:55:06 And I’m like, well, market forces would say
0:55:09 that there’s too many dancers or there’s not enough demand.
0:55:12 And they collect unemployment insurance for half the year
0:55:14 because they have to work other jobs
0:55:15 and they don’t have this and it’s just wild.
0:55:18 And so we got into a debate about,
0:55:20 what is the societal value of this?
0:55:22 Now, I found it valuable and I would go
0:55:24 and I would pay money and be a patron,
0:55:26 but I wouldn’t want my kids to go and pursue that
0:55:28 unless it was that they were solving for
0:55:32 just the aesthetic passion that they had.
0:55:35 But I think that there’s very few crevices
0:55:38 where AI will not creep in
0:55:41 and either be able to do the thing nearly as good,
0:55:45 including, you think about the compression
0:55:47 that we all enjoy of a Game and Throne’s episode,
0:55:51 the compression algorithm of all of that talent,
0:55:55 set design, special effects, screenwriting,
0:55:57 a lifetime of acting and performing,
0:55:59 you know, just swish down.
0:56:01 One of our companies Runway ML.
0:56:04 You can, and there’s others that you can conjure.
0:56:06 Today it’s 10 seconds, but tomorrow it’ll be two minutes
0:56:09 and full feature films with no key grips, no lighting,
0:56:12 no costume design, no set design, no actors.
0:56:15 And voice is entirely generated by AI.
0:56:20 And so does that strip this art of its soul kind of thing?
0:56:23 I don’t think so.
0:56:26 I think it just creates a new form of art, just like Pixar.
0:56:27 You know, for all the people that were doing
0:56:29 Disney Mickey animations by hand,
0:56:31 and then suddenly we get this 3D rendered graphics
0:56:34 and incredible storytelling, that will be timeless,
0:56:35 but the tools that we will use
0:56:38 will very rapidly replace these things.
0:56:41 So I think our kids should be embracing
0:56:43 and using all of these tools.
0:56:45 The only tool that I restrict them from is TikTok,
0:56:48 but otherwise, and that’s for a variety of bad influence
0:56:50 and Chinese Communist Party.
0:56:54 But otherwise I want them learning how to use every tool
0:56:57 as they would every appliance in a kitchen.
0:56:58 Outside of those tools,
0:57:01 where do you think the advantage comes from?
0:57:03 It is like a networking advantage
0:57:05 in who you know become more pronounced.
0:57:07 I think it’s this.
0:57:08 I think it’s human to human.
0:57:10 I think that if you always can frame things
0:57:12 as like what’s abundant and what’s scarce,
0:57:16 in a world where there’s gonna be an abundance of access
0:57:18 to information and abundance of access
0:57:22 to creative construction of things,
0:57:26 art and literature and movies and just by the way,
0:57:30 as an aside, I also will take boring PTA messages
0:57:33 from our school and I will put them into large language
0:57:35 models and send them to my wife where I’m like,
0:57:36 do this in the style of Matt Levine,
0:57:38 the Bloomberg deli writer,
0:57:39 or do this in the style of Shane Parrish,
0:57:43 or do this in the style of Al Swearingen from Deadwood.
0:57:45 And it’s just, it’s absolutely entertaining and brilliant.
0:57:48 It takes something that’s so boring
0:57:52 and cliched and hackneyed and like just brings it to life,
0:57:52 right?
0:57:53 – I do the same thing.
0:57:54 – So it’s a lot of fun, right?
0:57:55 It actually makes this stuff interesting,
0:58:00 but the advantage is gonna come in,
0:58:01 what do I do with that when I output it?
0:58:04 I enjoy it for a second, but then I share it.
0:58:05 And so all of these things, all the value,
0:58:07 all the market capitalization of all social media
0:58:09 is about sharing.
0:58:10 And I still think that we’re gonna produce,
0:58:13 we’re gonna share and what becomes scarce is this,
0:58:17 is like human connection, because we are still human
0:58:18 and we still want that.
0:58:20 We wanna be hugged and we want intimacy
0:58:22 and we wanna laugh with each other.
0:58:25 I mean, I’ll share just a quick aside from this,
0:58:27 like two or three months before Danny Kahneman
0:58:29 died last year.
0:58:31 Lauren and I went over with a filmmaker
0:58:35 and another woman and we had a dinner with Danny
0:58:39 and his partner, Barbara Tversky, it was Amos, his wife.
0:58:44 And we were talking about aging and getting older
0:58:46 and memories and Danny had this great point about
0:58:49 that the pleasure of pleasurable things
0:58:53 got less pleasurable,
0:58:57 but less than the pain of painful things got less painful.
0:59:01 So for him, the loss of a friend,
0:59:03 you became a little bit anesthetized to it.
0:59:05 The first time you lose a friend, it’s tragic,
0:59:07 but when all your friends are dropping dead,
0:59:10 it’s like, ah, this is just happening.
0:59:12 The first person you know gets divorced
0:59:14 and all these things over time,
0:59:17 the half-life of the pain just decreases.
0:59:19 You’re still losing the pleasurable things.
0:59:21 So food didn’t taste as good and wine didn’t taste as good
0:59:24 and music didn’t sound as good and sleep wasn’t as good
0:59:26 and sex wasn’t as good and all these pleasurable things.
0:59:29 But he thought that the pain was less painful
0:59:32 than the pleasure was less pleasurable.
0:59:33 And so I thought that was interesting inside.
0:59:35 Barbara had a different view
0:59:36 and she’s still alive
0:59:37 and I’m taking some license to share her view.
0:59:42 But she said, no, it’s still painful.
0:59:45 And the main reason she said that pain is painful
0:59:48 is that these memories I have of Amos
0:59:52 or I shared that moment with this person.
0:59:55 And the great human feeling is commiserating
0:59:57 about the thing that we experienced together
1:00:00 or the memory and laughing hysterically
1:00:02 which I still do with my childhood friends
1:00:04 of this shared moment.
1:00:06 And AI will never get that.
1:00:08 Another person won’t get the inside joke.
1:00:11 And so to your question about the advantage,
1:00:13 the advantage is in that human connection
1:00:16 because we are still human and we want that, we pine for it.
1:00:18 And so I thought that was a really profound counter
1:00:22 to Danny’s view, which is that when you lose these people
1:00:27 you lose the partner to amplify that emotion,
1:00:28 a good one or a bad one.
1:00:31 And I think that being able to have like uniquely human
1:00:34 experiences, understand each other, support each other,
1:00:36 that that is still gonna be an advantage.
1:00:38 – A lot of the things you said,
1:00:41 what sort of common thread with them in my mind
1:00:45 is we feel part of something larger than just ourselves.
1:00:46 – Yes.
1:00:48 – Like relate this to working remotely
1:00:52 or and how this interacts with other things, right?
1:00:55 Where we might not feel part of something larger
1:00:57 than ourselves and remote work is a great example
1:01:00 where you sort of, the ability is there
1:01:02 whether you do part-time or full-time
1:01:05 to shut off your laptop and the world looks like you.
1:01:05 – Right.
1:01:06 – You’re not forced to interact with people
1:01:08 with different political views,
1:01:10 different socioeconomic status.
1:01:13 You surround yourself so you don’t feel a part
1:01:15 of something bigger than, and that changes how you vote.
1:01:18 It changes a whole bunch of things in your life,
1:01:18 I would imagine.
1:01:21 I’m speculating here, but I’m sort of like thinking out loud.
1:01:22 – No way.
1:01:25 I think a lot of your values, a lot of our values.
1:01:28 And again, I’m gonna invoke Danny here in a conversation
1:01:32 before he passed away, was that you may think
1:01:36 you think the things you think because you analyzed
1:01:37 and you reasoned or whatever, but no.
1:01:38 The reason you think the thing you think
1:01:41 is because of the five or six most important people
1:01:44 around you and they sort of believe something
1:01:45 and you believe them.
1:01:47 And so again, an information contagion
1:01:48 and you will tell yourself that you believe it
1:01:50 because you really thought deeply about it
1:01:52 and you reasoned through this.
1:01:53 But no, the reality is you believe things
1:01:55 just because there’s the social phenomenon.
1:02:00 So working from home versus working in person,
1:02:02 there are so many, everybody is here Monday
1:02:04 through Friday now.
1:02:05 You know, at first it was Monday through Thursday,
1:02:07 take Friday, now I’m like, look, if you need to leave,
1:02:10 you go to family first, it’s a principle here,
1:02:13 never miss a concert or recital, a science fair,
1:02:15 but we need to be together.
1:02:16 Why?
1:02:17 Because there’s so many interstitial moments.
1:02:20 There’s a chance serendipity moment
1:02:21 because I come out of a meeting
1:02:24 and I’m able to introduce you to Grace
1:02:26 or Brandon is meeting with somebody and every day,
1:02:27 hey, do you have a minute?
1:02:28 You know, knock, knock,
1:02:29 and then somebody’s making an introduction
1:02:30 and you just never know what it unlocks.
1:02:33 That never happens on Zoom or on calls, it just doesn’t.
1:02:35 The structure of that doesn’t allow for that serendipity
1:02:37 and those sort of human connections.
1:02:41 The ability to really feel when somebody swallows
1:02:43 when you ask them a question and they’re feeling nervous
1:02:46 or like, hey, like something going on, you know,
1:02:47 and they’re, because everybody’s fighting
1:02:50 some epic battle, you know, they’ve got relationship issues,
1:02:53 they got parents issues, they got a sick person, like,
1:02:55 and we are just, we’re still human.
1:03:00 So I feel deeply that we should all be connected in person.
1:03:01 And I do think that that’s an advantage
1:03:05 in a world of abundant AI and sort of cold sterile,
1:03:08 even if it has the simulacrum of,
1:03:10 and again, a lot of people will use AI
1:03:13 as a beneficial way to share things
1:03:14 that they might not even be comfortable sharing
1:03:18 with a person and have a consultant therapist to tell them.
1:03:20 But I’ll give another example,
1:03:24 which is another colleague here moved from one city
1:03:27 to another and he happens to be religious observant,
1:03:29 but as an atheist, okay?
1:03:32 But he moved to this new city and he found a tribe
1:03:35 that he’s like, I immediately plugged in.
1:03:37 Like we didn’t know anybody, we didn’t have any friends,
1:03:40 but by being part of this religion,
1:03:42 we instantly had friends and peers.
1:03:45 And I was like, not that it’s cynical or selfish
1:03:49 or, you know, we could put a valence of meaning behind it,
1:03:51 but really at the end of it is like,
1:03:52 will you help me?
1:03:53 Is there reciprocity?
1:03:56 And that’s this like ancient sense
1:04:00 of whether it’s transactional or not, it’s over or not,
1:04:02 you’re signaling the depth of the sacrifices
1:04:03 that you would make for the group.
1:04:06 This idea that something is bigger than yourself,
1:04:10 belonging is arguably like the most pleasurable thing,
1:04:11 having friendships, you know, you do all these,
1:04:13 look at all these studies of people that age
1:04:15 and what was meaningful to them in leading a good life
1:04:18 and being ostracized or feeling rejected
1:04:20 or left as like the most painful thing.
1:04:23 So I think that that’s a timeless human truth.
1:04:26 And I certainly feel like,
1:04:28 and I encourage this with my kids, by the way,
1:04:31 I don’t want them just to have a group of friends in school,
1:04:34 because just like the diversity that you need in a portfolio,
1:04:35 you need to have hedges and all these other things
1:04:37 because maybe something is not going right
1:04:37 in that friend group.
1:04:41 And then you are more at risk of catastrophizing
1:04:43 that, oh my God, like nobody likes me.
1:04:47 And you know, and so if you’re in a soccer team
1:04:51 and you go to a religious group or you’re in Hebrew school
1:04:52 or you’re on a dance team
1:04:54 and you have a neighborhood group of friends
1:04:56 and my kids are in like six different things
1:04:58 outside of school each
1:05:00 and they then can bring these people together,
1:05:01 which itself adds this feeling of like,
1:05:04 oh, I connected people and I’m this node
1:05:08 and it sort of cements the network in a way
1:05:12 that is profound and meaningful and comforting.
1:05:13 I like that.
1:05:15 I want to come to aging a little bit here.
1:05:17 I don’t want to do it.
1:05:18 I don’t want to age.
1:05:19 – Talk to me about this
1:05:23 because I feel like based on the research happening now,
1:05:25 the amount of money going into this,
1:05:27 the progress that we seem to be making
1:05:31 at least on biomarkers that we understand,
1:05:33 we seem to be able to dramatically slow
1:05:36 at best our aging process.
1:05:37 You think we’re going to make a quantum leap
1:05:40 and in term of average lifespan for humans,
1:05:44 maybe adding 10 or 15 or 20 good years in the next 10 years?
1:05:48 – I think that’s possible, 10 or 15 or 20, not doubling.
1:05:51 I mean, we did that, you know, in a few generations,
1:05:53 you know, people would die at 40 years old or 50 years old
1:05:56 and people now regularly until their 70s or 80s.
1:05:57 You know, this is an interesting thing
1:06:01 because we don’t fund longevity work here
1:06:05 and I’m personally not invested in any of these things.
1:06:10 I will go to my doctor and I will get my blood tests
1:06:13 and he will suggest that I take some supplements
1:06:14 because I’m low in iron or this or that
1:06:16 and that’s entirely reasonable
1:06:19 just to maintain sort of homeostatic function.
1:06:22 But I don’t go absolutely crazy and intense,
1:06:24 but I appreciate the people, Brian Johnson and others
1:06:27 that are self-hacking and doing this in pursuit
1:06:29 and Ray Kurzweil was doing it back in the day.
1:06:32 Nobody has really seen Ray out that much, you know?
1:06:34 He was taking like, I don’t know, 100 supplements a day
1:06:35 or something like that.
1:06:38 And last I saw, I think he had a two-pay
1:06:40 and like, it was like a messy situation,
1:06:43 but I’m glad that these people are doing it
1:06:46 both in pursuit of staving off their own mortality
1:06:49 and a public service either interpreted as
1:06:50 I can’t believe you go to these extremes
1:06:52 and I’m not gonna do that ’cause it’s super stressful
1:06:54 or maybe you’re going to unearth something
1:06:57 and we’re all gonna be on metformin and all these,
1:07:00 but there are this timeless pursuit of avoiding death.
1:07:03 It’s very human.
1:07:07 It is and it goes back like the first form
1:07:09 of avoiding death was I’m not gonna die.
1:07:12 So the search for the fountain of youth and Ponce de Leon,
1:07:15 today modern pharmaceuticals and drugs and supplements
1:07:16 and lifestyle changes.
1:07:20 The second form was fine, I’m gonna die,
1:07:22 but I’m gonna come back.
1:07:25 And so reincarnation and maybe that was
1:07:26 the spiritual or religious sense.
1:07:29 And then today’s version of that would be Alcor
1:07:32 or any of these cryo, I’m gonna freeze my brain
1:07:35 so that when they figure this out, they’ll bring me back.
1:07:37 The third was, okay, fine, I’m gonna die,
1:07:39 but I am more than my physical self.
1:07:41 There’s an ethereal soul.
1:07:42 The modern technological version
1:07:44 would be the ghost in the machine,
1:07:46 endless sci-fi movies about people uploading themselves
1:07:49 and their likeness, which by the way,
1:07:51 are sort of a really interesting phenomenon
1:07:54 and how we deal with loss and the ability,
1:07:57 if there is an AI that is totally trained on my voice
1:07:59 and my likeness and everything I’ve ever said,
1:08:03 which I have done, that my kids would actually have a dad AI
1:08:05 and maybe they can consult it for questions
1:08:07 or is that a good thing or a bad thing?
1:08:10 I don’t know, but it is going to be a thing.
1:08:12 Then you have, okay, I’m gonna die,
1:08:13 but I’m gonna live on through my progeny,
1:08:15 through my children, through my genes,
1:08:17 which is the evolutionary impetus.
1:08:20 And I’m gonna live on through my works
1:08:23 and you won’t be there to experience either of those things
1:08:25 unlike the first three where you’re not gonna die
1:08:26 or you’re gonna come back.
1:08:28 And so I think about the people that,
1:08:30 whether it’s where I grew up in Coney Island, Brooklyn,
1:08:31 they put graffiti on the wall
1:08:33 and they put themselves up until it gets washed away
1:08:36 or if you’re Dave Rubenstein or Steve Schwartzman,
1:08:37 you put your graffiti etched in stone
1:08:39 on the New York public library
1:08:40 or but it’s really no different,
1:08:42 just $100 million instead of free
1:08:44 and the potential for being jailed.
1:08:47 And then it’s through your children.
1:08:49 And I think there it’s very Buffett-like,
1:08:54 the moral sort of mandate would be like he described
1:08:57 Don Keough of Coke that when you do die,
1:08:59 you want them to say what they said about him,
1:09:01 which was everybody loved him.
1:09:03 I don’t have that, like not everybody loves me,
1:09:05 but my kids is most important.
1:09:07 – I think the theory was the people
1:09:09 that you want to love you, love you.
1:09:10 – Which is interesting because the people
1:09:13 that we celebrate the most, if you think about Steve Jobs
1:09:16 and even Elon, like are the people that are closest to him.
1:09:19 I don’t mean millions of fans that don’t actually know him,
1:09:20 but I used to have this debate
1:09:22 with one of my best friends about Steve Jobs,
1:09:26 like the world loves Steve Jobs, but like–
1:09:29 – People in his orbit didn’t always love Steve Jobs.
1:09:32 – He’s like terrible, you know, like he’s so mean or like,
1:09:33 so I think that’s really interesting.
1:09:36 But going back, the common thing amongst all the people
1:09:39 that have tried to defeat death,
1:09:40 from the people that weren’t gonna die
1:09:42 through found of youth or modern biotech,
1:09:43 the people that were gonna upload themselves,
1:09:45 the people that were gonna come back,
1:09:48 the people that leave it to their kids or to their works,
1:09:51 the common thing amongst all them is that they’re dead.
1:09:53 You know, nobody has beaten death.
1:09:57 And so the mental model that I like on this is,
1:09:58 okay, take a piece of paper
1:10:00 and put the day you were born on the front
1:10:01 and then the day you’re gonna die,
1:10:03 roughly plus 80 years on the back.
1:10:06 And the only thing that you may have control over in part
1:10:09 is the story that you write between these two pages.
1:10:11 And my brother-in-law passed when he was early 40s,
1:10:14 stomach cancer, you know, he lived a tragic short-term,
1:10:16 you know, relative to others.
1:10:18 And, but maybe you get to live this epic tone
1:10:19 and it’s like, how do you write it?
1:10:20 And who do you spend your time with?
1:10:22 And nobody’s gonna look back and say, you know,
1:10:24 I wish I would have taken that extra meeting
1:10:26 or done that extra business trip or something.
1:10:28 It’s like, I’m really glad that I was there
1:10:30 for my kids or my spouse.
1:10:32 And yet you work really hard.
1:10:35 Yeah, and I always prioritize my kids.
1:10:38 Like, I think about all the time, like their judgment,
1:10:40 you know, some people are like, meet their maker
1:10:41 or they, you know, because I don’t believe,
1:10:43 I care deeply.
1:10:46 Well, they say my dad was there for everything.
1:10:47 And in part for me,
1:10:49 because my father was not present in my life,
1:10:51 my parents split when I was super young.
1:10:53 It’s my little guy who’s, I have two daughters and a son,
1:10:55 but my son, who’s nine,
1:10:58 always wants me to have a play date with my dad, you know?
1:11:01 And we’re, we’re civil and we speak a few times a year.
1:11:03 But I’m like, it’s just not that relationship.
1:11:04 I’m not gonna have that, you know?
1:11:06 And he’s like, yeah, but I really want you guys to,
1:11:09 and I’m like, no, I get to be the dad that I am to you
1:11:11 because I didn’t really have that.
1:11:12 And I’m making up for it now.
1:11:14 And this is what it is.
1:11:18 And then I worry that if they see such a present father,
1:11:19 do they take that for granted?
1:11:20 Oh, dude.
1:11:21 And then do they screw it up in the next generation?
1:11:22 You know, and so.
1:11:23 I have the same thoughts.
1:11:25 I actually talked to a therapist who had this
1:11:27 because I was like, am I too present in their lives?
1:11:29 Like they need some space, you know?
1:11:30 I’m home and they get home from after school.
1:11:34 And my wife and I talk about this like my parents split.
1:11:37 My father was married four times.
1:11:40 All I wanted was a stable nuclear family.
1:11:42 But if my kids grow up in a stable nuclear family,
1:11:42 do they take it for granted?
1:11:44 And does like one of them become a cheater
1:11:46 and infidelitus and all this kind of stuff because they,
1:11:48 and I have no idea.
1:11:50 But, but I know for me what is meaningful
1:11:51 and what makes me feel good.
1:11:55 And it’s a totally selfish thing that is about solving
1:11:55 for what I want.
1:11:58 Selflessly, I think it ends up being virtuous.
1:12:01 So what would your top like three or four priorities be then
1:12:03 if you were to outline them?
1:12:06 My kids and my wife call it family number one.
1:12:07 I mean, you know, you think about like the people
1:12:09 that have lost everything physically and materialistically
1:12:12 in the fires recently in LA.
1:12:14 Like family is like, you know,
1:12:15 and I don’t care over time as they grow up
1:12:17 and where they are and they’re, you know,
1:12:19 but that to me is like the most important thing.
1:12:22 And number two is purpose and meaning.
1:12:23 I think this is a universal thing,
1:12:26 but I feel lucky that I enjoy what I do.
1:12:28 It’s an intellectual puzzle.
1:12:32 There’s times of like great fierce competition.
1:12:34 We’re losing to, you know, I don’t like to lose.
1:12:38 We’re losing out to another firm that’s got an entrepreneur.
1:12:40 I like the intellectual gratification
1:12:41 of being right when other people are wrong.
1:12:43 I’m very intellectually competitive that way
1:12:46 and discovering something that people haven’t discovered.
1:12:49 I always talk about Linus Pauling, the double Nobel laureate
1:12:51 who wanted for both chemistry and peace.
1:12:54 And he has this quote about science,
1:12:56 which I just absolutely love.
1:13:00 And this is like, I hope until I’m 90 or 100 or 110
1:13:02 or whatever modern science lets me live till
1:13:04 that I continue to have this
1:13:05 ’cause it’s an addictive feeling,
1:13:08 which is that I know something that nobody else knows
1:13:11 and they won’t know until I tell them.
1:13:14 And I love that discovering a legal secret
1:13:17 and knowing that this is going to be announced
1:13:18 at the scientific breakthroughs coming
1:13:19 and nobody else knows about it yet.
1:13:21 So that to me is like meaning and purpose.
1:13:24 It’s intellectually competitive
1:13:27 and I understand the intellectual competitiveness.
1:13:27 I wanna be right.
1:13:29 I wanna make money.
1:13:31 I want the credit for it,
1:13:32 but it really is about like the status
1:13:33 ’cause otherwise you would just do these things
1:13:34 in private and totally in quiet.
1:13:37 But I like that feeling of that,
1:13:42 even if it’s being glorious and ego and all vanity.
1:13:43 So family, that sense of purpose
1:13:46 driven by this intellectual competitiveness.
1:13:49 And then I think I didn’t appreciate this as much,
1:13:51 but it’s sort of adjacent to the first.
1:13:54 There’s a handful of people who I imagine myself
1:13:56 like retiring with or like guy friends
1:13:58 and people that I enjoy spending time with
1:14:00 that have the same sort of values
1:14:02 and they’re very family driven.
1:14:05 And so my cousin in particular,
1:14:07 this amazing guy, Jason Redless,
1:14:09 one of my wife’s best friends,
1:14:10 this woman, Molly Carmel.
1:14:12 They’re both, we call it family.
1:14:13 They’re like friends and family.
1:14:15 But yeah, it’s a powerful thing
1:14:16 that I don’t wanna ever lose.
1:14:18 – That’s awesome.
1:14:20 You process a ton of information.
1:14:24 What’s that workflow like for information to get to you?
1:14:27 How are you using technology to filter?
1:14:29 How are you filtering information?
1:14:31 – So I typically go to bed between 12 and 12 30,
1:14:36 wake up around seven kids before they leave for school,
1:14:38 about 40 minutes.
1:14:39 I do a lot of physical activity,
1:14:42 usually three days a week between working out,
1:14:44 trainer, jujitsu, all kinds of interesting stuff.
1:14:47 But probably about an hour to an hour and a half
1:14:50 in the mornings of reading through
1:14:53 something like 40 different papers now.
1:14:54 It used to be like seven,
1:14:56 but then when I start to travel internationally,
1:14:57 if I went to the mid-east,
1:15:00 I would find an English version of some of the key papers,
1:15:01 same thing in Japan and elsewhere.
1:15:03 And so now I read a lot of international papers.
1:15:07 And when I say read, I use an app called PressReader.
1:15:09 It has the digital replica of the specific version,
1:15:10 which I really value.
1:15:11 And I know we’ve talked about this in the past,
1:15:14 but I like to know what the editor put on C22.
1:15:17 That’s not as visible on the website
1:15:20 because there’s meta information that the editor is saying,
1:15:21 this is not important to be on the front page,
1:15:24 but if I disagree and I think that there’s a magnitude
1:15:25 of informational importance,
1:15:28 that to me is like some sort of edge.
1:15:30 Then I will take screenshots of those.
1:15:34 And so I will sort of call it scout and scour
1:15:35 through all these papers,
1:15:37 take screenshots in some cases,
1:15:41 I may even take those screenshots and put them into an AI.
1:15:43 And basically say, give me a summary of this article
1:15:46 or give me the three key quotes that really matter.
1:15:50 And so I’ll go down all kinds of like rabbit holes with that.
1:15:50 So that’s the first,
1:15:54 which is just like call it 24 hours worth of information
1:15:56 that is basically put into editorial decree.
1:15:59 You can usually get the FT in New York time
1:16:01 like four, five, six PM Eastern time.
1:16:03 And so you have a little bit of information edge
1:16:05 because most people don’t get the FT
1:16:06 for another 12 or 14 hours,
1:16:08 but they don’t know that they could get it online,
1:16:10 but that is valuable.
1:16:11 And I care about all those things,
1:16:14 including not the sophisticated newspapers,
1:16:16 but like the less sophisticated ones like USA Today,
1:16:18 and I wanna know what is the average person going to read
1:16:19 when they wake up in a Marriott
1:16:20 and get the paper delivered under their door
1:16:22 and that kind of stuff.
1:16:25 Then Twitter, I have all kinds of lists that I follow.
1:16:25 And at any given time,
1:16:28 it might be something that is geopolitical
1:16:29 and war related where I’m going down to deep hole
1:16:31 or sometimes it’s AI and technology,
1:16:32 sometimes it’s my team
1:16:34 and what they’re posting and reading about,
1:16:35 but a lot on Twitter,
1:16:37 which I find truly invaluable.
1:16:38 I mean, I know a lot of people say
1:16:40 it’s like this dark cesspool of whatever,
1:16:42 but you can just filter through and cut out the people.
1:16:44 I’m muting and blocking people all the time.
1:16:46 And I’m discovering all kinds of
1:16:47 just absolutely incredible people.
1:16:49 It has been, as you know,
1:16:50 from one of our first conversations,
1:16:52 like this idea of randomness and optionality,
1:16:54 it is this huge randomness generator,
1:16:56 this huge optionality generator and the accessibility.
1:16:58 And I just, I absolutely love it.
1:17:01 So that’s another thing where I’m really rooting for Elon
1:17:03 in the continued success of X
1:17:05 because I’ll continue to pay a lot of money for it.
1:17:07 And I pay more money for it,
1:17:10 but I find it super valuable and it’s real time pulse.
1:17:12 And I’m excited for GROC to continue
1:17:15 ’cause I think that GROC and X just continue to sort of,
1:17:16 I mean, that’s a repository.
1:17:18 We talked about that before, repository of information.
1:17:21 – One of the first things Elon did was cut off access
1:17:23 to Google from the data. – Totally.
1:17:25 And I think that’s the right move, right?
1:17:28 This is our platform, same way as Meta has.
1:17:31 So I think that that repository of everybody’s tweets
1:17:33 and retweets and likes and the comments that they’ve made,
1:17:34 you can already go on and do this
1:17:37 in sort of a relatively superficial way
1:17:39 where you can say roast me.
1:17:40 And it will basically off your past,
1:17:43 I don’t know, like 25 or 30 tweets sort of roast you
1:17:45 based on what you’ve tweeted about,
1:17:47 but they’re a longitudinal access of people
1:17:50 that have 10,000, 100,000 tweets
1:17:53 is an amazing pastiche of like, you know what,
1:17:55 there’s an interesting thing here
1:17:58 which we were just riffing on internally,
1:17:59 which I’ll come back to.
1:18:00 Just remind me on sort of wrapping yourself
1:18:04 in this information mosaic and breaking free from it.
1:18:07 But papers in the morning, Twitter, internal Slack,
1:18:11 emails, texts, you know, just like processing
1:18:15 all this information, I use rewind on my Mac,
1:18:18 which is effectively doing nonstop screen capture.
1:18:20 And there will be other tools like this in part
1:18:21 because I do not remember the source
1:18:22 when I saw the information.
1:18:24 It’s sort of the same thing of like,
1:18:26 if you see a show and you were to ask me today,
1:18:28 like where did you, I don’t know, it was on Apple TV,
1:18:29 like was it on Paramount?
1:18:30 Was it on CBS?
1:18:31 Was it on Netflix?
1:18:33 Like I have no idea, right?
1:18:35 And in fact, usually when I do the Apple search
1:18:37 and it doesn’t show up, it means it’s on Netflix
1:18:39 ’cause it can search Netflix, right?
1:18:42 Yeah, just huge information omnivore, everything.
1:18:46 And then there’s some random writer that I follow
1:18:48 that’s like Catherine Schultz at the New Yorker,
1:18:51 Adam Gopnik and people whose style of writing
1:18:53 and the selection of their subjects.
1:18:54 I find really interesting
1:18:56 and then I’ll go deep into some of their themes.
1:18:58 – So you use rewind, press reader.
1:19:00 What are the other like technological tools
1:19:01 that you’re finding super valuable?
1:19:05 – Every AI, I might take an essay, read it,
1:19:08 ask it to summarize the key points,
1:19:10 ask it to put it in different voices,
1:19:12 take two different essays and say,
1:19:13 where do these things agree or disagree?
1:19:16 And so yeah, like just nonstop.
1:19:21 I’m on AI easily more than Google now,
1:19:23 but I don’t know, two, three hours a day.
1:19:24 – What have you learned about prompting
1:19:27 that would help everybody get better results?
1:19:31 – Usually very specific, like I give it a priming thing,
1:19:34 like you are the world’s, it’s a neuroscience paper.
1:19:36 You are the world’s greatest expert in neuroscience.
1:19:39 You have read every paper that has been published.
1:19:43 You have both a skeptical eye to new claims,
1:19:47 but you are also open-minded to interesting correlations
1:19:48 that might not have been considered.
1:19:53 Read this paper and give me the three most provocative,
1:19:56 non-obvious points and give me the three cliches.
1:19:57 And so just, and by the way,
1:20:01 I will put them into three different models at the same time.
1:20:03 So I will open three different browsers,
1:20:05 arrange them and put it into chat GPT,
1:20:06 put it into Claude, put it into one of the perplexing models
1:20:08 that’s not running on those two.
1:20:11 And sometimes I’ll mix and match them.
1:20:12 – I love that.
1:20:15 – It’s sort of like a pallet of mixing.
1:20:18 We have not yet done this as a partnership,
1:20:21 but we’ve talked about it, having an AI partner.
1:20:25 There’s still a behavioral discomfort
1:20:26 about recording conversations.
1:20:28 You and I are recording our conversation now.
1:20:30 But every partnership discussion we have,
1:20:34 we were confident that it was protected and encrypted
1:20:37 because we might say things that could be harmful.
1:20:38 – You don’t want them coming at.
1:20:40 – We could insult somebody or like,
1:20:42 or we have a piece of intel that we don’t want out.
1:20:45 But if we were comfortable that it was perfectly private,
1:20:48 which is a hard thing to promise, but if it was,
1:20:50 you would have a repository of every conversation
1:20:53 we’ve ever had over the past X number of years,
1:20:55 the decisions that we wrestled with,
1:20:57 you would be able to have somebody to advise us,
1:20:59 an AI to advise us,
1:21:02 where are we showing biases in consistency
1:21:04 between a decision we made three months ago and this?
1:21:06 What is different this time?
1:21:08 Which voices are not speaking up?
1:21:09 And you can already get this in some cases
1:21:11 with like certain Zoom calls or other recording things
1:21:14 where it’ll tell you who spoke for how long.
1:21:18 And then you could run like a Bayesian analysis of,
1:21:20 okay, given that we’re looking at these two companies,
1:21:21 give me the outside view,
1:21:24 the base rate of success historically,
1:21:25 which in venture honestly doesn’t matter,
1:21:27 but and then give me a Kelly criterion
1:21:28 of how you might size this
1:21:30 based on the projected internal confidence.
1:21:31 And so there’s all kinds of things
1:21:34 that we could internally do to use these tools,
1:21:37 which I think over time we’ll probably experiment with.
1:21:40 But the biggest thing is basically having like
1:21:41 a capture of everything, you know,
1:21:43 everything that you see, everything that you hear,
1:21:46 everything that my, I’ve already given over again
1:21:48 to the privacy gods, everything that my screen sees.
1:21:51 And so I trust that that siloed on my device,
1:21:52 it’s not going to the cloud,
1:21:53 but it’s super helpful when I’m trying to search
1:21:55 for something I’m like, was that a Gmail?
1:21:56 Was that a text?
1:21:57 Was that a thing on Twitter?
1:21:58 Was that a PDF I read?
1:21:59 Where did I see that?
1:22:03 And the ability to DVR my life is super valuable.
1:22:04 If I could do that with my conversations,
1:22:06 like who said that the other day?
1:22:08 In fact, Lauren and I just had somebody over,
1:22:10 you know, we host people at our house
1:22:12 and we couldn’t remember who told us this thing.
1:22:14 And we were like, I had to go through my calendar
1:22:15 to see who was over on Thursday or Friday.
1:22:17 Oh yeah, okay, it was, you know,
1:22:20 but being able to search your life instantly,
1:22:23 I think it’s going to be a generational change
1:22:25 in the same way that people were not comfortable,
1:22:26 you know, posting on Facebook.
1:22:28 And then they were comfortable.
1:22:30 And then like now people are like posting themselves
1:22:34 in swimsuits and bikinis and it just doesn’t matter.
1:22:36 That to me is going to be a big step change.
1:22:38 I want to come back to the info mosaic,
1:22:40 but one thing we never talked about YouTube
1:22:43 being like such a huge data source.
1:22:43 Incredible.
1:22:45 Closed and slightly open, I guess.
1:22:46 Yeah.
1:22:47 In some ways.
1:22:49 Yeah, like, well, I love that moment
1:22:52 when I think it was Mira from OpenAI was asked,
1:22:54 like, you know, so did you train?
1:22:56 And she was like, but you didn’t want to answer.
1:22:57 Crickets.
1:22:58 Yeah.
1:23:02 Okay, the info mosaic and breaking free from sort of like,
1:23:04 one thing I do love about acts is that
1:23:07 it shows you views that are contrary to your own,
1:23:09 like the algorithms gotten pretty good at.
1:23:10 Yes.
1:23:12 And there are, what is it, ground news
1:23:13 that you can sort of do this
1:23:16 where it will actually give you a bias on, you know,
1:23:18 certain things and it’ll give you both sides of the view.
1:23:19 So if you truly are objective
1:23:21 and like truly knowledge seeking,
1:23:23 then you would want to experience that.
1:23:24 And I feel like that will be an option
1:23:26 that you just click and enable feature
1:23:28 and you know, it’s able to identify some of the biases
1:23:29 and whatnot.
1:23:30 This idea of the information mosaic
1:23:33 was a recent conversation I was having with my colleague,
1:23:34 Danny Crichton, who runs like our risk gaming stuff
1:23:36 where we’re coming up with all kinds of crazy scenarios
1:23:39 and imagining these low probability, high magnitude events.
1:23:42 And the idea was that over time,
1:23:46 this perfect simulacrum of Shane or of Josh
1:23:47 is going to exist.
1:23:49 Everything that I’ve ever said on every podcast,
1:23:50 everything I’ve ever written publicly,
1:23:51 forgetting all my private thoughts,
1:23:53 but just everything that I’m out there publicly,
1:23:54 my voice, my tone, okay.
1:23:58 And so I almost imagined it like this matrix like mosaic,
1:24:01 like a Spider-Man costume that’s like form fitting.
1:24:04 It’s me or a close approximation of me.
1:24:06 But what if you want to break free from that?
1:24:08 In a sense, if I said,
1:24:10 give me something in the style of Shane Parish,
1:24:11 it might conjure something in the style of Shane Parish,
1:24:13 but in the style of Josh Wolf
1:24:17 or in the style of David Milch or Christopher Hitchens,
1:24:20 you know, I actually love invoking dead voices,
1:24:22 you know, to sort of bring them back from the dead, right?
1:24:24 And have them opine on the topic.
1:24:25 What would Christopher Hitchens say
1:24:27 about this article, blah, blah, blah.
1:24:30 But what if I wanted to break free stylistically?
1:24:35 If I said, give me a image of a horse in Tribeca
1:24:37 in the style of Wes Anderson.
1:24:40 You know, I can imagine the pastel palettes
1:24:42 that it would conjure and you could imagine that too
1:24:45 with the, you know, rectilinear framing and whatever.
1:24:48 But what if Wes Anderson suddenly had like a new
1:24:51 stylistic change in his over and wanted to just shift?
1:24:53 Like he’d be constrained, you know,
1:24:56 in the same way that people hate when, you know,
1:24:57 I don’t know, maybe when Dylan went electric
1:25:00 or like, you know, somebody else changes their style
1:25:01 or their genre.
1:25:05 And so there’s this aspect where AI constrains you.
1:25:10 And they’re just sort of playing with this idea of,
1:25:12 you know, how do you break free in the same way
1:25:15 that there might be like the right to be forgotten
1:25:16 that maybe you want to change your style.
1:25:19 The great virtue of college for most people
1:25:22 is this quartet of years where you can break free
1:25:25 from who you were for the past four years.
1:25:27 And nobody knows who you were and what you cared about.
1:25:31 And maybe you were into heavy metal,
1:25:34 but you were in like the band, you know,
1:25:35 and you couldn’t break free or maybe you were gay
1:25:38 and nobody knew or all these things that you can just
1:25:41 suddenly like be yourself and explore new things.
1:25:45 And there’s this element where the great virtue of college
1:25:48 is self-expiration against the constraints of high school,
1:25:51 but could AI be this constraining force?
1:25:53 Because the more content that you put into it,
1:25:55 the more it knows you,
1:25:59 the more you may have trouble varying from it.
1:26:01 And so there’s something interesting there.
1:26:02 – I like that a lot.
1:26:05 Let’s talk military and technology
1:26:08 and you guys are big investors in Antrol.
1:26:10 Where’s that going in the future?
1:26:13 – Well, there’s gonna be a lot more brilliant minds,
1:26:16 I think that feel comfortable, motivated,
1:26:20 not only by a sense of purpose, patriotism,
1:26:23 but also principle and capital making that they see,
1:26:25 the things that they doubted early on,
1:26:26 like why is this time different
1:26:29 in another defense company of which there weren’t very many,
1:26:32 but seeing Andrew’s ascendancy and valuation
1:26:34 and success and program wins,
1:26:35 I think has inspired a lot of people
1:26:37 like wait, there’s something going on here.
1:26:39 We went from 50 primes down to five,
1:26:42 you’re seeing the rise of these neoprimes.
1:26:44 I deeply believe that Antrol in the next few years
1:26:48 will be a $30 to $50 billion publicly traded business,
1:26:52 doing single digit mid billions of revenue
1:26:53 with software like margins
1:26:55 that are not like these cost plus margins.
1:26:57 So that is gonna usher in a big wave
1:26:59 and they’re buying companies,
1:27:00 they’re acquiring smaller businesses,
1:27:03 but you’ll continue to see that sort of evolution
1:27:05 in a world that people realize
1:27:07 is not kumbaya peace and safety.
1:27:12 There are bad actors that when we take a step back
1:27:15 or on our back foot or a little bit permissive
1:27:18 that they arm up, it happened with Iran.
1:27:21 And I think the prior administrations from Obama and Biden
1:27:24 were well-intentioned in trying to bring them
1:27:28 into the Western world, but it was a sort of ruse
1:27:29 from Iran standpoint.
1:27:33 Same thing with Gaza and Israel and Russia
1:27:35 and who thought that we were gonna see a land war
1:27:38 in the 21st century where Russia would invade Ukraine
1:27:43 and China and Taiwan and North Korea
1:27:45 and the African continent,
1:27:47 if we talked about in the Sahel Maghreb,
1:27:49 infiltration of a lot of these groups into South America.
1:27:52 I mean, there’s just lots of conflict waiting
1:27:57 and the best way to avoid conflict is to have deterrence.
1:28:00 And if Ukraine had nuclear weapons,
1:28:01 Putin wouldn’t have invaded.
1:28:03 Most of the West and NATO really said,
1:28:04 “Don’t worry, we got your back.”
1:28:06 And even though you’re not part of NATO
1:28:08 and they never nuclearized,
1:28:13 I think the world, timelessly through all of human history,
1:28:18 is gonna face enormous conflict, resource wars,
1:28:20 water may be next.
1:28:24 I think there’s something like 1,900 active conflicts
1:28:26 around the world, around water rights.
1:28:28 You look at China and Pakistan, control of the water.
1:28:32 I mean, there’s just like a lot of resources.
1:28:36 You look at disrupting undersea cables,
1:28:38 sabotage efforts.
1:28:40 You look at deep sea mining.
1:28:42 You look at space as another frontier.
1:28:46 There’s just a lot of opportunities for zero-sum conflict.
1:28:49 And when you can’t reconcile those conflicts
1:28:52 through diplomacy or negotiations or agreement,
1:28:54 it goes to violence.
1:28:58 And the people that can bring or effect or export violence
1:28:59 typically have the upper hand.
1:29:03 And part of what has made this country great
1:29:05 and made it powerful and made it the economic juggernaut
1:29:10 is that it is allowing for the low entropy system,
1:29:14 even though the country at times seems chaotic,
1:29:16 that allows for the high entropy production
1:29:20 of entrepreneurial ideas and free market capitalism
1:29:23 and booms and busts is that we have
1:29:25 the most powerful military on the planet.
1:29:28 You could argue that that didn’t just benefit the United States.
1:29:31 It benefited Canada, Europe, it benefited a lot of–
1:29:32 Mexico are allies for sure.
1:29:34 You can watch many fictional movies
1:29:36 that have run these counterfactuals
1:29:38 of what would have happened if Nazi Germany had won
1:29:43 or the Russians had landed for all mankind
1:29:45 before we did in the moon.
1:29:48 But we’re getting away from an era
1:29:53 of like, here’s a trillion dollar boat effectively.
1:29:56 It depends who you talk to, if shouldn’t we?
1:29:58 Like, I mean, if that boat can be taken out
1:30:01 by a $3,000 drone, how effective is it?
1:30:03 Yes, for sure, the asymmetry of a threat
1:30:08 of an aircraft carrier against a large fleet of drones.
1:30:10 It is very much, if you talk to Sam Paparo,
1:30:11 who’s the head of Indo-Pakarm,
1:30:13 he will say it is all about mass on target.
1:30:18 There’s certain things that automation cannot do.
1:30:19 And he wants what he calls,
1:30:20 which I guess is a technical term,
1:30:24 a hellscape in that region, the Taiwan Straits
1:30:29 and South China Sea, so that you make it really impossible
1:30:32 for them to have any military dominance.
1:30:36 But it is an era where it’s about,
1:30:38 you saw this again with Iran and Israel and Gaza
1:30:42 and Syria’s missiles and counter missiles
1:30:44 and rockets and intercontinental ballistic missiles
1:30:47 and hypersonics and space weapons.
1:30:50 It is just about going back to almost like
1:30:53 Planet of the Apes, one ape through another,
1:30:56 a rocket another or a twig or a stroller.
1:30:58 The weapons get more powerful,
1:30:59 but the behavior doesn’t change.
1:31:01 We’re back to throwing projectiles at each other,
1:31:03 and it’s just they’re automated,
1:31:07 they’re at speeds or at levels of
1:31:10 a treatable, overwhelming defensive forces
1:31:12 that that is the battlefield.
1:31:15 Do you think values become a disadvantage in some ways then?
1:31:18 Like for example, if the United States were,
1:31:21 we need a human operator to pull the trigger
1:31:22 and another country was no,
1:31:24 it could be completely automated.
1:31:24 For sure.
1:31:26 And therefore in a dogfight,
1:31:28 we’re likely to more win.
1:31:31 Look, this is already happening in the information space
1:31:35 where we have certain and in the autonomous space.
1:31:40 I was in the Pacific region with SOCOM,
1:31:43 and there’s a drone operator who’s flying the drone.
1:31:44 There’s another drone operator
1:31:48 who’s piloting the weapon system,
1:31:49 and there’s two lawyers.
1:31:51 So they’re helping the commander
1:31:54 who is effectively given like a God shot of,
1:31:57 how many combatants and civilians can be killed
1:32:00 and what ratio, and sometimes it’s like five to one
1:32:03 or 10 to one, but there’s lawyers that can authorize
1:32:06 because we have a certain rule of engagement
1:32:09 that frankly gives these military personnel
1:32:13 the ethical comfort that this is a superior system.
1:32:15 But for sure, if there are people that don’t have
1:32:17 that same moral code, in some cases,
1:32:20 they can be at least temporarily advantaged.
1:32:22 – Well, you can think of that through AI too,
1:32:23 not just military, right?
1:32:26 If we restrict, we put restrictions on any technology
1:32:28 and another country doesn’t,
1:32:31 sometimes that can cause an advantage to another country.
1:32:34 – China has the 50 cent army.
1:32:37 These people are getting 50 cents for every tweet
1:32:38 and information they put out.
1:32:40 The State Department, when they want to tweet something out
1:32:43 through groups, there’s literally like a disclaimer
1:32:45 that says, and it’s like one woman in Tampa
1:32:48 that’s doing this, like this was sponsored by the state.
1:32:50 So we have these ethical restrictions,
1:32:54 which definitely tie our hands behind our back
1:32:55 in some cases.
1:32:57 And our enemies will always try to weaponize this.
1:32:59 So, I mean, you can look at many vectors today
1:33:01 that don’t seem like they’re threat vectors,
1:33:03 but they have been weaponized.
1:33:05 Social media information, we know.
1:33:09 And the best fix for that is identifying the bad actors
1:33:11 and also inoculating people with a heightened degree
1:33:13 of skepticism, but the vast majority
1:33:15 of the American population will not be inoculated.
1:33:17 They will see the things that they want to see.
1:33:19 They will follow the accounts that they want to follow.
1:33:22 And then occasionally those accounts will start to pepper
1:33:24 in other information that they want people to believe.
1:33:27 And that’s how information cascades can go.
1:33:29 We have open systems, immigration.
1:33:31 You’re seeing a lot of the rise
1:33:34 of the populist anti-immigrant movement in part
1:33:37 because in some cases it’s a result of good intentions
1:33:40 of providing sanctuary cities and wanting to help people
1:33:43 and provide amnesty and help immigrants come here
1:33:44 because that’s what our country was built on.
1:33:46 And then you want those people to assimilate.
1:33:48 And when they’re not assimilating,
1:33:50 but then you also have bad actors like Putin
1:33:52 who has weaponized immigration and put migrants
1:33:54 on people’s borders to create pressures
1:33:56 so that you can get a political movement
1:33:58 from inside the country that will be sympathetic
1:33:59 to the nationalist sensibilities
1:34:02 and he’s orchestrated that very well.
1:34:05 And so infiltration into our university systems,
1:34:08 which accept foreign capital and you see Qatar
1:34:12 that has influenced very massively domestic US universities.
1:34:13 China doesn’t allow that.
1:34:16 US is not able to come in and sponsor Chinese universities.
1:34:18 TikTok, of course, a huge one, right?
1:34:20 We banned it years ago,
1:34:23 right when it had become TikTok for musically,
1:34:24 in part because at the time,
1:34:26 I seemed like a conspiratorial nuts saying,
1:34:28 I don’t trust this with the Chinese Communist Party
1:34:29 having control over this.
1:34:32 And then behind the scenes myself and many others
1:34:35 have played a role in helping to orchestrate what I hope
1:34:38 will not be thwarted by Trump to see this divested.
1:34:41 I have no problem with people using TikTok,
1:34:44 but it should not be in the hands of the algorithm
1:34:46 and control of the Chinese Communist Party.
1:34:48 And it’s on a lot of government phones.
1:34:49 It’s insane.
1:34:50 It’s really interesting.
1:34:52 Why haven’t we seen more isolated attacks
1:34:54 that are cheap and using technology?
1:34:56 And what I mean by that is,
1:34:59 nefarious actor can probably for two or three grand
1:35:03 effectively plot an assassination.
1:35:04 I guess the question would be to what end?
1:35:07 And so you still have to realize that many people,
1:35:09 even if they’re like evil geniuses,
1:35:12 have an objective in mind.
1:35:14 And do they just want to sow chaos
1:35:15 and create distrust in the system
1:35:17 and have people scrambling?
1:35:19 Whereas they’re an opportune time to strike.
1:35:24 Think about the Israel operation with the Beeper plan.
1:35:25 This was 10 years in the making.
1:35:27 Now they could have done it at any point in time
1:35:30 five years ago, but they waited until a precise moment.
1:35:32 And so being able to do the thing
1:35:34 and deciding when you do the thing
1:35:36 are two different decisions.
1:35:39 But I think that we’ve been warned for a very long time
1:35:41 about hacking and infiltration
1:35:44 into our physical infrastructure, for sure.
1:35:46 Somebody could shut down air traffic control.
1:35:47 And what we saw just recently
1:35:49 between the Black Hawk helicopter
1:35:52 and this regional plane from, was it Kansas City?
1:35:54 You know, crash in Washington, DC.
1:35:58 You could see the FAA shutdown and have a glitch.
1:36:01 You can have infiltration into our banking system
1:36:03 and just like the Sony hack, right?
1:36:05 The big thing with the Sony hack
1:36:07 was not that the systems were disabled.
1:36:09 It’s that information was revealed.
1:36:10 You want to create civil war in this country.
1:36:13 Just reveal everybody’s emails for the past year.
1:36:15 The things that we’ve said about each other, you know.
1:36:17 I mean that like reveal truth in a sense, right?
1:36:18 It was the great irony.
1:36:21 So the obfuscation of these things in private
1:36:23 helped to create a civil society.
1:36:26 Our water systems, our infrastructure, our traffic lights,
1:36:27 you know, I mean, all the things that you’ve seen
1:36:31 in sci-fi movies when like things just start breaking,
1:36:36 I’m actually amazed that our infrastructure globally,
1:36:39 but you know, even in New York City, even in this office,
1:36:42 there’s a million skews in this office.
1:36:44 You know, above our heads right now,
1:36:45 there’s an HVAC system.
1:36:46 Like the fact that we trust this system
1:36:49 and it’s not gonna fall and explode or blow up like,
1:36:52 and then we’re shocked when these things do,
1:36:54 but I’m constantly amazed that the entropy,
1:36:57 the forces of entropy are constrained
1:36:59 by either really good engineering
1:37:02 or inspection of systems or whatever it might be,
1:37:03 the maintenance of systems,
1:37:05 which is another really interesting thing.
1:37:07 This idea of maintenance.
1:37:11 Like the past 10 years have been all about growth,
1:37:12 growth, growth, growth.
1:37:15 You go to financial statement on CapEx,
1:37:16 you’ve got growth and you’ve got maintenance.
1:37:19 And I think in a world where rising cost of capital
1:37:21 keeps going up for a variety of reasons.
1:37:22 I think there’s a ton of dry powder
1:37:24 and venture capital and private equity,
1:37:25 but a lot of it I call wet powder
1:37:27 because this money is basically reserved
1:37:29 for companies and people don’t realize.
1:37:32 Reshoring, all of these things, you know,
1:37:33 tariffs, they’re gonna be inflationary,
1:37:35 they’re gonna be a rising cost of capital.
1:37:37 If you have a rising cost of capital,
1:37:40 if you are a CFO or you’re on a board
1:37:41 and you’re thinking about good governance
1:37:43 and capital allocation,
1:37:44 we’re not buying the next new hot thing
1:37:46 unless we really have to like AI today.
1:37:47 – Yeah.
1:37:49 – We’re thinking about how do we maintain
1:37:50 the existing assets we have?
1:37:52 And those assets could be satellites up in space,
1:37:54 they could be military installations,
1:37:56 they could be our telecom infrastructure,
1:37:59 our bridges, our waterways, sanitation, processing,
1:38:01 our HVAC systems, our industrial systems,
1:38:03 all those things that need to be maintained.
1:38:06 So I’m increasingly interested in new technologies.
1:38:08 This could be software, services, sensors,
1:38:12 all kinds of things that can help apply to old systems
1:38:12 to maintain them for longer,
1:38:15 depreciate them for longer, let them last for longer.
1:38:16 I think there’s gonna be increasing demand
1:38:18 for maintenance of systems.
1:38:20 But I’m amazed that everything around us
1:38:21 is just not constantly breaking.
1:38:24 It truly is like miraculous.
1:38:25 – It is when you think about it, right?
1:38:26 Yeah.
1:38:28 What do you think of Doge?
1:38:30 – The currency or the initiative?
1:38:32 – The initiative, no talk crypt.
1:38:36 I think it’s a virtuous thing because
1:38:39 it’s shining a spotlight on a lot of things
1:38:40 that were just done because they were done
1:38:42 and you get this bloat.
1:38:45 Or in some cases there was like overt obfuscation.
1:38:48 So I think sunlight heals all and putting a spotlight
1:38:52 on ridiculous spending or ridiculous inefficient things.
1:38:56 I will say I grew up sort of a center left Democrat
1:38:57 my entire life.
1:38:59 The first time you go to the DMV,
1:39:00 you become a Republican.
1:39:04 Like it’s just like you want systems that have competition
1:39:06 because competition makes things better
1:39:08 because if you have a monopoly on something
1:39:10 you don’t have to improve.
1:39:12 If there’s one regional carrier for an airline,
1:39:14 if there’s one restaurant,
1:39:17 if there’s one place you have to go to for your passport,
1:39:19 you don’t want that sort of centralized control
1:39:20 because the service is going to suck
1:39:22 because they don’t have to do any better.
1:39:26 So I think if you can put a spotlight on excess
1:39:29 and waste and bureaucracy and at least begin
1:39:32 the conversations at a bare minimum of wait,
1:39:34 we’re spending how much money on what?
1:39:36 I think that that’s a virtuous thing.
1:39:39 Whether or not these things will be effective
1:39:42 at really reducing costs, TBD,
1:39:44 but it actually seems quite positive
1:39:46 that they may hit some of their targets
1:39:49 of trying to reach what is it a billion a day or more.
1:39:52 And if that could end up reducing the deficit by 10%
1:39:54 or 20% let alone 50% going from two trillion
1:39:56 to a trillion would be incredible.
1:39:59 So whatever the motivation,
1:40:00 I don’t believe it’s patriotism.
1:40:02 It could be intellectually competitive.
1:40:03 It could be power.
1:40:04 Whatever the motivation,
1:40:07 I think that the means to the end,
1:40:09 I think the end is a virtuous pursuit.
1:40:12 If you were to take over a country effectively
1:40:15 and you were in charge of policies and regulations,
1:40:17 what would you do to attract capital
1:40:21 and become competitive over the next 20 or 30 years?
1:40:22 What sort of things would you implement?
1:40:25 What would you get away from and not do?
1:40:27 Well, I have an adjacent answer
1:40:29 of if I were Secretary of State or Secretary of Defense
1:40:31 for the day, which I’ll give you first,
1:40:34 which is I would really put priority on Africa
1:40:36 as a continent and particularly that Sahel Maghreb
1:40:38 because I do believe between violent extremists,
1:40:41 Russian mercenaries, China infrastructure,
1:40:43 you are one terror event away projected into Europe
1:40:45 that creates the next Afghanistan
1:40:48 and suddenly NATO and the US are in there dealing with ISIS.
1:40:50 You’re already seeing the first authorized strike
1:40:53 by Trump on ISIS in Somalia.
1:40:58 And so you’ve got Sudan, Chad, Mali, Niger,
1:41:00 like it is just a hotbed of people
1:41:02 that were coming from Syria and Afghanistan,
1:41:05 Islamic extremists, it is a bad situation.
1:41:07 And I think that we should be proactive there
1:41:08 before we have to be reactive
1:41:10 and it’s a lot more costly in lives and money
1:41:12 and blood and treasure.
1:41:12 The second thing I would do is a
1:41:15 hemispheric hegemony declaration.
1:41:19 I just went to Nicaragua, Nicaragua for a variety of reasons.
1:41:20 We went instead of Costa Rica,
1:41:22 but I feel much safer in Costa Rica.
1:41:23 And I was worried that I was not gonna be able
1:41:24 to leave Nicaragua and we went with a friend
1:41:26 who happens to be a prominent journalist.
1:41:28 He was not allowed entry into the country.
1:41:31 So it really through our family vacation,
1:41:35 his family of five, my family of five for a wrench
1:41:39 because the government is trying to take over
1:41:42 the banking system and they don’t want it to be covered
1:41:43 by financial journalists and these kinds of things.
1:41:46 And you look at who is in there
1:41:48 and you literally have presence from Hezbollah,
1:41:51 from China, CCP, from Iran.
1:41:52 It’s a bad situation.
1:41:58 The places that we think in most of Central America’s,
1:42:01 Caribbean, South America are vacation spots
1:42:04 where we get our coffee and we go on a nice vacation,
1:42:06 massive infiltration from adversaries.
1:42:09 And so I think we are losing the game
1:42:11 and I would declare almost like a new monorail doctrine
1:42:15 kind of thing where we say the entire Western hemisphere,
1:42:17 you’ve got a billion people,
1:42:21 both ability to project into the Pacific and the Atlantic.
1:42:24 You’ve got mostly English and Spanish speaking people
1:42:25 safer Brazil with Portuguese.
1:42:27 You’ve got a ton of resources,
1:42:30 a ton of brilliant educated doctors and whatnot.
1:42:31 And I would just shore up this hemisphere,
1:42:33 particularly against a crink,
1:42:34 you know, China, Russia, Iran, North Korea
1:42:36 and their influence.
1:42:38 If we were worried about like Cuba
1:42:39 and the Cuba missile crisis
1:42:43 and things proximate 90 miles or what from Florida,
1:42:47 I think that China is doing very smart strategic things.
1:42:48 So us going back in and saying,
1:42:52 we’re gonna reclaim the Panama Canal and our influence on it,
1:42:55 like forgetting about provocations of Mexico,
1:42:58 of like the Gulf of America versus Gulf of Mexico.
1:43:00 But I think that having influence in that region
1:43:01 is really important.
1:43:03 So those would be the two things as sect F or sect state
1:43:06 that I would do is declare a hemispheric hegemony
1:43:07 and make sure that we shore up our allies in the region
1:43:10 and get out our adversaries and their presence in part
1:43:11 because there’s so much commerce
1:43:13 and money and infrastructure that’s going in
1:43:15 and then focus on the Sahel and Maghreb in Africa.
1:43:18 For country competition, brain drain,
1:43:19 you want the best and the brightest.
1:43:21 You need to fund basic research and basic science.
1:43:22 It should be undirected.
1:43:24 That’s the serendipity and the randomness
1:43:26 and the optionality that leads to great breakthroughs.
1:43:29 You want capital markets to be these low entropy carriers
1:43:32 for high entropy entrepreneurial surprise.
1:43:35 So predictable rules and regulations.
1:43:37 I would lower, I don’t know why we don’t have a flat tax.
1:43:38 I mean, I know why we don’t,
1:43:42 but I would just have a flat tax, make it super simple.
1:43:43 Rich people are gonna scout around
1:43:45 and figure out how to get around the tax system anyway
1:43:47 and poor people are burdened by it.
1:43:49 I get progressive versus regressive,
1:43:53 but I would just simplify our tax scheme massively.
1:43:55 Anybody that is coming here
1:43:56 and getting an education in this country,
1:43:58 I would staple a visa as long as they stay here,
1:44:01 work for American company for at least five years,
1:44:03 let them become, we want the best and the brightest here.
1:44:06 We are, as an example again,
1:44:07 and I don’t mean to make this like all China,
1:44:09 but they are our most dominant adversary,
1:44:13 50% of all AI undergrads in the world today
1:44:15 are being graduated by China.
1:44:17 In our own country in the United States,
1:44:22 38% of researchers in AI are from China.
1:44:24 So we’re outnumbered even domestically, it’s a big deal.
1:44:27 And we used to attract something like 23%
1:44:31 of all foreign graduates here, that’s down to 15%.
1:44:32 People are either going to other countries
1:44:33 or staying in their own country.
1:44:37 And so we need that, that’s what won us World War II.
1:44:40 You know, if Einstein would have stayed in Germany or…
1:44:43 – What causes that to happen is that the tax rate,
1:44:45 is it opportunities available?
1:44:46 Is it housing costs?
1:44:49 Like what are the factors that go into people leaving?
1:44:51 – Well, start with the attracting part.
1:44:53 You know, as Walter Riston said,
1:44:54 people go where they’re welcome
1:44:56 and stay where they’re well treated.
1:44:58 So we should be welcoming.
1:44:59 Now there’s a debate about immigration
1:45:03 and we should distinguish between the bad people
1:45:05 and like brilliant people and we should want them here.
1:45:08 – That just comes down to like basic batting.
1:45:10 – Right, but you know, some of that is exploited.
1:45:13 You know, a lot of these consultants with Wipro
1:45:16 and some of the Indian business processes,
1:45:20 BPO is the business processes operations, or outsourcing.
1:45:23 Housing is a difficult one, but people can always figure out,
1:45:24 New York City is expensive,
1:45:26 but you can live in Long Island City
1:45:27 and Brooklyn and Queens.
1:45:33 But I think, yeah, housing availability,
1:45:35 we have, I mean, our cities are so rich
1:45:36 and filled with culture and people.
1:45:37 And particularly if you’re a young person,
1:45:39 you want to be around the density of that
1:45:41 because you’re trying to find peers and a mate
1:45:44 and all of that, even if you’re from another country,
1:45:45 you go to New York, you can find your enclave
1:45:48 of Korean or Chinese or Russian or Ukrainian
1:45:52 or Israeli and Caribbean, it’s all here.
1:45:54 So, yeah, I think just having a culture
1:45:57 that embraces this and encourages it,
1:45:59 you already have a robust venture capital system
1:46:00 of risk taking, many other countries don’t have that.
1:46:02 So that’s another thing.
1:46:05 But if you were to design the system from scratch,
1:46:08 you want openness with security.
1:46:10 So some means of vetting.
1:46:12 You want a great education system that can attract people
1:46:15 that view it as a status to have graduated
1:46:17 from that particular school.
1:46:18 And people want to be around people,
1:46:20 whether this is in a company or in a country
1:46:21 that are like them, that are competitive
1:46:24 and highly intellectual, that they respect or admire,
1:46:25 want to compete with.
1:46:26 So that’s number two.
1:46:29 Their work, something that we did in the 1980s,
1:46:31 I think it was in 1980 was the Bidol Act,
1:46:33 where government funding for research
1:46:36 would allow the university and the principal investigator
1:46:38 to actually own the intellectual property
1:46:39 that became an asset.
1:46:42 That asset could be licensed to a company.
1:46:45 It absolutely opened the floodgates for venture capital
1:46:46 to be able to commercialize that.
1:46:49 That happened to coincide with Orissa
1:46:52 and allowing retirement plans to go into venture capital.
1:46:54 So now you have a pool of risk capital,
1:46:58 which you need for taking risk on unseasoned people
1:46:59 and unseasoned companies.
1:47:02 And then you need a robust capital market system
1:47:04 to be able to continue to allocate money.
1:47:05 But again, capital goes where it’s welcome,
1:47:06 stays where it’s well treated,
1:47:09 true of human capital, true of financial capital.
1:47:12 And so a rules-based system, a strong military,
1:47:13 if you were starting a country from scratch,
1:47:14 you’re not gonna have that.
1:47:17 But you need great allies then, think about Singapore.
1:47:19 Yeah, I think that’s a phenomenal model.
1:47:20 Singapore is a great model.
1:47:21 – That’s awesome.
1:47:23 I hope some of the government people we have
1:47:25 are listening to this.
1:47:27 I wanna come to IP for a second and copyright
1:47:29 and then wrap it up here.
1:47:33 So do you think that AI should be able to create IP
1:47:34 or copyrighted material?
1:47:36 Like if I tell AI to write a book,
1:47:38 is should that be copyrightable?
1:47:41 And who owns the copyright, the AI or me for the prompt?
1:47:44 – It’s super complicated because the first debate
1:47:45 about this, which is the great irony, right?
1:47:48 Because open AI investors and stakeholders were up in arms
1:47:50 that are one stole from open AI.
1:47:51 But you can make the argument that open AI
1:47:53 is trained on the repository of like the public internet
1:47:56 and every art that’s ever been produced and whatnot.
1:47:59 Now, if you were an art student and you went to the Louvre
1:48:03 or to the Met or to MoMA and you sat there and studied it
1:48:04 or took a picture of it.
1:48:07 And then we learned through copying.
1:48:09 We learned through symmetry and imitation
1:48:11 and we remixed these things.
1:48:14 And there’s this great, what is his name?
1:48:17 Kirby who did everything as a remix.
1:48:18 I just sent it to a friend,
1:48:20 but it was like updated last year.
1:48:24 And it’s so brilliant in its compilation
1:48:29 of every facet of culture that you love from books
1:48:33 to your Tarantino movie, to the Beatles, to art,
1:48:34 to scenes and movies.
1:48:37 Like it was all copied from something, you know?
1:48:40 And you’re like, wait a second.
1:48:42 That riff came from this 1940 song
1:48:44 from this African American blues guitarist
1:48:47 that John Leonard or Paul McCartney stole.
1:48:50 And you’re like, and so everything was sort of stolen
1:48:52 from somebody.
1:48:54 It was imitated, tweaked slightly.
1:48:56 And by the way, that’s what we are, right?
1:48:58 I mean, you get two people who exist.
1:49:00 And then there’s this genetic recombination
1:49:02 of their source material.
1:49:03 And every one of my kids are different,
1:49:04 but they came from the,
1:49:09 and so remixing is like how everything happens.
1:49:12 And it’s like Matt Ridley said, ideas having sex.
1:49:15 And so to your core question, yes,
1:49:18 I think that if I do a calculation
1:49:20 and I’m using a calculator instead of like
1:49:22 doing math by pencil,
1:49:25 that calculation is still an input into my output.
1:49:29 If I’m using AI to generate art
1:49:33 and it’s my prompt instead of the gesture of my brush
1:49:35 and the strokes of my hand,
1:49:37 then I think it should still be mine.
1:49:41 Even if it was trained like a great art student was
1:49:44 by staring and learning and studying and then emulating.
1:49:45 And then these things evolve.
1:49:47 You look at Picasso through all the different phases
1:49:49 of his style of art,
1:49:54 from like realism and portraiture to cubism and abstract.
1:49:57 These things are just evolved until you find the white space
1:49:59 that defines you.
1:50:01 And that goes back to like,
1:50:03 if I train all my AI and everything I’ve ever wrote,
1:50:06 but then like my voice is,
1:50:08 a new voice is rare and hard to create.
1:50:13 So I actually think we should probably worry more about
1:50:16 how do you break free from this constraints of these things?
1:50:20 Then, should they be copyrightable?
1:50:22 – Well, that goes back to our earlier conversation.
1:50:24 Do we just end up in this lane that we can’t get out of?
1:50:26 Or we don’t even recognize we’re in a lane, I guess.
1:50:27 – Right.
1:50:29 – In some ways, even more devastating.
1:50:31 – And the brilliance of all of this, again,
1:50:33 like I’m a big believer that we make our fictions
1:50:35 and our fictions make us.
1:50:37 And if you’ve watched Westworld,
1:50:38 I don’t know if you’ve-
1:50:40 – Watched an episode or two, yeah.
1:50:41 – The first episode, you know,
1:50:44 you have a guest who comes to the park
1:50:46 and he’s sort of squinty-eyed looking at the host
1:50:48 who’s actual robot that you learn later on,
1:50:50 but he doesn’t know at the time.
1:50:52 And he’s like looking at her and she goes,
1:50:53 you want to ask, so ask.
1:50:55 And she knew what he was going to ask her.
1:50:56 And he goes, are you real?
1:50:57 – Yeah.
1:51:00 – And she goes, well, if you have to ask, does it matter?
1:51:03 And, you know, I’m going to sort of spoiler alert
1:51:04 on Westworld.
1:51:08 It’s all about these hosts interacting with the guests
1:51:10 and they’re there to serve the guests.
1:51:12 But in fact, it’s the opposite
1:51:14 because every host is watching and learning
1:51:17 every small nuance, every gesture,
1:51:18 every inflection of your voice,
1:51:21 every cadence of your speech.
1:51:24 And it’s learning you so that it can basically create
1:51:25 a perfect simulacrum of you
1:51:28 and 3D print biologically, a version of you.
1:51:32 And so it’s a really profound philosophical question
1:51:34 about how we’re interacting with these things.
1:51:37 But all these things have been trained
1:51:41 on the sum total of all human creation.
1:51:42 And now they’re being trained
1:51:43 on the sum total of human creation
1:51:45 plus artificial creation.
1:51:47 And some of that is done with human prompts
1:51:49 and some of it is going to be done automatically.
1:51:50 But I just think it’s going to be part
1:51:54 of the total overture of creation.
1:51:56 And I think it’s a beautiful thing.
1:51:59 So does anything about this scare you?
1:52:02 About AI, like the direction we’re heading?
1:52:06 I think in the near term, the thing that scared me
1:52:10 is what again, scarcity and abundance,
1:52:13 what becomes abundant is people’s ability
1:52:14 to use AI to produce content.
1:52:17 And I don’t know if I’m getting an email from somebody.
1:52:20 Did an AI ride it or optimize it
1:52:23 or was it really a thoughtful note from John?
1:52:26 This young college student who’s persistent,
1:52:28 was it really them?
1:52:31 And can I infer something about their persistence
1:52:32 and their style of writing?
1:52:34 Or did they put it into an AI and know
1:52:37 from the repository of what influences me
1:52:39 and what I’ve talked about and what I care about
1:52:40 that they, you know, so many people are like,
1:52:41 oh, I heard you on this podcast
1:52:43 and I felt compelled to write you
1:52:46 because I too care deeply about family
1:52:46 and you know, blah, blah, blah, right?
1:52:48 I mean, those are surface level stuff
1:52:51 but somebody that’s more nuanced about it.
1:52:53 Am I being manipulated by them or by the AI?
1:52:55 And if it’s them, there’s a cleverness to it
1:52:56 that I might admire.
1:52:59 If it’s just the AI, I feel suddenly more vulnerable.
1:53:02 So what becomes abundant is the sort of,
1:53:04 not just information, misinformation or whatever
1:53:05 but the production of it,
1:53:08 what becomes scarce is veracity and truth.
1:53:13 And that to me was less scary
1:53:16 but more you need to be inoculated
1:53:18 and immunized, vaccinated
1:53:19 and you’re almost going to become
1:53:21 a little bit more distrusting
1:53:24 but like your reactions right now,
1:53:25 I might say something and you might say,
1:53:28 oh, and maybe you actually thought it was profound
1:53:30 or maybe you’re like, this is not interesting at all
1:53:32 but there’s something authentic, right?
1:53:35 About this and we are reading each other
1:53:36 and reacting to each other.
1:53:38 That to me is going to become ever more valuable.
1:53:43 So our humanity, the interactions are the scarce thing.
1:53:48 Even if and as through other mediums, it’s hard to tell.
1:53:50 I love that.
1:53:51 I always love talking to you too.
1:53:55 So I get so much energy and ideas out of our conversations
1:53:56 and I’ll be chewing on this for weeks.
1:53:59 I know we always end with, what is success for you?
1:54:00 You’ve answered this before,
1:54:02 I’m curious to see how it changes.
1:54:05 It really is the eyes of my kids.
1:54:07 It is for me, them saying my dad did that
1:54:10 or my dad made that or my dad was present for me.
1:54:13 And I think it’s the story I tell myself
1:54:16 about my own life and my relationship with my father
1:54:17 and wanting to invert that.
1:54:22 And so for me, success is like them saying I’m proud.
1:54:24 He was my dad, he was a great father
1:54:26 and I’m proud that he does all these things.
1:54:30 And when we find a company or like some of these secrets
1:54:33 that I talked about, I share them with my kids.
1:54:35 And so I was taking my middle daughter
1:54:37 to my oldest place tennis, my middle does soccer.
1:54:40 My little guy plays basketball like 10 days a week.
1:54:42 He’s better at nine years old than I was at 19
1:54:44 and I was reasonably good.
1:54:47 And I like sharing these stories.
1:54:49 So I’m like, you know, next week there’s a story
1:54:52 that’s gonna come out about this particular thing.
1:54:55 And nobody knows about it except the company and now you.
1:54:57 And they’re like, oh my God, really?
1:54:58 And I’m like, yeah.
1:54:59 And like, you can’t tell anybody, you know?
1:55:01 And I just, I love that feeling.
1:55:02 – That’s awesome.
1:55:03 – And I do it in part.
1:55:05 Not because I want them to learn about it,
1:55:06 but I want them to be proud of me
1:55:09 as selfish and Vainglorious as that is.
1:55:11 And to be like, oh, my dad’s cool, you know?
1:55:12 So that’s a success.
1:55:13 – I think you’re cool.
1:55:15 You’re not my dad, but man.
1:55:17 – I’ll tell you, my 15-year-old daughter
1:55:19 definitively does not think I’m cool.
1:55:21 She says, you are so cringe.
1:55:24 – And I think everybody’s kids say that, right?
1:55:25 It’s the same with my kids.
1:55:26 Like, instead of telling them something,
1:55:28 all sometimes they might be listening to this,
1:55:31 but I’ll get my friends to tell them.
1:55:32 And then all of a sudden it holds weight.
1:55:35 But if I tell them the same thing, like, yeah, whatever.
1:55:37 – Same thing with our spouses and that, yeah.
1:55:38 – Thank you very much.
1:55:40 – She ain’t always great to do with you.
1:55:43 I admire what you’ve built and the repository
1:55:46 and compendium of the ideas and the minds
1:55:46 that you’ve assembled.
1:55:48 It’s like a great thing for the world.
1:55:49 Thank you.
1:55:50 – Thank you.
1:55:52 Thank you for listening and learning with me.
1:55:54 If you’ve enjoyed this episode,
1:55:56 consider leaving a five-star rating or review.
1:55:58 It’s a small action on your part
1:56:00 that helps us reach more curious minds.
1:56:03 You can stay connected with Furnham Street on social media
1:56:06 and explore more insights at fs.blog,
1:56:08 where you’ll find past episodes,
1:56:11 our mental models, and thought-provoking articles.
1:56:13 While you’re there, check out my book, “Clear Thinking.”
1:56:16 Through engaging stories and actionable mental models,
1:56:20 it helps you bridge the gap between intention and action.
1:56:24 So your best decisions become your default decisions.
1:56:25 Until next time.
1:56:27 you
1:56:29 you