Skip to content

Podcast

Josh Wolfe- Human Advantage in the World of AI

The Knowledge Project

Source ↗ ← All highlights
  • Limits of Human Intelligence
    • Human intelligence limitations are rooted in biology.
    • AIs will eventually understand us better than we understand ourselves. Transcript: Josh Wolfe How do we create an unfair advantage in a world of AI? I think the limits to human intelligence are rooted in our biology. AIs over time will understand us in many ways better than we understand ourselves. (Time 0:00:12)
  • Shift to Memory
    • AI growth may shift from compute to memory.
    • On-device inference could make memory a key player like GPUs were for AI. Transcript: Shane Parrish You mentioned sort of the limitations, like biology moves at a certain speed. There’s a couple subset of questions here, but one is, where’s the limitation in AI growth right now? It seems like we have energy as a key input, we have compute as a key input, and then we have data slash algorithms as like the next key input. What am I missing? What’s the limitation on those? Josh Wolfe Well, start with conventional wisdom, which has heretofore been correct, which is you need more compute, you need more capital, scaling laws for AI, just throw more GPUs and processors And money and ultimately energy to support that, and you will get better and better breakthroughs. The counter to that, which you’re seeing in some cases with open source or people that have now trained on some of the large models, is that there’s going to be a shift towards model efficiency. And so that’s number one, that people are going to figure out how do we do these more efficiently with less compute. Number two, which is a big sort of contrary thesis that I have, is that a significant portion of inference, so if you break down training, you still need large 100,000 clusters of H100 Top-performing chips. It’s expensive. Only the hyperscalers here before have been able to do that. You can do some training if you’re using things like Together Compute in some of our companies without having to do that yourself, sort of like going back to on-prem versus colo versus Cloud transition 15 years ago. But I think that you’re going to end up doing a lot of inference on device, meaning instead of going to the cloud and typing a query, like 30 to 50% of your inference may be on an Apple or an Android device. And if I had to bet today, it’s Android because of the architecture over Apple. But if Apple can do some smart things, maybe they can catch up to this. But some of the design choices and the closed aspects, which have been great for privacy as a feature, may hurt them in this wave. And you could already see like perplexity can actually be an assistant on my Android device. I carry both so I can understand both operating systems. You can’t do that yet on iOS and Apple. But here’s the insight. If 30 to 50% of my inference is my cached emails and my chats and my WhatsApp and all the stuff that I keep on my device, my photos, my health data, my calendar, then the memory players may Play a much bigger role. So now you have Samsung, SK Hynix, Micron that are important here. If I had to come up with somewhat pejorative analogy, I think Samsung is going to be more like Intel. I think that they’re just a little bit sclerotic and bureaucratic and slow moving and it’s going to sort of ebb and decline. I think Micron is a US company which is going to be more constrained by restrictions that are put on export control. And I think SK being a Korean company is going to be able to skirt those in the same way that NVIDIA has with, you know, distribution to China through Singapore and Indonesia and Malaysia, Which is now being investigated. But the memory players are likely to be ascendant. And so what heretofore was a bottleneck on compute could shift attention, talent, money into new architectures where memory plays a key role on small models on device for inference. Ryan Reynolds here from Mint Mobile, with a message for everyone paying big wireless way too much. Please, for the love of everything good in this world, stop with With Mint, you can get premium wireless for just $15 a month. Of course, if you enjoy overpaying, no judgments, but that’s weird. Okay, one judgment. Anyway, give it a try at mintmobile.com/switch. Upfront payment of $45 for three-month plan equivalent to $15 per month required. Intro rate first three months only. Then full price plan options available. Taxes and fees extra. See full terms at mintmobile.com. (Time 0:07:54)
  • Nvidia’s Future
    • Nvidia’s high margins and market cap allow for reinvention.
    • Jensen Huang’s capital allocation skills are key to Nvidia’s success. Transcript: Shane Parrish You think NVIDIA has got a long runway? Josh Wolfe I think that their ability with that market capitalization and that cash and the margins that they have, they can reinvent and buy a lot of things. So I think Jensen is a thoughtful capital allocator. I think he’s benefited and caught lightning in a bottle over the past 10 years and now just particularly in the past six. I would not count NVIDIA out. Now, what’s the upside from two and a half or three trillion to five trillion or 10 trillion? Or do they go back down to, I have no idea. So that’s more of a fundamental valuation based on the speculation. But if you have 90, 95% margins and your chips are $30,000, could you shrink them down and sell for $3,000 and take small margins, but get more volume? And this was some of the debate that I think happened with the release of DeepSeek, where you had Satya basically talking about Jevin’s paradox, that any one thing might get more efficient, But the result is not that you have less demand. In aggregate, you have much more demand. You know, the classic example of this is like refrigerators. A single refrigerator back in the day was an energy hog. All of a sudden, you make these things more efficient. And what happens? It becomes much cheaper. So if something’s cheaper, you’re going to buy more of it. And they shrunk refrigerators down. And so now everybody had one in their garage and in their office and in their basement. And so the aggregate demand for electricity and then for all the components and coils and refrigerant went up, not down. Same thing with bandwidth. You know, if you have a 56K baud modem, which was like my first modem, you know, dial up internet and all that kind of stuff on CompuServe. And, you know, then you go to a T3 and fiber optic, you know, at the speed of light, it is way more efficient. But your usage now is just huge. You’re streaming 4K videos and watching Netflix and trading on Bloomberg. Whereas if you actually want to decrease use, the non-obvious thing that you would want to do is actually slow down speed. I mean, put me on a 56K baud modem today, I’d just like pull out my hair and never use it. (Time 0:12:12)
  • AI in Medical Diagnosis
    • Josh Wolfe uses LLMs to analyze his medical records and find potential missed diagnoses or relevant correlations.
    • He highlights the inherent biases in human judgment, even in elite professions like law and medicine. Transcript: Shane Parrish Think what I’m interested in is at what point will machines, like if intelligence is the limiting factor on progress in certain domains or areas, it strikes me that in the near future, The machines will be able to surpass human intelligence. For sure. And if that’s the case, then those areas are rife for either disruption or rapid progress. Josh Wolfe Well, I think, you know, what Peter Drucker never imagined when he was talking about knowledge workers and the shift from like blue collar workers to white collar workers was that machines Would actually most threaten those professions. And so you take some of the most elite professions, doctors, the ability to do a differential diagnosis today. I take my medical records as soon as I get them from top doctors and I still put them into LLMs to see what did they miss. And sometimes it unearths some interesting correlations. Is there a scientific paper from the past 10 years that might have bearing on this particular finding or something in a blood test? So that is really interesting. Lawyers, language is code. Multi-billion dollar lawsuits sometimes come down to the single placement and interpretation by a human of a word. One of the things that Danny Kahneman recognized, and he published this in his last book, Noise, was that the same case, the same information, the same facts presented to the same judge At different times of day or presented to different judges, they are not objective. And he actually thought that for justice and fairness, that you would want the human intelligence applied to these situations with their biases to be either complemented or replaced By AIs that had a consistency and a fidelity in how they made decisions. So those are all high-paying jobs with lots of training, lots of time to gain the experience, the reasoning, the intelligence, and many of those things are at risk. Government itself, the ability to legislate, make decisions, you want to be able to capture and express the will of the people. But increasingly, we have social media that’s able to do that. It could be corrupted, but there’s mechanisms to figure out how do you really surface what does the populace care about? And some in Europe have tried to do these things. The key thing is always like, what’s the incentive and what’s the vector where somebody can come in and corrupt these things? But interestingly, I actually think that the people whose jobs are like most protected in this new domain are blue collar workers. A robot today can’t really fully serve a meal and they cannot effectively, even though every humanoid robot tries to do folding laundry. And there’s still basic jobs that are not low paying, but they’re arguably safer than ever. And this ties into immigration and technology and human replacement. But people that are doing maintenance, people that are plumbers, many of these things are standardized systems. But it’s like the old joke about the plumber that comes in and, you know, he comes in and he taps a few things and, like, suddenly the pipes are fixed. He says, you know, how much is it? It’s a thousand bucks. A thousand bucks for that? He’s like, it was a $5 par. There’s still a lot of this like tacit knowledge and craft and maintenance that is going to be protected against the rise of the machines that are going to replace most of the white collar Intelligence and knowledge workers. (Time 0:15:48)
  • Impact of AI on Coding and Margins
    • AI coding will democratize coding but empower high-skilled coders further.
    • Company margins may not see permanent increases due to competition, except for monopolies like Nvidia. Transcript: Shane Parrish You believe, I think it was Zuck who came out and said, you know, by mid this year, 2025, that AI would be as good as a mid-level engineer. Encoding. Encoding. Yeah, for sure. What are the implications of that? Walk me through like the next 18 months, if that’s true. Josh Wolfe Well, again, if you take this in the frame of Jevon’s paradox, then a lot more people are going to be able to code in ways that they never have. And in fact, I think it was like Andre Kaparthi, who was on Twitter a day or two ago, talking about who he himself as a coder was basically just talking in natural language and having the AI, and I forget which one he was particularly using, generate code. And then if it tweaked something and he wanted to change a design and make a particular sidebar a little bit thicker or thinner, he would just say, make the sidebar. And it was able to do that. So I think the accessibility for people who never coded, never programmed to be able to come up with an idea and say, oh, I wish there was an application that could do X, Y, and Z to be able To quickly do that is great. For the big companies who employ many coders that are competing at an ever faster speed, you know, you have somebody like Mark Benioff who’s saying that they’re not hiring any more coders At the same time that he’s still talking about the primacy of SaaS, which is this weird contradiction. But I would suspect that maybe you lose 10 to 30% of the people that you normally would have hired, but the people that are there are still like these 10x coders, and now they have a machine That’s helping them be like 20 to 100x. Do you think margins go up then for a lot of these companies? I don’t know. I always feel like margins are always fleeting in the sense because it’s like a fallacy of composition. One company stands up a little bit higher and then everybody else is on their tippy toes. So I think it just changes the game. But I don’t think that you have some like permanent margin. The only time you get really large margins is when you truly have like a monopoly. NVIDIA today, until there’s an alternative, whether in architecture or algorithms or in something else, you know, they’ve got dominant margins because they can charge super high Prices because there is no alternative. So when you have that, there is no alternative. But in many domains, given enough time, there’s an alternative. And then margins just resettle and look at cars. You know, the average margin on cars. Cars today are 10,000 times better by every measure of fuel efficiency, comfort, air conditioning, satellite radio. But those margins never persisted as being like permanently high. (Time 0:19:02)
  • Benefit from Competition
    • Wait for competing AI services to improve and become cheaper for consumers.
    • Competition among providers will benefit users. Transcript: Shane Parrish I always come back to sort of Buffett in the 60s with the loom, right? I always relate everything to the loom because everybody was coming to him when he first took control. And they’re like, oh, we got this new technology. And he’s like, yeah, but all the benefits going to go to the customer. It’s not going to go to me. Josh Wolfe I mean, look at AI right now. Billions of dollars being spent. All the foundation models, all the competition. The second that DeepSea comes out, it suddenly accelerated the internal strategic decisions from OpenAI of when are we going to release models. And so the flurry of all of this benefits us as consumers always and everywhere. And so we were looking at internally installing some new AI system to surface all of our disparate documents. And there’s a bunch of these and our Gmail, our Slack, our Google Docs, our PDFs, our legal agreements, and just have a repository with permissions and all this. And it’s expensive in part because going back to that ignorance arbitrage, somebody could charge us a lot of money to do that implementation. And my default was just wait. Why don’t we wait six months? Because this is going to be available from all of the major LLM providers today that want to get the enterprise accounts. And let’s just wait and they’ll compete and compete and it will accrue to us as users. Talk (Time 0:21:22)
  • Sovereign AI and Data Moats
    • Sovereign AI models are emerging to avoid reliance on US and China models.
    • The real value in AI will be in proprietary data repositories, like Meta’s WhatsApp and Instagram data. Transcript: Shane Parrish To me about all these models. People are spending hundreds of billions if not trillions of dollars around the globe competing on a model. Do you think that’s the basis of competition or how does that play out? And then you have Zuck who’s trying to open source it. And he spent, I don’t know, what, 60 to 100 billion probably by the end of June this year open sourcing it. So he’s basically like, I wouldn’t say he’s doing it for free, but what’s the strategy there? Josh Wolfe First, you have just straight head to head competition, you know, Anthropic and Claude and Chachapiti and OpenAI and others. Then you have sovereign models. So there are countries that are saying we don’t want to be beholden to the US or to China. We funded a company in Japan called Sakana. This is one of the lead authors from the Google Transformer and a guy, David Ha, and just incredible team. And they are actually trying to do these super efficient novel architecture. So they’re not trying to train these multi-hundred thousand clusters. Their latest model, which was based on this evolutionary technique, was like eight GPUs, which was wild. So that’s one trend. But on the strategic question for Zuck, I actually think that he’s probably playing at the smartest of everybody, which is – and he’s been open about this. We’re going to open source the models with Llama, and we’re going to let people develop on them. Why? Because the real value is going to be in the repository of data, longitudinal data, deep data. If you go back 10, 15 years, like the number one thing in tech was big data, big data, big data. Okay, well, now if you actually have big data, you want to use whatever models are out there to run on your proprietary silo of data. So the people that I think are going to be advantaged, meta, why? They’ve got all my WhatsApp message. Apple doesn’t, meta does. They’ve got all my Instagram likes and preferences and every detail of how long I spend and linger on something and what I post and all of that content, my Facebook, which I don’t really Use anymore other than when Instagram cross posts to it. But that is super valuable. And they care about that in part because Zuck needs to route around both Apple and Google. He does not have a device. I mean, you’ve got Oculus and MetaQuest and whatnot, but that’s not the one. Orion with the neural band from the company control labs that we funded, which was for the non-invasive brain machine interface to be able to use free gestures, which is an absolute Directional arrow of progress, right? Disintermediating the control surfaces you have, remote controls and all that kind of stuff. And just being able to gesture, map a device to your human body is absolutely the trend. But he’s thinking about how do I route around these devices and how do I have a long repository of everybody’s information and use the best model that’s out there. And the great thing about open source is it’ll continue to improve over time. So I think that that’s a winning strategy. I think the people that are continuing to develop ever better models, unless they have proprietary data, are going to be sort of screwed. Bloomberg should do really well. I mean, the huge amount of proprietary information, all the acquisitions that they’ve done over time, being able to normalize merger data and historic information and the longitude Of price information and correlations between different asset classes, being able to run AI on top of that is like a quant’s dream. So I think that people that have hospital systems, arguably some governments have used efficiently, but anybody that has a proprietary source of information, clinical trials, failed Experiments inside of pharma companies, being able to do that is the real gold. And the large language models are effectively like over time, I think, going to trend towards zero in a commodity excavator of that data. (Time 0:22:31)
  • AI Art Recreation
    • Josh Wolfe uses AI to recreate art he likes.
    • He felt bad about recreating an artwork with AI after describing it. Transcript: Shane Parrish Aside from meta, what counterintuitive sort of public companies would you say have like really interesting data sources? Josh Wolfe That’s a good question. I haven’t really spent a lot of time on that to figure out who’s got crazy amounts of proprietary data. Pharma would be a good one because obviously they’re, you know, tracking both their successful but their unsuccessful clinical trials. There’s a lot of information in the unsuccessful data, like the things that fail that you can learn from. You could argue that Tesla, of course, who I’m very publicly critical of, like if they truly are collecting a ton of road user data from, you know, every Tesla that’s being driven, that Would be valuable. Anything where there’s a collection, a set of sensors, a repository of information that is owned by them. Anything that we’ve signed off on that your data is free for us to use, like Meta. Shane Parrish You think of Tesla, they should have the best mapping software in the world. They literally drive millions of miles every day. They can update everything. They can locate police. They can locate speed cameras. They can get real-time traffic. Weather patterns. Yeah, totally. Josh Wolfe But okay, the flip side of that though, right? Taking sort of like the opposite view for a moment, Netflix. Netflix has all of our viewing data. They know what we like. They know what you like. They can make a perfect set of channels for you. And the recommendations are reasonably good approximations of adjacencies to things that you liked, but they haven’t been successful, nor has the human algorithm at, say, HBO in The past of, like, perfectly creating the next show that you really want to see. And what’s interesting about that is they’ve put a lot of money into this, but it hasn’t yielded the recipe maker for like the next perfect show. And oftentimes the thing that you want to see is almost something that’s orthogonal from what you’ve been watching. Like I heard Anthony Mackie, who’s in the latest Marvel movie, talking about an expectation that he’ll be – somebody was like, how long do you think you’ll be doing marvel movies he’s Like i think probably like the next 10 years and i’m like probably two or three because people are just bored of this stuff after a while like nobody wants to see another i don’t want to See another marvel movie i i like the adjacency of like the boys yeah which was like the dark you know superhero kind of movie and i think trying to find groups that have proprietary data That have some predictive value, the most value probably for society, I don’t know if it’ll be entirely captured by companies, is just all the scientific information that we have. Because I’m absolutely convinced that (Time 0:26:20)
  • AI-Driven Science
    • Automated labs and AI will revolutionize scientific discovery by enabling 24/7 experimentation.
    • Josh Wolfe believes AI-driven science will be a game-changer. Transcript: Shane Parrish We are going to have machines doing science 24-7. Well, so talk to me a little bit about, I want to come back to Tesla in a sec, but let’s go down the science. Why has nobody sort of taken every study published in a domain, say Alzheimer’s research, popped it into GPT and be like, where are we wrong? What studies have been fabricated or proven not true that we’re investing research in, right? Like, because studies get built on studies and studies. And so if something from the 80s came out and it’s like completely false, we’ve probably spent $20 billion down this rabbit hole. And what’s the next most likely thing to work? Is anybody doing that? Josh Wolfe I have to imagine they are because deep research came out today or in the past 24 hours from OpenAI, which is sort of their model with a better engine, so to speak, than Google’s deep research, Which itself was impressive, both because of its ability to search many sources and then the ability to, you know, sort of, I think it was either there or through Notepic LM to conjure The podcast, which at first was a static presentation of near human quality voice. But now you can interrupt it like a radio caller, which is super cool. But you can say, go through the past 15 years of PNAS papers or science and nature papers around this particular topic and find correlations between papers that do not cite each other Or tell me any spurious correlations. And the beauty of all of that on sort of the information or informatic side is eventually you will have a materials and methods output of that, that you can feed into something like Benchling Or some of the automated lab players to actually say like run the experiment. So I’m absolutely convinced, like high certitude. I don’t know exactly which company will do it. We’ve invested in some, they haven’t worked. We’ll invest in more, hopefully they will. But this directional arrow of progress of the idea by analogy, of machines doing science 24-7 automated is going to happen. I’ll give you one or two analogies. If you were a musician back in the day, if you and I were starting a band, we would have to go and get studio time here in New York City or Electric Ladyland or whatever. You bring your instruments. Okay, maybe you could rent the instruments there. And then GarageBand and Logic and Pro Tools pops up. And now we don’t have to be in the same physical space. My instrument is virtualized. I can create a temporal sequence of notes. I can layer them in. You could play drums, you could do vocals, blah, blah, blah. Okay. Science is the same thing. I can be on the beach in the Bahamas and conjure a hypothesis and use one of the AIs to test the hypothesis and look at past literature searches, see freedom to operate, see if there’s white Space, and then tell one of these cloud labs that is literally like sending something to AWS back in the day and say, run this experiment. And the beauty of this is the robot will do the, well, here’s the beauty, the virtue and the vice. The robot will do the experiment and it should do it perfectly because it’s digital and it’s high fidelity. The vice is so much scientific breakthrough has often happened because of serendipitous screw ups. Right. And so you want almost to engineer like a temperature on an AI model, a little bit of stochastic randomness so that the machine can sort of screw it up to see what might happen. Because, you know, penicillin and Viagra and rubber and vulcanization, all these things happen by like random processes and then post-factor like, huh, that’s funny. And then, you know, you run with it. But then the machine will say, here’s the results. And it will then reverse prompt you and say, do you want to run the experiment again, but changing the titration of this to 10 milliliters instead of five. And you just click a button from the beach and you’re like, yes, and the robots run it. Whoever ends up creating and building that, I think is going to make a fortune. (Time 0:28:50)
  • Creative AI Use
    • Josh Wolfe uses AI to rewrite boring school PTA messages in entertaining writing styles.
    • He finds it fun to make mundane content interesting using AI. Transcript: Shane Parrish Isn’t that low-hanging fruit, though? Just like, look at all the work that we’ve done so far. Tell us where we’re on the right track and where we’re going astray. There’s nothing preventing that from happening today. Josh Wolfe I mean, it’s very like David Deutschian. It’s like, if it obeys the laws of physics, it should be possible. There’s nothing about this that is totally speculative and fantastical that it doesn’t obey the laws of physics. Shane Parrish The other project that I wanted – I was thinking about sort of doing is just calling like a priorart.org or something and having AI read through all the patents and make the next adjacent Sort of like patent, like improvement, and then just publish it because then there’s prior art. Right. Well, yeah, there will for sure be AI patent trolls, if you put it negatively. This would dissuade that. I mean, in a sense, it would sort of be making prior art for as much as you can 24-7. Josh Wolfe There are companies that do this where they have creative patent filers for continuations in part so that they can keep sort of the life of this going. But the rise of agentic AI, you can have some crazy idea, some brain fart at 9.30, 10 o’clock at night. And you just had a cocktail with friends. And you’re like, oh, like imagine if that exists. Well, do the research. Does this exist? And if it doesn’t exist, can you write a patent and sketch a diagram for me and file it? And start and incorporate a company. All those things have to go at the speed of like certain processes, but all of that could be done overnight where you literally wake up and there are multiple agents working on your behalf That have filed a patent, created a design, incorporated a company, possibly even put it out to some group and raised money for it overnight that have opted into it. And I don’t know if there was like successful enough and it hit a bunch of criteria, I might allocate some capital into an account to allow a robot AI to actually receive pitches, respond To it and create a small portfolio to allocate as an experiment. So you can see this whole thing is just like a human idea, or maybe one inspired by interactions with an AI, by the time you wake up, you have a company started and the basis for people to Actually do work. Like that in 10 or 20 years, people look back and be like, how did we not see that coming? And look at all the jobs that are being created because every single person is now creating and has like six virtual companies. The (Time 0:33:56)
  • AI Tools for Kids
    • Embrace and learn AI tools like kitchen appliances.
    • Restrict children’s TikTok use due to negative influence and CCP ties. Transcript: Shane Parrish Future is going to be wild. Yeah. Talk to me, let’s go back to Tesla for a second. Why the hate on Tesla? Josh Wolfe Let me say, I think Elon is amazing at certain things. Elon is arguably the greatest storyteller, fundraiser, inspiration for anybody in the past, maybe all time, truly. I think his relationship with the truth has been questionable. And so in Tesla particularly, I think there was a time where the short sellers started to identify things, not because they just hated the company or hated the future or any of this. And he was able to very shrewdly weaponize the us versus them. They’re trying to kill us, right? Most short sellers I know happen to be very disaffected people and they have a chip on their shoulder. And to me, the motivating force and the incentive for them is not that they just want to make money, but they want to be right. And they want to be right because they’ve identified somebody that they think is intentionally doing wrong. It’s the same thing as an investigative journalist. It’s the same thing as a opposition research for a politician. It’s the same thing for somebody that is trying to debunk a Sunday preacher charlatan that basically is almost intellectually competitive to say, you are trying to pull the wool over These people’s eyes and I know what you’re doing, and I’m going to call you out on it. And so for me with Tesla, I think that they got away with accounting fraud on warranty reserves and a whole bunch of other things. I think there was a lot of prestidigitation and magic of look over here while we’re doing this. And today it doesn’t matter because they got away with it. But I think that there was not the same kind of honesty that I would ascribe to Jeff Bezos in how he built Amazon and raised a few hundred million dollars of equity. You could look at stock-based compensation. You could look at debt as capitalization, but created this monster that is profitable, cashflow positive, and never raised another dollar of equity. And Elon raised north of $50 billion, took out $50 billion, treated it like an ATM, said, I’m never selling a share and sold lots of shares. And then whether he had to do it to buy Twitter or whatever, I just, I don’t feel it was done as honestly as other entrepreneurs that I greatly admire. Now, that said, SpaceX, I have no issue with. I think SpaceX is an extraordinary company. I think it’s an incredibly important American company. I think without it, we would be at a massive disadvantage. I think it is truly a national treasure run by Gwen Shotwell, incredible engineers. We’ve backed a bunch of these engineers that have come out from Tom Mueller to that I just, I think the world of. I’ve just, yeah, been much more critical about Elon’s relationship with the truth as it came to Tesla. And in many ways, I felt like the whole thing was unnecessary. Do (Time 0:36:15)
  • Human Connection Advantage
    • Human connection and shared experiences will be the scarce, valuable advantage in the age of AI.
    • AI can never replicate the unique human ability to connect and share emotions. Transcript: Shane Parrish Like what are the ways that we can, like your kids, like, how are you teaching them to navigate this in a way that gives them an advantage, an advantage, a behavioral advantage? Josh Wolfe So, you know, you go back some years and it was like, okay, we can still beat computers and chess, done. We can still beat them and go, done. We can beat them in video games, done. Okay, but we still have creativity, done. Now, of course, human creativity is not dead. But every day I am doing something creative on AIs that I can’t do. I cannot paint. I cannot draw. I cannot conjure. I like taking photographs. And I like the composition of that. But I can engineer prompts, which itself is an act of creativity, and get the most inspiring muses and results. And I can take works of art that I like and put them in and ask it to describe it and do it six times, particularly in like mid-journey, and recreate from the prompt some alternative of it. There was even an artwork that I loved, which was this mix mash of superheroes that looked like it was put through a blender. I was going to buy the artist’s work and then I couldn’t describe it to Lauren, my wife. And so I put it into Mid Journey and described it. And then I just clicked a button and I made four versions of it. I guess 16 versions because each one was four and it was insane. And I was like, why am I going to buy this? Because I just recreated it and I felt morally bad because I wasn’t copying, but I had a perfect description of the style. And so I thought that that was pretty wild. So these are tools that I think kids should be using. They should be learning. It’s just like a language. And I think they need to be versed with it in part to understand the domains to avoid because they’re going to be not void of emotional or aesthetic or moral value, but you’re not going To make money. And so my wife and I debate this all the time about dance. Dance is amazing. We started dating and she took me to this dance thing. I was really not into dance at all. At Parsons Theater, it was this guy, David Parsons, at the Joyce. And he has this performance called Caught. And me being into science and technology, I’m watching a bunch of people dance. I’m not really into it. And all of a sudden, it’s a weird ambient sound. Very electronic, very Tron-like. And there’s a guy in white pants and that’s it. And then a strobe light goes on. Anyway, and the strobe starts flashing and all you see as a viewer is this person floating in the air. But behind the scenes, what they’re really doing is jumping, jumping, jumping, jumping, perfectly timed to the choreography of the strobe. So you see them caught, but they’re doing this crazy kinetic athletic thing, right? And I was just like, that’s super cool, right? Everybody should see this, right? It was just like inspiring. Now that said, we had this debate afterwards because she’s like, you know, these dancers make no money. And I’m like, well, market forces would say that there’s too many dancers or there’s not enough demand. And they collect like unemployment insurance for half the year because they have to work other jobs and they don’t have this and it’s just wild and so we got into a debate about you know What is the societal value of this now i found it valuable and i would go and i would pay money and be a patron but i wouldn’t want my kids to go and pursue that unless it was that they were solving For just the aesthetic passion that they had. But I think that there’s very few crevices where AI will not creep in and either be able to do the thing nearly as good, including you think about the compression that we all enjoy of a Game Of Thrones episode, the compression algorithm of all of that talent, set design, special effects, screenwriting, a lifetime of acting and performing, you know, just swoosh down. One of our companies, Runway ML, you can, and there’s others that you can conjure. Today it’s 10 seconds, but tomorrow it’ll be two minutes and full feature films with no key grips, no lighting, no costume design, no set design, no actors, and voices entirely generated By AI. And so does that strip this art of its soul kind of thing? I don’t think so. I think it just creates a new form of art, just like Pixar, you know, for all the people that were doing Disney Mickey animations by hand, and then suddenly we get this 3D rendered graphics And incredible storytelling. That will be timeless, but the tools that we will use will very rapidly replace these things. So I think our kids should be embracing and using all of these tools. The only tool that I restrict them from is TikTok. But otherwise, and that’s for a variety of bad influence in Chinese Communist Party. But otherwise, I want them learning how to use every tool as they would every appliance in a kitchen. Outside (Time 0:53:53)
  • Diverse Social Circles
    • Encourage diverse friend groups for children to build resilience and belonging.
    • Multiple social circles provide hedges against rejection and isolation. Transcript: Shane Parrish Lot of the things you said, what sort of the common thread with them in my mind was we feel part of something larger than just ourselves. Yes. Like relate this to working remotely or, and how this interacts with other things, right? Where we might not feel part of something larger than ourselves. And remote work is a great example where you sort of, the ability is there, whether you do part-time or full-time to shut off your laptop and the world looks like you. You’re not forced to interact with people with different political views, different socioeconomic status. You surround yourself. So you don’t feel a part of something bigger. And that changes how you vote. It changes a whole bunch of things in your life, I would imagine. I’m speculating here, but I’m sort of like thinking out loud. (Time 1:02:22)
  • Lifespan Increase
    • While lifespan may increase by 10-20 years, doubling it is unlikely.
    • Josh Wolfe appreciates longevity research but doesn’t intensely pursue it. Transcript: Shane Parrish I like that. I want to come to aging a little bit here. I don’t want to do it. I don’t want to age. Talk to me about this because I feel like based on the research happening now, the amount of money going into this, the progress that we seem to be making, at least on biomarkers that we Understand, we seem to be able to dramatically slow at best our aging process. You think we’re going to make a quantum leap in terms of average lifespan for humans, maybe adding 10 or 15 or 20 good years in the next 10 years? (Time 1:06:57)
  • Information Consumption
    • Josh Wolfe reads widely, from diverse news sources to niche writers.
    • He uses PressReader, Twitter, and AI tools like Rewind for information processing. Transcript: Shane Parrish Thing that I don’t want to ever lose. That’s awesome. You process a ton of information. What’s that workflow like for information to get to you? How are you using technology to filter? How are you filtering information? Josh Wolfe So I typically go to bed between 12 and 1230, wake up around seven. Kids, before they leave for school, about 40 minutes. I do a lot of physical activity, usually three days a week between working out, trainer, jujitsu, all kinds of interesting stuff. But probably about an hour to an hour and a half in the mornings of reading through something like 40 papers now. It used to be like seven, but then when I start to travel internationally, if I went to the Mideast, I would find an English version of some of the key papers, same thing in Japan and elsewhere. And so now I read a lot of international papers. And when I say read, I use an app called PressReader. It has the digital replica of the specific version, which I really value. And I know we’ve talked about this in the past, but I like to know what the editor put on C22 that’s not as visible on the website because there’s meta information that the editor is saying, This is not important to be on the front page. But if I disagree and I think that there’s a magnitude of informational importance, that to me is like some sort of edge. Then I will take screenshots of those. And so I will sort of call it scout and scour through all these papers, take screenshots. In some cases, I may even take those screenshots and put them into an AI and basically say, give me a summary of this article or give me the three key quotes that really matter. And so I’ll go down all kinds of like rabbit holes with that. So that’s the first which is just like call it 24 hours worth of information that is basically put into editorial decree you can usually get the ft in new york time like 4 or 5 6 p.m eastern Time and so you have a little bit of information edge because most people don’t get the ft for another 12 or 14 hours but they know they don’t know that they could get it online. But that to me is valuable. And I care about all those things, including not the sophisticated newspapers, but like the less sophisticated ones like USA Today. And I want to know what is the average person going to read when they wake up in a Marriott and get the paper delivered under their door and that kind of stuff. Then Twitter. You know, I have all kinds of lists that I follow. And at a given time, it might be something that is geopolitical and war related where I’m going down a deep hole, or sometimes it’s AI and technology. Sometimes it’s my team and what they’re posting and reading about, but a lot on Twitter, which I find truly invaluable. I mean, I know a lot of people say it’s like this dark cesspool of whatever, but you can just filter through and cut out the people. I’m muting and blocking people all the time. And, and, and I’m discovering all kinds of just absolutely incredible people. It has been, as you know, from one of our first conversations, like this idea of randomness and optionality, it is this huge randomness generator, this huge optionality generator, And the accessibility, and I just, I absolutely love it. So that’s another thing where I’m really rooting for Elon in the continued success of X, because I’ll continue to pay a lot of money for it, and I pay more money for it, but I find it super Valuable. And it’s a real-time pulse. And I’m excited for Grok to continue, because I think that Grok and X just continue to sort of, I mean, that’s a repository. We talked about that before. Repository of information. Shane Parrish One of the first things Elon did was cut off access to Google from the data. Josh Wolfe And I think that’s the right move, right? This is our platform, same way as Meta has. So I think that that repository of everybody’s tweets and retweets and likes and the comments that they’ve made, you can already go on and do this in sort of a relatively superficial Way where you can say roast me. And it will basically off your past, I don’t know, like 25 or 30 tweets sort of roast you based on what you’ve tweeted about. Longitudinal access of people that have 10,000, 100,000 tweets is an amazing pastiche of like, you know, what, there’s an interesting thing here, which we were just riffing on internally, Which I’ll come back to. Just remind me on sort of wrapping yourself in this information mosaic and breaking free from it. But papers in the morning, Twitter, internal Slack, emails, texts you know just like processing all this information i use rewind on my mac which is effectively doing non-stop screen Capture there will be other tools like this in part because i do not remember the source when i saw information it’s sort of the same thing of like if you see a show and you were to ask me today Like where did you i don’t know it was on apple tv like was it on paramount was it on Netflix? Like, I have no idea, right? And in fact, usually when I do the Apple search and it doesn’t show up, it means it’s on Netflix because it can’t search Netflix, right? Yeah, just huge information omnivore, everything. And then there’s some, you know, random writer that I follow that’s like Catherine Schultz of the New Yorker, Adam Gopnik, and people whose style of writing and the selection of their Subjects I find really interesting. Then I’ll go deep into some of their themes. Shane Parrish So you use Rewind, Press Reader. What are the other technological tools that you’re finding super valuable? Every AI. I (Time 1:16:00)
  • Effective AI Prompting
    • Use specific and detailed prompts when using AI, setting context and expectations.
    • Test the same prompt on multiple AI models and compare results. Transcript: Josh Wolfe Might take an essay, read it, ask it to summarize the key points, ask it to put it in different voices, take two different essays and say, where do these things agree or disagree? And so, yeah, like just nonstop. I’m on AI easily more than Google now, but I don’t know, two, three hours a day. Shane Parrish What have you learned about prompting that would help everybody get better results? Josh Wolfe Usually very specific. Like I give it a priming thing. Like you are the world’s, if it’s a neuroscience paper, you are the world’s greatest expert in neuroscience. You have read every paper that has there has been published. You have both a skeptical eye to new claims, but you are also open-minded to interesting correlations that might not have been considered. Read this paper and, you know, give me the three most provocative, non-obvious points and give me the three cliche, you know, and so just, and by the way, I will put them into three different Models at the same time. So I will open three different browsers, you know, arrange them and put it into chat GPT, put it into Claude, put it into one of the perplexity models that’s not running on those two. And, you know, sometimes I’ll mix and match them. I love that. Yeah, it’s sort of like a palette of, you know, mixing. We have not yet done this as a partnership, but we’ve talked about it. Having an AI partner. (Time 1:20:47)
  • Anduril’s Impact
    • Anduril’s success has inspired others to enter the defense tech space.
    • The company’s projected growth and high margins are attracting attention. Transcript: Shane Parrish Technology. And you guys are big investors in Antril. Where’s that going in the future? Josh Wolfe Well, there’s going to be a lot more brilliant minds, I think, that feel comfortable, motivated not only by a sense of purpose, patriotism, but also principle and capital making that They see the things that they doubted early on. Like, why is this time different in another defense company of which there weren’t very many? But seeing Andrel’s ascendancy and valuation and success and program wins, I think has inspired a lot of people like, wait, there’s something going on here. We went from 50 primes down to five. You’re seeing the rise of these neoprimes. I deeply believe that Andrel in the next few years will be a 30 to $50 billion publicly traded business doing single digit mid billions of revenue with software like margins that are Not like these cost plus margins. So that is going to usher in a big wave in their buying companies that are acquiring smaller businesses, but you’ll continue to see that sort of evolution in a world that people realize Is not kumbaya peace and safety. There are bad actors that when we take a step back or on our back foot or a little bit permissive, that they arm up. It happened with Iran. And I think the prior administrations from Obama and Biden were well-intentioned in trying to bring them into the Western world. But it was a sort of ruse, you know, from Iran standpoint, same thing with Gaza and Israel and Russia and who thought that we were going to see a land war, you know, in the 21st century where Russia would invade Ukraine and China and Taiwan and North Korea and the African continent, as we talked about in the Sahel and Maghreb, infiltration of a lot of these groups into South America. I mean, there’s just lots of conflict waiting. And the best way to avoid conflict is to have deterrence. And if Ukraine had nuclear weapons, Putin wouldn’t have invaded. Most of the West and NATO, you know, really said, don’t worry, we got your back. And even though you’re not part of NATO and they never nuclearized, I think the world timelessly, you know, through all of human history is going to face enormous conflict, resource Wars, water maybe next. I think there’s something like 1900 active conflicts around the world around water rights. You look at China and Pakistan control of the water. I mean, there’s just like a lot of resources. You look at disrupting undersea cables, you know, sabotage efforts. You look at deep sea mining. You look at space as another frontier. There’s just a lot of opportunities for zero sum conflict. When you can’t reconcile those conflicts through diplomacy or negotiations or agreement, it goes to violence. And the people that can bring or affect or export violence typically have the upper hand. Is that it is allowing for the low entropy system, even though the country at times seems chaotic, that allows for the high entropy production of entrepreneurial ideas and free market Capitalism and booms and busts, is that we have the most powerful military on the planet. Shane Parrish You could argue that that didn’t just benefit the United States. It benefited Canada, Europe. It benefited a lot of- Mexico, our allies, for sure. Josh Wolfe You can watch many fictional movies that have run these counterfactuals of what would have happened, you know, if Nazi Germany had won or, you know, the Russians had landed in, like, For all mankind, you know, before we did in the moon. Shane Parrish But we’re getting away from an era of, like, here’s a trillion dollar, you know, boat effectively. It depends who you talk to. Josh Wolfe You know, if- (Time 1:27:50)
  • Modern Warfare Analogy
    • Modern warfare involves automated projectiles, similar to primitive conflict.
    • Automation and overwhelming force are key characteristics of the modern battlefield. Transcript: Shane Parrish Shouldn’t we? Like, I mean, if that boat can be taken out by a $3,000 drone, how effective is it? Josh Wolfe Yes, for sure. The asymmetry of a threat of an aircraft carrier against a large fleet of drones, it is very much, if you talk to Sam Paparo, who’s the head of IndoPaycom, he will say it is all about mass On target there’s certain things that you know automation cannot you know do and he wants what he calls which i guess is a technical term a hellscape in that region uh the taiwan straits And south china sea and so that you make it really um impossible for them to have sort of any military dominance. But it is an era where it’s about, you saw this again with Iran and Israel and Gaza and Syria’s missiles and counter missiles and rockets and intercontinental ballistic missiles and Hypersonics and space weapons. It is just about going back to almost like Planet of the Apes, You know, one ape threw another, a rock at another, or a twig, or a stone. Shane Parrish Yeah, the weapons get more powerful, but the behavior doesn’t change. Josh Wolfe We’re back to throwing projectiles at each other, you know, and it’s just they’re automated. They’re at speeds or at levels of attributable, overwhelming defensive forces, that that is the battlefield. (Time 1:31:41)
  • Foreign Policy Priorities
    • Prioritize Africa and the Western Hemisphere in foreign policy to counter adversaries’ influence.
    • Encourage hemispheric unity and preemptive action against extremism. Transcript: Shane Parrish The means to the end, I think the end is a virtuous pursuit. If you were to take over a country effectively and you were in charge of policies and regulations, what would you do to attract capital and become competitive over the next 20 or 30 years? What sort of things would you implement? What would you get away from and not do? Josh Wolfe Well, I have an adjacent answer of if I were Secretary of State or Secretary of Defense for the day, which I’ll give you first, which is I would really put priority on Africa as a continent. And particularly that’s Sahel and Maghreb because I do believe between violent extremists, Russian mercenaries, China infrastructure, you are one terror event away projected Into Europe that creates the next Afghanistan. And suddenly NATO and the US are in there dealing with ISIS. You’re already seeing the first authorized strike by Trump on ISIS in Somalia. And so you’ve got Sudan, Chad, Mali, Niger, like it is just a hotbed of people that were coming from Syria and Afghanistan, Islamic extremists. It is a bad situation. And I think that we should be proactive there before we have to be reactive. And it’s a lot more costly in lives and money and blood and treasure. The second thing I would do is a hemispheric hegemony declaration. I just went to Nicaragua, Nicaragua, for a variety of reasons. We went instead of Costa Rica, but I felt much safer in Costa Rica. And I was worried that I was not going to be able to leave Nicaragua. And we went with a friend who happens to be a prominent journalist. He was not allowed entry into the country. So it really threw our family vacation, his family of five, my family of five for a wrench, because the government is trying to take over the banking system and they don’t want it to be Covered by financial journalists and these kinds of things. And you look at who is in there and you literally have presence from Hezbollah, china ccp from iran it’s a bad situation uh the places that we think in most of central caribbean central America’s caribbean south america are vacation spots where we get our coffee and we go on a nice vacation massive infiltration from adversaries and so i think we are losing the game And i would declare almost like a new monroe doctrine kind of thing where we say the entire Western hemisphere, you’ve got a billion people, both ability to project into the Pacific And the Atlantic. You’ve got mostly English and Spanish speaking people, say for Brazil with Portuguese. You’ve got a ton of resources, a ton of brilliant educated doctors and whatnot. And I would just shore up this hemisphere, particularly against CRINK, you know, China, Russia, Iran, North Korea, and their influence. If we were worried about like Cuba and the Cuba missile crisis and things proximate 90 miles or what from Florida, I think that China is doing very smart strategic things. So us going back in and saying, we’re going to, you know, reclaim the Panama Canal and our influence on it, like, you know, forgetting about provocations of Mexico, of like, you know, The Gulf of America versus Gulf of Mexico. But I think that having influence in that region is really important. So those would be the two things that sect F or sect state that I would do is declare hemispheric hegemony and make sure that we shore up our allies in the region and get out our adversaries And their presence, in part because there’s so much commerce and money and infrastructure that’s going in. And then focus on the Sahel and Maghreb in Africa. For country competition, brain drain, you want the best and the brightest. You need to fund basic research and basic science. It should be undirected. That’s the serendipity and the randomness and the optionality that leads to great breakthroughs. You want capital markets to be these low entropy carriers for high entropy entrepreneurial surprise. So predictable rules and regulations. I would lower, I don’t know why we don’t have a flat tax. I mean, I know why we don’t, but I would just have a flat tax, make it super simple. Rich people are going to scout around and figure out how to get around the tax system anyway, and poor people are burdened by it. I get progressive versus regressive, but I would just simplify our tax scheme massively. Anybody that is coming here and getting an education in this country, I would staple a visa as long as they stay here, work for an American company for at least five years. Let them become, you know, we want the best and the brightest here. We are, as an example again, and I don’t mean to make this like all China, but they are our most dominant adversary. 50% of all AI undergrads in the world today are being graduated by China. In our own country, in the United States, 38% of researchers in AI are from China. So we’re outnumbered even domestically. It’s a big deal. And we used to attract something like 23% of all foreign graduates here. That’s down to 15%. People are either going to other countries or staying in their own country. And so we need that. Shane Parrish That’s what won us World War II. You know, if Einstein would have stayed in Germany or- What causes that to happen? Is it the tax rate? Is it opportunities available? Is it housing costs? Like what are the factors that go into people leaving? Josh Wolfe Well, start with the attracting part. You know, as Walter Riston said, people go where they’re welcome and stay where they’re well treated. So we should be welcoming. Now there’s a debate about immigration and, you know, we should distinguish between, you know, the bad people and like brilliant people and we should want them here. Shane Parrish That just comes down to like basic vetting. Right. Josh Wolfe But, you know, some of that is exploited. You know, a lot of these consultants with Wipro and some of the, you know, Indian business process, BPO is the business process operations or outsourcing. Housing is a difficult one, but people can always figure out. New York City is expensive, but you can live in Long Island City and Brooklyn and Queens. But I think, yeah, housing availability, we have, I mean, our cities are so rich and full with culture and people. And particularly if you’re a young person, you want to be around the density of that because you’re trying to find peers and a mate and, you know, all of that. Even if you’re from another country, you go to New York, you can find your enclave of Korean or Chinese or Russian or Ukrainian or Israeli and Caribbean. It’s all here. So yeah, I think just having a culture that embraces this and encourages it, you already have a robust venture capital system of risk-taking. Many other countries don’t have that. So that’s another thing. But if you were to design a system from scratch, you want openness with security. So some means of vetting. You want a great education system that can attract people that view it as a status to have graduated from that particular school. And people want to be around people, whether this is in a company or in a country that are like them that are competitive and highly intellectual, that they respect or admire want to compete With. So that’s number two, their work, something we did in the 1980s, I think it was in 1980 was the Bayh-Dole Act, where government funding for research would allow the university and the Principal investigator to actually own the intellectual property that became an asset. That asset could be licensed to a company. It absolutely opened the floodgates for venture capital to be able to commercialize that. That happened to coincide with ERISA and allowing retirement plans to go into venture capital. So now you have a pool of risk capital, which you need for taking risk on unseasoned people and unseasoned companies. And then you need a robust capital market system to be able to continue to allocate money. But again, capital goes where it’s welcome, stays (Time 1:41:51)
  • Attracting Capital and Talent
    • Attract capital with open yet secure systems, strong education, and robust capital markets.
    • Staple visas to graduates who work for American companies to retain talent. Transcript: Shane Parrish Awesome. I hope some of the government people we have are listening to this. I want to come to IP for a second and copyright and then wrap it up here. So do you think that AI should be able to create IP or copyrighted material? Like if I tell AI to write a book, should that be copyrightable? And who owns the copyright, the AI or me for the prompt? Josh Wolfe It’s super complicated because, you know, the first debate about this, which is the great irony, right? Because OpenAI investors and stakeholders were up in arms that R1 stole from OpenAI, but you can make the argument that OpenAI is trained on the repository of like the public internet And, you know, every art that’s ever been produced and whatnot. Now, if you were an art student and you went to the Louvre or to the Met or to MoMA and you sat there and studied it or took a picture of it, and then we learn through copying. We learn through imagery and imitation and we remix these things. You know there’s this great what is his name uh kirby who did everything as a remix i just sent it to a friend but it was like updated last year and it’s it’s so brilliant in its compilation Of every facet of culture that you love from books to your tarantino movie to the beatles to art you know to scenes of movies like it was all copied from something you know and and you’re Like wait a second that riff came from this 1940s song from this african-american blues guitarist that john leonard or paul mccartney stole and you’re you know you’re like and so everything Was sort of stolen from somebody it was imitated tweaked And by the way, that’s what we are, right? I mean, you get two people who exist and then there’s this genetic recombination of their source material. And every one of my kids are different, but they came from the, and, and so remixing is like how everything, you know, happens. And it’s like Matt Ridley said, like, you know, ideas having sex. And, and so to your core question, yes, I think that if I do a calculation and I’m using a calculator instead of like doing math by pencil, that calculation is still an input into my output. If I’m using AI to generate art, uh, it’s my prompt instead of the gesture of my brush and, you know, the strokes of my hand, then I think it should still be mine. Even if it was trained like a great art student was by staring and learning and studying and then emulating. And then these things evolve. You look at Picasso through all the different phases of his style of art, you know, from like realism and portraiture to cubism and abstract to these things are just evolved until you Find the white space that defines you. And that goes back to like, if I train all my AI and everything I’ve ever wrote, but then like my voice is, is a new voice is rare and hard to create. So, um, I actually think we should probably worry more about how do you break free from this constraints of these things than, you know, should they be copyrightable? (Time 1:49:06)
  • AI and Copyright
    • AI-generated content blurs the lines of authorship, but human creativity remains important.
    • Breaking free from AI constraints is crucial for true creative expression. Transcript: Shane Parrish Well, that goes back to our earlier conversation. Do we just end up in this lane that we can’t get out of? Or we don’t even recognize we’re in a lane, I guess. Right. In some ways, even more devastating. Josh Wolfe And the brilliance of all this, again, like I’m a big believer that we make our fictions and our fictions make us and if you’ve watched westworld i don’t know if you’ve watched an episode Or two the first episode you know you have a guest who comes to the park and he’s sort of squinty-eyed looking at the host who’s actual robot you learn later on but he doesn’t know at the Time and he’s like looking at her and she goes you want to ask so ask and she knew what he was going to ask her and he goes are you real yeah and she goes well if you have to ask does it matter and You know i’m going to sort of spoiler alert on westworld it’s all about these hosts interacting with the uh with the guests and they’re there to serve the guests. But in fact, it’s the opposite because every host is watching and learning every small nuance, every gesture, every inflection of your voice, every cadence of your speech. And it’s learning you so that it can basically create a perfect simulacrum of you and 3D print biologically a version of you. And so it’s a really profound philosophical question about how we’re interacting with these things. But all these things have been trained on the sum total of all human creation. And now they’re being trained on the sum total of human creation plus artificial creation. And some of that is done with human prompts and some of it is going to be done automatically. But I just think it’s going to be part of the total overture over of creation. Shane Parrish And I think it’s a beautiful thing. So does anything about this scare you about AI, like the direction we’re heading? Josh Wolfe I think in the near term, the thing that scared me is what, again, scarcity and abundance, what becomes abundant is people’s ability to use AI to produce content. And I don’t know if I’m getting an email from somebody. Did an AI write it or optimize it or was it really a thoughtful note from John? This young college student who’s persistent, was it really them? And can I infer something about their persistence and their style of writing or did they put it into an ai and know from the repository of what influences me and what i’ve talked about And what i care about that they you know so many people are like oh i heard you on this podcast and i felt compelled to write you because i too care deeply about family and you know blah blah Blah right i mean those are surface level stuff but somebody that’s more nuanced about it am i being manipulated by them or by the ai and if it’s them there’s a cleverness to it that I might Admire. If it’s just the AI, I feel suddenly more vulnerable. So what becomes abundant is the sort of not just information, misinformation or whatever, but the production of it. What becomes scarce is veracity and truth. And that to me was less scary, but more, you need to be inoculated, immunized, vaccinated, and you’re almost going to become a little bit more distrusting. But like, your reactions right now, I might say something and you might say, oh, and maybe you actually thought it was profound, or maybe you’re like, oh, this is not interesting at all. But there’s something authentic, right, about this. And we are reading each other and reacting to each other. That to me is going to become ever more valuable. So our humanity, the interactions are the scarce thing. Shane Parrish Even if and as through other mediums, it’s hard to tell. I love that. I always love talking to you too. So I get so much energy and ideas out of our conversations and I’ll be chewing on this for weeks. I know we always end with what is success for you? You’ve answered this before. I’m curious to see how it changes. It really is the eyes of my kids. Josh Wolfe It is for me, them saying my dad did that or my dad made that or my dad was present for me and i think it’s the story i tell myself about my own life and my relationship with my father and wanting To invert that and and so for me success is like them saying i’m proud he was my dad he was a great father and i’m proud that he does all these things and and and when we fund a company or like Some of these secrets that i talked about yeah i share them with my kids. And so I was taking my middle daughter to my oldest plays tennis. My middle does soccer. My little guy plays basketball like 10 days a week. He’s better at nine years old than I was at 19 and I was reasonably good. And I like sharing these, these stories. So I’m like, you know, next week there’s a story that’s going to come out about this particular thing and nobody knows about it except the company and now you. And they’re like, oh my God, really? And I’m like, yeah. And like, you can’t tell anybody, you know? And I just, I love that feeling. And I do it in part, not because I want them to learn about it, but I want them to be proud of me as selfish and vainglorious as that is. And to be like, oh, my dad’s cool. Shane Parrish You know, so. I think you’re cool. You’re not my dad, but man. Josh Wolfe I’ll tell you, my 15-year daughter definitively does not think I’m cool. She says, you are so cringe. And I think everybody’s kids say that, right? It’s the same with my kids. Shane Parrish Like instead of telling them something, I’ll sometimes, they might be listening to this, but I’ll get my friends to tell them. Yeah. And then all of a sudden it holds weight. But if I tell them the same thing, yeah, whatever. Josh Wolfe Same thing with our spouses. Shane Parrish Thank you very much, Josh. Josh Wolfe Shane, always great to be with you. I admire what you’ve built and the repository and compendium of the ideas and the minds that you’ve assembled. Shane Parrish It’s like a great thing for the world. Thank you. Thank you. Thank you for listening and learning with me. If you’ve enjoyed this episode, consider leaving a five-star rating or review. It’s a small action on your part that helps us reach more curious minds. You can stay connected with Farnham Street on social media and explore more insights at fs.blog, where you’ll find past episodes, our mental models, and thought-provoking articles. While you’re there, check out my book Clear Thinking. Through engaging stories and actionable mental you’ve it helps you bridge the gap between intention and action. So your best decisions become your default decisions. (Time 1:52:05)