260 - Reid Hoffman: Artificial Intelligence and the Future of Humanity

31:02chat bots yeah a whole like we could spend hours on just even this topic so first is like a simple way of saying why is dog not man's best friend is if someone said my dog's my best friend my first question is well did you have your dog spayed and the frequent answer is yes you're like well then i don't want to be your friend uh you know so interestingly i didn't have my dog fixed because i of that i was like i wouldn't fix my best friend yes exactly well this is among the other things but it's like look it's it's part of friendship is i think the antithesis of friendship is the exerting of power right so the fact that like you can't i don't think be friends with your kids you can grow to become friends with your kids but while you are in power a state of power over your kids you're a parent not a friend right now by the way there's of course all kinds of attributes that are shared, loyalty, belief for the best, willingness to suffer and die for the person.

32:04Like there's a bunch of great things, but it's still a different relationship. And so, and so I think that one of the things that's central in friendship is two folks who are, two entities who are not dependent upon each other, who can kind of be as fully present and as autonomous and then choosing to help the other person become their best possible self as part of it, as I think what's really grounding and what friendship happens. And so, you know, dogs, pets can serve really excellent roles, can even play, you know, therapy animals as we get to the kind of therapist part of this.

32:53But it's until like there is a call it an equals relationship of, you know, a whole bunch of nuance and that, but like that's that that's, I think, fundamental for friendship. Now, I think AIs already can be very effective companions in various ways. You know, like I think Illinois just passed this dumbass law about like AI can't be therapists. And you're like, look, what's your theory of that person who's at 11 p.m. feeling very depressed, etc., and is lonely and needs someone to talk to and may even be self-harming? What's your theory? Right. And frankly, my theory is, it's better for them to have access to an AI companion than not.

33:48Right. At that point, the therapist not available, not there, not there for most people, not affordable for most people, et cetera, et cetera. Not even in the landscape. And AI already can do really good, call it therapeutic companionship. Very patient, listen to you, be there and do things. And obviously you want to be trained in a way that's not like, well, I want to harm myself. Oh, here, let me help you. It's like, no, no, no. Let's talk about why you want to harm yourself and what's going on. And there's a whole bunch of very thorny issues to be navigating here, but I think it can be that.

34:22Now, can it be a therapist? I think the answer is yes, ultimately. I think there may be a training standard that's not there. Therapist is also not a friend's relationship because again, it's like a service relationship and kind of equivalent. And so I think that that's, but I think already today, one of the things that's under discussed is that these models by being trained, like for example, in my own kind of co-founded AI companion startup, Pi, personal intelligence from inflection, it's trained to be kind. Now, if you ask it, you know, you tell it, you're my best friend. It says, no, I'm your companion let's talk about your friends friends are really helpful for life and so on but being kind and and and and having very emotionally positive interactions is good for us in various ways and like there was this whole you know critique of open ii is having a certain set of models that were too sycophantic which of course can be a rare like would you rather it be like sycophantic or would you rather it be you know um uh you know kind of nasty and brutish like well let's go let's go let's let's air a little bit on that side like i thought it was like a little bit of a storm in a teacup again can be have errors there but not not that so finally on the friendship side you know all of this gives you the kind of baseline for how i would think about friends in AI, which is part of the reason I don't think AI can be friends today, is that it's a tool.

35:57It's a service. It's a, you know, it's a, you know, it's kind of a, it's an engaging, it's not an independent entity. And, you know, obviously there's a whole bunch of different questions that come around. Well, when could it be? And when should it be? And when would we want it to be so and when we not want it to be so. And, you know, you could just spend, you know, I think we will spend, you know, many, many years philosophically working out a bunch of this stuff and there will be a bunch of different human positions which will have different points of view on these questions. But that being said, I think right now, I think it clearly cannot be.

36:34I literally had a friend of mine said, you know, Claude is my best friend. I'm like, it absolutely is not. It's extremely helpful to you, right? And I'm not trying to deny your experience of it being so useful to you, like on every different thing that you're turning around to. That's great. That's a good feature. It's good that you're engaging with it. But, you know, a friend, to some degree, like here's a simple way of thinking about the kind of friend. You might say, oh, I'm having a really bad day. And you're going to lunch with your friend. You sit down and find, I'm planning on talking about my bad day with my friend and kind of, you know, working through this.

37:07And then you discover your friend has had a much worse day. Your agenda has completely changed. We're talking about my friend's worst day, right, for this. And that kind of different entity and navigating is part of what's good for you and part of what's good for the two of you in exploring friendship together. Because that kind of the, hey, we are helping each other navigate life in this way is part of the thing that is kind of fundamental to friendship here. I would like to focus for a moment on this question of whether or not AI can be a good therapist. Because one, it's an interesting question.

37:43And it also jogs a funny anecdote that I think I got out of Daniel Dennett's, one of Daniel Dennett's books. A therapeutic stance? Yeah, the therapeutic stance. No, it was the intentional stance. Yes, that's what I was really... The intentional stance. That's how it was junk it. Yeah, yeah, yeah. Yeah, I got it. In the early days of AI, some researcher created a very, very rudimentary program that mimicked a therapist. Oh, yeah. Eliza. Yeah. And I think, but the story is it would ask some very basic canned questions. It would listen. and then if it got stumped or it got a prompt it didn't understand it would just say does this remind you of your father or something to that effect and the developer of this program apparently left his office and his secretary was there and when he came back his secretary was communicating with this program and she wanted him to like leave her with it so that they could be alone and continue talking.

38:54And it does actually connect to the intentional stance because we are just programmed to attribute agency to things that might not have it. And so you hear in literature and philosophy about therapy that a huge part of the therapeutic process is developing a relationship with your therapist. And it doesn't seem based on the anecdotes I've just heard from you that in my own experience that people have trouble forming what they believe to be a relationship with an AI. And so here's what I'm wondering. I think it would be very fascinating to see studies about whether or not it would be more efficacious to deal with a therapy-trained AI, for instance, at all hours of the day, whenever you need it, versus seeing a therapist 45 minutes a week.

39:52And let's just say that it does blow the therapist out of the park in terms of the average results. Now, some people, for instance, like those who are very opposed to AI art because they think it displaces human artists, would be, I'm assuming, extremely opposed to AI therapists because it's going to displace human therapists. And I'm wondering how you deal with this attitude in general that people have toward AI that we simply should not continue producing it because even if it's better than humans at some jobs, it's displacing them and depriving them of economic opportunities. When you need to build up your team to handle the growing chaos at work, use Indeed Sponsored Jobs.

40:41It gives your job post the boost it needs to be seen and helps reach people with the right skills, certifications, and more. Spend less time searching and more time actually interviewing candidates who check all your boxes. Listeners of this show will get a$75 sponsored job credit at Indeed.com slash podcast. That's Indeed.com slash podcast. Terms and conditions apply. Need a hiring hero? This is a job for Indeed sponsored jobs. Hey, it's Kelly Rowland. You may not know this, but I have eczema, so I get how it can steal your time. But why let eczema take over when you can talk to your doctor about ebglis?

41:17Ebglis Lubrikizumab LBKZ, a 250 milligram per two milliliter injection, is a prescription medicine used to treat adults and children 12 years of age and older who weigh at least 88 pounds or 40 kilograms with moderate to severe eczema, also called atopic dermatitis that is not well controlled with prescription therapies used on the skin or topicals or who cannot use topical therapies. Ebglis can be used with or without topical corticosteroids. Don't use if If you are allergic to EpGliss, allergic reactions can occur that can be severe. Eye problems can occur. Tell your doctor if you have new or worsening eye problems.

41:48You should not receive a live vaccine when treated with EpGliss. Before starting EpGliss, tell your doctor if you have a parasitic infection. Paid partnership with Lilly. Respect your time. Ask your doctor about EpGliss and visit epgliss.com or call 1-800-LILLY-RX or 1-800-545-5979. Well, part of how we make economic progress in society is by displacing human labor. And so, for example, the Industrial Revolution displaced a whole bunch of agricultural labor, and we had a lot of improvement because of that. Now, people were, by the way, during the Industrial Revolution, it was part of, again, the challenges of navigation.

42:29They, you know, ultimately ended up much better, but it was the transition. The transitions were difficult. it's the same thing with like why we have this term luddite from, from blood and his followers who were like, we should not have the electric loom because everyone has great leaving careers, you know, and it's kind of this small business. And like, well, it's like, Nope, actually, in fact, we're better off. We've clothed much more of the world. Clothing is much cheaper. The careers change. And that kind of progress, I think is very important. And so I don't think stopping technological progress that has economic economic construction is fundamentally wise at all.

43:12And even if you succeed at doing it, like this is one of the frequent problems you have with, you know, kind of call it unions or labor forces that go, we're going to like, we're going to try to stop AI from being used in the classroom. So the teachers, you know, can continue to teach the way they have now, which is the same way they've been teaching for the last 50 to 100 years. But there's a lot of university professors who have a very similar point to do with this, right, just since we're in a quasi-academic discussion here. And it's just, it's a disaster and ultimately ends in a kind of decaying collapse of society.

43:48And eventually, where else, it's built elsewhere that you just become a very late adopter and your whole society and your children and grandchildren really, really suffer from that. So I tend to think it's like, look, it isn't that you don't, you ever kind of try to hold back for, you know, kind of for this reason. Now, that being said, that doesn't mean you're callous. Like the question is, is how do we navigate these challenges? Like part of the reason I describe AI in superagency as a cognitive industrial revolution is not just because of the upside, but because, well, actually we're going to have these transitions.

44:23And that's a very human suffering moment that we need to do. and it requires some real work for how to navigate this stuff. And, you know, one of the fortunate things with having AI is like we can help. Like, for example, I was mentioning if customer service people are all transitioning, how do we have AI people, you know, AI help someone who used to be working in a customer service job, you know think of other kinds of work they can do learn the other kind of work be assisted in doing the other kind of work finding the other kind of work all of this stuff is is kind of really important that's how it ends up being a therapist now by the way i think a lot of humans will still choose to walk human therapists do that's great nothing wrong with that but let me i'll finish with a small thing which i don't think is a replacement to doctors and i think we have infinite demand for doctors.

45:23So I'm, I'm totally good with it. But like one of the things we have a line of sight to do today is building a medical assistant on every smartphone that could run for call it under$5 an hour, which I think we should provision across as much of the world as we can, of course, as much across, let's say American society as we can. And the reason is, is, you know, the vast majority of people do not have access to doctors on quick basis, and they don't know when they should or not. And if you have a medical assistant that can tell you, no, it's worth a three-hour drive to get to the hospital, right, then it's worth it if, say, you have access to a hospital, et cetera.

46:03And so having that kind of medical assistant broadly available is super important. It will change some ways of doctors' jobs. And already today, one of the things I think that if you're not using, you know, ChatGBT as a second opinion or choose your preferred, you know, highly developed model as a second opinion, you're being foolish because I think it can already help with all that stuff. And so I think that's my direction about how we lead to a much better humanist future, which is the same path that we've been on in building technology today. And actually, one last quick comment just for why for people maybe read super agency is a lot of the ways in which we talk about AI now is the same way the educated elites talked about the printing press.

46:57Right. It's like it's going to destroy human cognition. It's going to remove an ability to discern the truth. It's going to reduce the roles of these experts who have been trained for this and know how to use knowledge responsibly and so forth. And it's like, okay, that's right. That was clearly, you know, anyway, it's kind of understanding, you know, stepping out of yourself and looking back at this moment from good moments in the future. And how do we get to those good moments in the future? I'm going to circle back to medicine in a moment and maybe this medical assistant too, but thank you for allowing me this digression on friendship.

47:38But before we leave existential risk, you said you think it's useful to adopt AI to solve economic problems and that we shouldn't hold back. And I want to talk about using AI to solve existential risks. And one existential risk, though it's not immediately it's it's an extension existential risk for individuals rather than us as a as a species and that's cancer and medical problems and what I wanted to ask you about is Manus AI is that how you pronounce it yep yeah cool I want to ask you about it quite broadly what is it what is it meant to solve Among the kind of amazing benefits that I think are literally line of sight with AI is all versions of human medical care.

48:35Like, you know, today you use ChatGPT and others for medical advice. Generally speaking, they try to tune it down because of medical liability. We need to change those laws in order to do that. But then you get to more advanced questions. And like, for example, cancer is a universal killer. Kills rich, poor, young, old, male, female, you know, all races, etc. And, you know, essentially, you know, part of even though we have one term for it, it basically means that, you know, some cell function in your body gets unregulated and grows. and ultimately, you know, and we, by the way, we both need growth, cell replication, and health, and we need, like, containing it to not die from growth.

49:26And that's what the universal format is, but it's many, many different forms of cancer. And, you know, at a very simplistic broad brush, what we tend to do is try to detect it early so we can remove the very specific cells. but if it metastasizes or we can't do that, then you can't. And then we go to chemo, which is, can we poison you overall just enough to kill all the cancerous cells and hopefully you survive, which is a kind of brutal way of doing this. It's, you know, difficult and not very successful. So the question is to say, well, can we use AI to actually build much better cancer, you know, kind of cures, because there's lots of different kinds of cancer.

50:16And if you could just target the specific cancerous cells versus, you know, dip you in chemo, you know, et cetera, can you detect things early? All these things is something that AI can really help with. It will require specific AI technical work. It isn't just, you know, kind of go grab your favorite language model and make it work. And that was a discussion of Siddhartha Mukherjee, who is the famous author of Emperor All Maladies and one of the great oncology researchers from Columbia. And I were talking about, we're like, yes, we think we could do this. You know, I helped bring a bunch of the AI stuff.

50:58He brings a broad swath of scientific genius and understanding of cancer. And how do we pull that together? Because obviously the more next level cancer therapeutics we can create, beneficial for the quality of human life across the entire globe. how specifically does this work the the process of ai drug treatment and discovery well um some of this we we we haven't really talked about publicly so i can't go go into but the short answer is to look if you if you think about ai is as a as a kind of a massively better search engine because like you know search engines some degree are the early versions machine learning and ai and you say what we're looking for is we're searching for the the kind of the right molecules that will bind with, you know, bind with the cancerous, you know, kind of cells and help, you know, those die, right?

52:19And they'll allow the other cells to continue to grow. That's the kind of thing. And like the search algorithm is the biological, is that biology is a very complicated space. How do you identify like that, the finding the the target molecules within that much larger, complicated search space. Okay. And is there something about cancer in particular that made you, maybe it was just your conversations with Siddhartha, but that made you gravitate toward disease? Well, there was a couple of different reasons, and it's kind of a combination of what makes the biggest difference for human life. There's also part of why we live in economic circumstances.

53:08Cancer cures are extremely valuable because, you know, if you, five-year-olds might die of cancer and, you know, if you can save a five-year-old's life, you know, that's a whole life in front of it. If, like, cancer was only a disease for 80-year-olds, right, It would be much less, you know, kind of both either both human quality life valuable and and and life valuable. And it's it's one of the ones that is least like like I said, our solution set is very, very limited. You know, obviously working on heart disease and working on other kinds of things, you know, really matter for this kind of stuff.

53:50But it's and in most of the things when people are trying to counteract mortality, most people don't realize that what you're really trying to do is things that would apply to children. And so when it's a, you know, create health. You know, situations that aren't basic, like feed them, give them clean water. Very super important things. not something you can solve through a pharmaceutical, but like cancer is one of the really fundamental ones that you can attack there and makes a very big difference. At the beginning of our conversation, we talked about existential risks and how the development of AGI could really help with a lot of them, this being one of them.

54:33And I was just up north talking to Eliezer Yudkowsky, and that's why this is so fresh on my mind. And one of the scenarios he laid out as something he thinks is maybe the most plausible way that AI would wipe out humanity is by producing some sort of, I don't know, virus that would be very transmissible and that would kill everyone. And presumably, this is the sort of development of AI technology that would facilitate an end like that, if that were something that was going to happen. And I'm not suggesting that this is a risk that you have to worry about with Manus AI, but I do wonder if there are any unique risks or risks in general to this modality of drug development.

55:29So there definitely are.

55:37The biggest risk that I see with AI is bioterrorism with AI enabling essentially bioterrorists, whether they're state actor terrorism or independent and so forth. Now, when you get to humanity impacting diseases. Sorry, could I just interrupt for one moment and ask? So your idea of the chief risk of AI is that some terrorist group or a state or whatever will use AI, even if AI is not super intelligent, to help create damaging biologics. Yeah. Think movie or TV series 12 Monkeys, you know, as an example. And that would be a terrorist instance. And so an AI does increase a risk factor here, even with, and this is probably one of the most intense things I put work into in alignment, which is trying to align against this is one of the reasons why it's important that, you know, all players, not just, you know, OpenAI, Anthropic, Microsoft, Google, you know, do safety and alignment stuff, but it's important.

56:55The reason why you should invest in it, like if you're a Grok or XAI or other things, that's part of the reason why the investing in the safety and alignment and red teaming is all very important. Now, the first thing is, is that there's probably some number of tens of thousands of human beings today, could be 20, could be 200, that could create this kind of pandemic today if they were so inclined because it's what do you mean they could do it today well there's ways to create biological um uh biological pandemics that the the larger the the that like uh and i obviously don't go lot of depth in public in this because I don't want to give anyone any nutty people any ideas.

57:49But there's ways to create diseases that are far more dangerous than the diseases we saw today through, you know, current techniques of biological engineering. It is a doable thing for a human being to do the biggest thing ai does is it now moves that that number from whatever tens of thousands to millions if you have ais because oh let me help you do this you know and so that creates a a negative attack surf and that's the reason it's my highest risk factor which is like let's not have ais help with that matter of fact people frequently in this biological risk factor environment say the standard is can google can a person googling help figure it out and i think that standard is actually in fact too too weak because i think we also want to be from search indices and so it's redacting information to make you know crazy people using this for biological terrorism to make it harder to do because there will be some there's eight billion plus people there are some very crazy people, a la the movie 12 Monkeys.

59:04Now, so you go, okay, that increases a service vector. The problem is, is there's no way to reduce that service vector. We're already in the rat risk factor today to zero. It does get increased some of AI, but the only way I know to protect it is AI. Because AI can, like, I think we're gonna get to a lot of monitoring of biological substances and so forth. Are there viruses spreading? By the way, there could be natural ones, not just man-made ones. Nature could stumble across something that would be very destructive.

59:43And you go, okay, how do you detect it? How do you analyze it? How do you create a vaccine or an antidote quickly, etc.? all of that stuff, AI is the only thing that can do it. So I think AI is net positive. It's one of the reasons why I tend to be, quote unquote, an accelerationist about getting there because getting to that point where you could have AI being on the defense column is part of what you do on this kind of existential risk side of it. Now, it's absolutely right. You know, I've also had the thought that one of the things that's entertaining about the Terminator movie is if you had a superintelligence robotic species that would say, let's get rid of this carbon biological species, you wouldn't go, I'm going to create robots with guns and plasma rifles.

1:00:33It's I'm going to create diseases. Diseases affect biological things. It doesn't affect me. That's what you do. And so you want to, you know, kind of minimize the surface of that. That would be a, that would be a attack thing. Now, when it gets to Manus, Manus, you know, is, you know, cancer, like a lot of things are described as disease. Cancer is not an infectious disease. The principal thing in these risk factors is infectious diseases. And yes, if you're dealing with things that have infection variants, you got to be very, very careful. And it's one of the things that's, you know, I've had friends of mine make arguments to me that we You shouldn't do anything that requires high-level biological safety labs because lab leaks and other kinds of equipment is just inevitable once you have a lot of that.

1:01:25And there's no way to get that to zero. And that's a, you know, that's a, you have to have like lab leaks of things that are infectious is like one of the things that is like, like I think one of the biggest dangers for human race. Unfortunately, we have that without AI. And I think AI is ultimately the best defense to it. You mentioned the theory of action a few minutes ago. Did you study philosophy of action here with Michael Bratman at Stanford? Yes, I did. Okay, that's funny. Mind, matter, and meeting. Yeah. Well, I wanted to ask you, pivoting now that we've talked about exit risks to sociological risks.

1:02:11I mean, you wrote in your recent New York Times guest essay, and this is also a core theme of super agency, which you mentioned a few times, that AI can enhance individual agency. And before we talk about how it can do this, what does it mean to enhance individual agency for you? um you know as you know just referred to you know um brackman's work and a bunch of philosophical work on agency there's a whole bunch of different you know kind of theories here so rather than getting into a deep philosophical discussion i'll just kind of be simple which is i think agency is the question of of being kind of as a one hand self-possessed um like being able to make choices for yourself, being able to implement things in the world, to be able to kind of execute within a certain thing, and then capabilities, being able to do stuff.

1:03:12And part of the reason why I think AI is very positive on both of these things is one is, and obviously people worry about AIs that are intensely manipulative or something else, you know, not to say there's zero worries there, but the question is like, for example, and AI can easily help you become self-possessed. You know, we were talking about AI companions and therapists, but like help you understand your own mind, your own interest, to educate you in various ways, to be your advocate. I mean, we were talking about a medical assistant earlier. Well, what also about a legal assistant for everybody?

1:03:50You know, that's available free. Obviously, you know, still the people who can pay for better legal assistance always have some edge, but still valuable in having one. which a lot of people don't have. Just curious. I know this is digressing again, but where would you see a legal assistant helping the average person in the way that a medical assistant would? Well, I mean, look, most people can't afford lawyers. So like, for example, they sign rental contracts without knowing what's in them. They don't know when they have an ability to take a legal action or not or what the kind of things to do are.

1:04:30They don't necessarily know what their rights are. If a collection agent is talking to them, there's all kinds of things that go on there. Okay. Actually, since South Park is very topical right now, there was this South Park episode many years ago, I think called like Human Centipad, where people, anybody who signed like the iTunes agreement unknowingly was committing themselves to be part of a human centipede, if you remember that horror movie. And I can see this would be a great case where you have the AI legal assistant looking over your shoulder to just skim through the terms and conditions of everything you sign.

1:05:14So, yes, exactly. And look, there's just all kinds of places. There's a little, like, part of why we have laws is we live in a society where we have, you know, obligations, responsibilities, entailments with each other, with society and so forth. And part of why we have lawyers to help navigate that. And this becomes a much easier way to navigate all that. Anyway, one part of that is kind of this knowledge of, you know, what are my available options? What are my rights? How am I self-possessed? And so forth. And AI can help with that. And then there's capabilities in terms of doing stuff. Like, say, for example, we're on a legal theme.

1:05:49I want to write a letter to say, hey, you know, you did me wrong, you know, legally. Well, I can help you write the letter, right? You know, and kind of cross check that and so forth. And so that extends capabilities. And so I think that that's part of the individual agency. And obviously, you know, that was the specific question. And then the super agency in part, just so people think about it, is to not only think about individual agency in terms of individuals, but also what happens with our agency when we all get it together or when a lot of us get it together. And part of super agency is the thesis that when we have a broad based technology for many of us, that all of our agencies go up, not just because of our own individual capabilities with it, because your capabilities with it.

1:06:38And so with like cars, you know, created super agency as well. And I think AI has that capability of kind of raising all of our agencies, not just by us individually, but what happens when many of us are all raised together. And a very last point, people say, well, but isn't AI going to differentially help the wealthier or someone else? And the answer is maybe yes, but it's also like a smartphone. And Tim Cook has the same smartphone, you know, that some of your Uber drivers do. Because when it's that broadly provisioned, we all get the same one. There isn't actually, in fact, any particular benefit to give one person a much better one because it's much better to be for the hundreds of millions and the billions for doing this.

1:07:33And I think that's part of the increase of agency we'll be having with super agency. And part of this New York Times guest essay, though, is that AI will enhance individual agency by always watching us. And I think this also connects with the digital twin experiment where your digital twin kind of knows everything you know. And how does AI, by knowing so much about us, amplify agency? Well, it's again the, you know, kind of broadly the two elements I just mentioned, at least the two elements I just mentioned, which is one, you know, kind of helping you know yourself and what kinds of things you might want and might be available to you.

1:08:23And then the one is helping get them, helping move to that. And by knowing you, it can both suggest the first and be a good companion, catalyst in here are the things that you might want and might consider. And then and then in doing them and in finding your way to them and making that happen. You know, part of the, like, it's a very simple example. I tend to use deep research from AI every day. And part of that kind of deep research is because I'm learning and doing things. And that enablement to do that, to say, hey, I've got this question around this, you know, particular project that I'm working on.

1:09:12Well, you know, part of what AI deep research gives you is a pretty amazing, very quick research assistant to help you sort through things. For the time we have remaining, I really want to talk about employment and economics and AI. And this is particularly for me a major issue. So we've talked a bit about academia. um i'm an academic at this point i'm getting a phd here at stanford and we've talked about law i'm starting law school here in a couple of weeks too and once upon a time i i toyed with the idea of going to med school and being a psychiatrist and uh that's still i mean there are still pathways to becoming one.

1:10:05And all of these things appear like they can be displaced by AI. So could we just to satisfy my curiosity, I'm curious what your thoughts are about how academia will change in the years to come because of AI? Well, the short answer is massively, although I think academics is a very, you know part of it's being a you know uh university professors being a centuries-long institutions it's one of the you know the oldest things that exist and along with cities and religions and whatnot they tend to be very resistant to change and part of it tend to be like no we're going to keep doing the things we're doing now and please be brutal i've kind of come to the conclusion that this is not a viable career pathway anymore yeah well it's a very it's not It's not viable the way it is right now.

1:11:02It'll change massively.

1:11:07I think the old industrial paths of, for example, the way that we teach is we hand out some books. People come to lectures. They submit written essays and they do written exams. I think that's all not long for this path. I don't know if it's small end years or very resistant institutions longer, but, you know, obviously the ability for AI to do very intense one-on-one education is extremely important. The ability for AI to do one-on-one assessment is extremely, extremely good. Obviously it can do, you know, AI can actually even lead seminars and all the rest. So I think that the that this will all be transformed and also obviously, you know, to whatever degree the savant super intelligence involves helping research is there, too.

1:12:07And so all of this stuff really changes, you know, kind of academics. Now, that that being said, you know, I could easily see a universe where you say, look, all individual learning, teaching and assessment is done by AI. But there's still a bunch of role for human beings in doing stuff. It's like... This episode is brought to you by Google Chrome. You think you know a browser, but Gemini and Chrome? That's new. It can help you with practically anything on the web, like restoring a vintage motorcycle from a 50-page restoration block, or finally break down that long article you've had open for weeks.

1:12:45Gemini and Chrome is here for it. Ready to make anything online make sense? There's no place like Chrome. Check responses set up required. Compatibility and availability varies 18+. This episode is brought to you by Google Health. Stop chasing someone else's definition of health. What matters is what's healthy for you. Google Health offers a new kind of coach built with Gemini for effortless tracking, sleep insights, and holistic coaching tailored to you. Visit GoogleStore.com to learn more and start a new relationship with your health. Requires Google account, Google Health app, Internet, and Google Health premium subscription.

1:13:18Features subject to change. Availability and results vary. Not intended for medical purposes. Works independently of Gemini apps. Check responses for accuracy. Facilitating seminars, seeing if the target of, you know, like what kinds of things should we be assessing for? How should the assessment be driven? You know, are people engaging the right way in the learning and teaching process? What are ways that we might prompt different options with AI for making that happen? I think there's a variety of things. I think doing research and thinking about, you know, philosophical thoughts, for example, is still a viable thing, even in quite advanced, you know, kind of superintelligent circumstances.

1:14:02But I think the notion of, you know, like the way that we've had universities be, you know, this kind of multi-century bulwark, I think AI is the most amazing learning and idea development technology we've created in history. And that's going to completely transform how they work. I wonder if rather than speaking about philosophy professors, speaking about mathematics professors might be easier in a certain way, because whether or not it really is feasible, it's it's I don't know, maybe easier to imagine AI solving. extant mathematical problems than it is to think of them solving all philosophical problems.

1:14:58And in a future where AI is just so much better than Terry Tao and any other great mathematicians, is there a role for math professors doing research at universities in the same way that they're doing today? Well, I mean, there's this interesting question. It depends a little bit on what the savant stuff does. I mean, there could very well be just like, look, in one simple form, we have more people watching chess tournaments than we've had in history. And yet, computers are the best chess players. So there could be roles even, you know, for math research and everything else, even if computers are the best math researchers.

1:15:52So, you know, there's, you know, there's many, many different ways the world could be. That being said, while there has been some good, you know, kind of,

1:16:10you know, work from Google and from others about like, you know, doing math research and math Olympiads and so forth, we're still a fair ways off from these AIs actually, in fact, you know, anywhere substantively really pushing these kind of research barriers and not just with math. I mean, I think, for example, if you go use ChatGPT5, Deep Research, you get some really amazing results, but it's still not the same thing as the thing that a highly qualified human being doing real work, you know, can get to. Now, it can accelerate that human being a great deal. So again, it depends on where these super savants, you know, get to and how broadly they get.

1:16:55And it's very possible that there's a number of different domains, one of which might be math, you know or some areas of physics or other things um but there's a there's a sorry about that um uh the um there's a number of of areas where you know kind of this um these kind of areas might actually in fact have a fairly long participation of human beings like one is by themselves. But the other is, I actually think person plus machine will actually have a much longer lifespan than people who gesture at the chess analogy will have. I mean, there was this funny brief period where there was like, okay, humans better, computers better, humans plus computers better, computers better.

1:17:52And that was a brief human plus computers. But one of the things about chess is chess is a is a kind of a rare very specified environment it's it has zero epistemological uncertainty everybody knows the same board it is a very determinist fitness function and a very limited environment to be operating in even though there's a lot of configurations and movements and i think life is different and i think that will give a much broader range for the person plus machine being a higher competence fitness function than just the machine. Speaking of this partnership and the long-lived possibility of humans and machine working together, I mean, you've told people, you've told young people that their strength in the job market would be their status as AI natives.

1:18:47And I'm wondering, like, just take me, for example, again, somebody who might be interested in law or medicine or academia, where does being an AI native fit into the future of these fields? Well, to some degree, I mean, we don't know exactly how the field will go, but part of the thing is human organizations change, you know kind of slowly and resistantly and part of being young like for example going into a law firm going i know how to use ai right there's like and i'm going to help like revitalize our our work processes what we do what targets are how we look at how you even again and and and grow customers may very well be very different you know in a ai process with law and all the rest And so the kind of this process of these things may actually, in fact, be, you know, kind of central to how this works.

1:20:03Like, I don't have a simple answer, but I think the path of the answer is by engaging with it and using the fact that you have the kind of younger, more AI generation, AI native experience to help bring about the future is then helpful to your career path, to organizations, to why people might pay you, you know, et cetera, et cetera. and that what the new career paths are, are being created. Now for me, you know, today, and also if I'm younger like you, that would be exciting versus concerning. But I do understand that people have concern about the, oh shit, we're creating the new path. I'd rather just learn it and do it.

1:20:45And it's like, yeah, we're in a creation moment. The last thread I'd like to tug on today involves taking advantage of your expertise as an entrepreneur and investor in this space. And also it's tied to economic uncertainty. But for anybody who has some capital to spare, how would you go about if you were just a retail investor, which you're not, but if you were just a retail investor, how would you go about trying to leverage these great advances in AI and AI driving a lot of future economic gains through investment? Well, I mean, I am an investor. I'm doing stuff. But look, I think part of it is when you invest in capital, you're trying to invest in things where, you know, it's like public market.

1:21:43You buy a share to dollar. You hope the share price goes up, et cetera. Part of how you choose things is the pattern of where you think technologies, products and services will be successfully delivering the future behind people who are organizations that are building them and you think they're going to be economically differentially successful. That's the abstraction of that. And I think that part of that is to like be thinking like investing is a competitive game. It's like, for example, you hire some people to be lawyers, not others. You hire some people to be doctors, not others. You know, I can't say, hey, I'm a doctor.

1:22:19You know, pay me to be a doctor. It's not just because the licensing, it just doesn't work that way. So it's like, well, where do you have a differential, like, edge in investing? Do you have access to a certain pool of talent? Do you have a better kind of judgment? Do you have a better understanding of a market? You know, all of that kind of goes into it. And part of how, you know, AI is changing all of that. And you may also have a differential use of AI as a tool for identifying, for driving, and you have something that you're doing. But you have to think about it as a competitive game. And part of it as that competitive game is like, you know, how, like, you know, I use a poker analogy frequently when training investors or getting people to think about investing is like, you know, in a poker table, if you can't spot the sucker, you are the sucker because it's a competitive game.

1:23:12And if you're the weakest, you're the person who's net contributing money and you should opt out of that poker table. You want to play at poker tables where you're the strongest person. And that's the same kind of competitive advantage question that we have in a variety of economic questions in the field. And so that's the kind of thing to be looking at, you know, given how AI is changing investment. And then you have to look at what are your competitive advantages. Well, it sounds quite broadly like the person that I am describing, just the random retail investor, they're not going to have the competitive edge when it comes to investing in AI.

1:23:55So it sounds like I'm putting words into your mouth here. You would say find some place where you have a competitive advantage. If you work in, I don't know, the textile industry, maybe cotton futures are your game. But unless you're close to Silicon Valley in some way, maybe tech isn't the right space for you to be investing in. Or if you knew the cotton industry, for example, do you understand how the AI technology wave is going to transform the cotton industry and what bets you would take or not bet because of that? Right. If you also understand AI. But yes, you know, I wouldn't walk up to some kid at Stanford and say, are you working on AI?

1:24:35Here's some money to invest. That's not a good way to do it. It's part of the reason why, you know, people put a lot of energy to being professional, like professional tech investors or professional, you know, executives or other kinds of things as ways of doing that. So it's a, and it's one of the reasons why in the private investment every now we have this notion of an accredited investor, which is, you know, private companies are not supposed to take money from people who aren't accredited investors, which are, you know, signals of a certain amount of expertise and a certain ability to lose money for doing it.

1:25:14Well, without getting into the specifics of companies, and this is your area of expertise, are there any sectors of the economy that AI will be applied to where you expect that it will have outsized economic gains? Like, I'm assuming medical research is one of them, given what we talked about with Manus.

1:25:39So, look, I think the productivity implications from AI are massive. I think that every person who uses language in a working environment will get amplified. There will be co-pilots for everything that involves use of language and information, maybe even more, but that's the minimum. And that's a broad set of things. Like you mentioned the content industry. Well, you know, the supply chain is all physical bits, but analysis of the supply chain, risk factor analysis, meetings, financial, you know, sales and marketing, you know, those are all use of language. And so I tend to think there's going to be massive productivity increases across all of the stuff.

1:26:24As a matter of fact, you know, part of your earlier legal thing, I think the whole notion of how law is going to apply is going to get much more sophisticated because we have these and it will be the whole legal activity and profession and accountability and liability will all change because of AI in doing this. And so I think the productivity increases across all this stuff is going to just be massive. And part of the way to understand it is, you know, the notion of individual work is going to be, there may still be individual human work, but individual human work working with agents as a way of playing this out.

1:27:06And so I think, again, the productivity implication is going to be massive. Now, I think it's more massive the closer you are to the world of bits, the world of knowledge, the world of information, because it's more natural. So, like, for example, you know, yes, financial analysis for the cotton industry, much slower for the physical manipulation of the world of it. Doesn't mean it doesn't bleed into that, but that's those are the kinds of transformations I think we'll see. Reid, this has been a really excellent conversation. I'm a total neophyte in these areas. So I'm really thankful that you let me pick your mind for me and the audience today.

1:27:49So thanks again. Well, my pleasure. And I look forward to when this comes out.

1:28:03When it comes to my policy, Farmers gives it to me straight, so there's no surprises. It's your coverage. You should never be confused. You hear that? An insurance provider with a conscience. We're a choir, but yes, we have that too. Straightforward coverage you can literally understand. And that's what you get when you have farmers. Bum, bum, bum, bum, bum, bum. Learn more at Farmers.com. Underwritten by Farmers Truck or Fire Insurance Exchanges or Affiliate. Products not available in every state. Shipping, billing, admin, payroll, marketing. You're managing all the things. So why waste time sending important documents the old-fashioned way?

1:28:41Mail and ship when you want, how you want with Stamps.com. Print postage on demand 24-7 and schedule pickups from your office or home. Save up to 90 % with automated rate shopping. That's why over 1 million small businesses trust Stamps.com. Go to Stamps.com and use code PODCAST to try Stamps.com risk-free for 60 days.