The AI Bubble Thesis with Doug Clinton

[00:00:00] Welcome to Excess Returns where we focus on what works over the long term in the markets. Join us as we talk about the strategies and tactics that can help you become a better long-term investor.

[00:00:30] We get Doug's take on where we are in the current AI driven cycle and the opportunities for investors. In the back half of the conversation, we talk to Doug about building investment strategies using chat GPT and other AI engines and the results of those strategies so far.

[00:00:46] As always, thank you for listening. Please enjoy this discussion with DeepWaters, Doug Clinton.

[00:00:50] Hi, Doug. Thank you very much for joining us again.

[00:00:53] Great to be back guys. Good to see you.

[00:00:55] We wanted to have you on to talk about technology, of course, but I think we'll get into some discussion around AI, maybe the opportunities that exist today and also sort of the constant evolution or sort of understanding technology cycles.

[00:01:17] I think what investors can learn from past history and your experience, I think in the technology space in general.

[00:01:27] And then if we have some time, we want to take the opportunity at the end to talk to you about some of the indexes that you're building using artificial intelligence.

[00:01:36] And then we want to talk about the features that are interested in sort of how the AI is being used to construct investment strategies, try to stick around because that's going to be on the back half if we have time.

[00:01:46] So really appreciate it. You're always so gracious with your time and it's been super thoughtful on a lot of these topics.

[00:01:52] I wanted to start with you and just reference a quote and I'm not going to read it because it's way too long but it's on your ex handle.

[00:02:02] I think you have it pinned and it's this quote from the engines that move markets.

[00:02:07] And it's about understanding the cycles of technology sort of boom and bus.

[00:02:12] So I think that's a good place to start could you just kind of walk us through what what that quote is trying to express?

[00:02:19] I can and for anybody who is interested in sort of like understanding the normal pattern of technologically driven booms engines is a great book.

[00:02:29] It's written by Alistair Naren, who is a UK based investor.

[00:02:34] But the general idea of the quote is that these boom to bus cycles driven by technology all kind of follow this generally same pattern with investors.

[00:02:45] First you have pretty much broad skepticism like everybody is if they're not ignoring it, they actually disbelieved that this new technology is ever going to mount anything.

[00:02:56] Then you sort of get this phase where you have some early optimists maybe some technological breakthroughs happen and some people in the market start to see there's opportunity there.

[00:03:06] That's probably where we are right now with AI maybe in the middle to tail end of this early optimist phase.

[00:03:12] Then you just get broad right market optimism everybody starts to believe like you see the light bulb you see electricity.

[00:03:19] We're not there yet with AI, but I think there will be a moment where everybody all of a sudden just graphs that you know this is going to be profound.

[00:03:27] And then once you get that broad optimism, you get insanity. That's when you get the mania. That's when you get kind of blow off top when you look at the charts in 2000, the bubble and then of course you get the busts so that's that's the genesis or sort of the meat of the quote is you go from skepticism a little bit of optimism.

[00:03:45] You know everybody gets optimistic to optimistic and then bust.

[00:03:50] How do you think someone can better learn to become an early optimist because to me like I'm thinking my own experience like with some crypto stocks and even me trying to use chat GPT in my daily sort of work life and production and what I do.

[00:04:08] Like I feel like I have always tried to like embrace like those technologies like when I see them.

[00:04:14] But I'm curious from your perspective, I mean you guys have your finger on the pulse like a lot of these new technology I think before you just you know we're showing us the apple vision probe thing that you have on the side of your desk and I think like party your job is you know being one of those early optimists but if you know taking yourself out of your professional role.

[00:04:34] How do you think investors can better become early optimists and look at this kind of stuff.

[00:04:39] I think it can be hard like I actually think like take a step back from that broader question I actually think people generally fall in one of two categories like they're either natural optimists or there's sort of natural skeptics or pessimists.

[00:04:54] And what I found hard is I actually think naturally I'm more of the skeptic or the pessimist in many cases which is probably an odd combination for a tech investor who I think most tech investors happen to be sort of natural optimists.

[00:05:07] So for me I try to do exactly what you're asking about Justin by being curious just just going out and you know like you mentioned with the vision pro you know it was if we have video it is sitting on my desk I was using it a little bit yesterday.

[00:05:23] Just going out and like getting the new devices in the case of AI going out and I try every day almost like exercise to go and use some of the chat interfaces and just.

[00:05:36] Use use the tools and see what they can do tried to make them do different things that are crazy I had one experiment I ran with my wife actually where we designed a handbag company which is random we just thought it'd be fun but we had mid journey make and design kind of a cool looking handbag.

[00:05:52] And then we had chat GPT like create the name and the brand and the story behind the handbag so just go and and try stuff I think that's the best way to be optimistic and then what I think you find is when you actually go and use the technology instead of reading like second hand what someone else is saying about it.

[00:06:11] You can figure out where you think stuff is real and not real so we were talking before the podcast I'm a little bit skeptical of the vision pro still like I had to hear it's out of my desk for two weeks unused is just not a killer use case for me yet.

[00:06:25] And I think it's a really difficult thing for people to put on headsets which we can talk about more if you guys want.

[00:06:31] So that one I'm skeptical of but AI and you mentioned kind of investing with AI earlier that's something I've been experimenting with a lot trying to really push the boundaries and that's something I'm super optimistic on just from that experience.

[00:06:44] Going back to the sort of cycle with technology companies and that engines that moves market what where do you think we are with AI currently like what ending would you say.

[00:06:57] I think we're probably around eating four ish something like that three to five would be my my gut take right now I think we're still we use that framework from from engines that we.

[00:07:09] List data before something around that like early optimist cycle I don't think we're in broad market optimism yet I know some people do think so I think the skeptics think we're like way further down.

[00:07:20] The line then we really are but I mean chat you be tea I think last I saw it's like they have a hundred million weekly active users it's really not that many in the scale of things where the internet right I mean we've got four five billion people who use the internet every day.

[00:07:35] So the penetration of AI like AI's most important product is still super low and so I think we have a long way to go before we get to and I think we will get there by the way you know a mania driven by AI but I think it's probably still three to five years away.

[00:07:51] And just to get at the overall premise here to before I ask about a couple more quotes you had in there you think this could has the potential to be significantly bigger than the dot com bubble correct.

[00:08:00] I do think so because right now the internet was connecting all of us really for the first time ever and I think that interconnected nature helped some of the spike we saw then now we're already all connected we're even more connected than we were twenty twenty five years ago and actually think AI in terms of the value it can create for the world is bigger than what the old the internet ultimately did so I think that that necessitates we could have a bigger by.

[00:08:30] So the global do you have any like if we think about like some of the mind blowing things that this might do like in terms of our everyday lives do you have anything up top of your head you can think of of like the uses of this then maybe someone like me who's not in it or not realizing.

[00:08:44] I'll give you to what's coming I think in the near term is this concept of AI agents so like right now we just have these chat bots that are pretty straightforward you can ask a question you can get an answer.

[00:08:56] In the future we will have bots that if you say you know book travel for me to go on vacation to San Diego the bot will understand your preferences will understand your calendar it'll know your credit card information and it will be able to perform actions for you in the real world

[00:09:16] and ultimately I think they'll be able to perform actions in the physical world hail an uber for me make sure my ubers on time tomorrow morning to take me to airport things like that.

[00:09:24] So I think that's the next evolution that people will feel and sort of it will impact their daily lives.

[00:09:30] The other thing like I mentioned earlier that I'm super excited about is using AI to actually make investments and I know we'll talk a little bit more about that later but I think we're going to see a world in the near future where AI is acting as a portfolio manager acting as a whole.

[00:09:46] There's a stock picker and I feel very confident that AI can beat markets now and I think it will have a good chance of continuing to do that as the models get smarter.

[00:09:56] You had another great quote in your tweet here from Mark Twain it is the changes that tell you how to pilot the boat not stasis.

[00:10:03] And I was wondering if you talk a little bit about how you think this situation might be different than some other things we've seen with tech booms in the past.

[00:10:12] What I think is probably the biggest difference here especially if I compare it to the internet is the infrastructure needs.

[00:10:18] So I know people like to use kind of parallel Cisco and the network build out in 2000 versus Nvidia today and sort of the compute build out.

[00:10:28] But the thing I think is different is that with the AI boom like it is wholly dependent right now and I think it will continue to be wholly dependent on.

[00:10:38] The amount of compute that's available so I mean it essentially like compute is equivalent to intelligence the more compute we have the more machine intelligence we have.

[00:10:47] And until something changes in that paradigm until somebody creates a model that you know maybe is more efficient and uses data in a different way.

[00:10:55] And obviously there are people who are trying to figure that out until somebody does I think we just have this natural and ever increasing demand for compute.

[00:11:06] And so I actually think as long as that happens as long as that stays persistent this infrastructure build out may last longer than people really realize.

[00:11:15] So I think that's the biggest thing that's different is that we do have this specific product right which is compute and all of the network infrastructure that goes around it.

[00:11:27] That we really need and you could almost say there's like no practical limit to how much of that we might want because more machine intelligence is good I think for the world.

[00:11:40] It's interesting on that Nvidia versus Cisco thing there is a chart going around Twitter recently on this idea of like where we are you know relative to the tech or the dot com bubble with them.

[00:11:49] And if you look at the chart and I love like overlaying these charts from different time periods I don't think it's the greatest thing but it just did give you an idea like if you look at the chart Nvidia is like at the beginning of Cisco is a cent in terms of like.

[00:12:00] Meering the charts with each other and also Nvidia doesn't trade anything close to the valuation Cisco traded at at the top so it just says for like value guys like me you say well this things over you know we're in the night dinner or whatever.

[00:12:10] This could go a long way to even be similar to what the dot com bubble was.

[00:12:15] It's really hard I mean you look at Nvidia's chart and you go whether it's year to date I mean it's up almost 100% you pull back to a full year it's up almost 300% I think.

[00:12:26] And any rational human being would say wow that is scary to like buy it right now what's funny is as I've mentioned like I've kind of experimented using AI to make investments in one of the funds that I'm testing with AI.

[00:12:43] It's top large cap pick has been in video for like five months straight and I think what's telling for me when I think about that like why is AI so excited about Nvidia and why has it been so excited because AI was just looking at the fundamental realities of the business the qualitative realities of the business.

[00:13:01] And looking at the stock chart it actually doesn't have any comprehension of what a stock chart is.

[00:13:06] It could describe what one is to you but like it's purposefully I sort of don't let it have that data and it was just saying look on the fundamentals and if it grows like people think it's going to grow.

[00:13:16] You should own Nvidia and AI has been dead right so I do think there's there's something in that where when you extract or you you pull away some of the things that make humans battered investing emotion in particular.

[00:13:28] AI has an advantage there that could be hard to hard to sum out.

[00:13:32] Yeah and to your point like if you look at the actual like sales growth of Nvidia it's almost unprecedented for a company of that size and I think somebody was doing it the other day then maybe you can hit with one or something but to grow that fast on that size is pretty crazy.

[00:13:44] Yeah, it's a man and it's followed you know the stock chart is kind of follow the sales growth chart and ultimately the bottom graph line growth as well.

[00:13:52] You had another quote which I thought was great this this idea about genius and luck is something I've thought a lot about and you have this quote from the mind of Napoleon by J. Christopher Harold it says it takes genius to be lucky.

[00:14:03] And I was wondering if you talk a little bit about that the relationship between genius and luck and how it related to this tweet on tech blooms you were talking about.

[00:14:11] I think the thing with genius and luck is and I think the point that Napoleon was trying to make is.

[00:14:19] Like when when you're a genius you kind of put yourself into positions to be lucky and I think other people have said that probably after him and before him to timeless lessons often are kind of retold in different ways.

[00:14:32] And I think it's a pertains to investing around technological sort of innovations and booms.

[00:14:39] You know the it doesn't take a genius to really say okay like AI is going to be this really big thing.

[00:14:45] It takes a little bit more intelligence to go deep on it and say well what are the components that are super important what are what are the companies that are really going to benefit from this.

[00:14:54] And then just placing yourself ahead of that even if you're too early.

[00:14:58] You know then I think you start to invite the potential for luck by actually being in the game actually investing maybe building companies in the case of some entrepreneurs and then anything can happen luck can take hold.

[00:15:12] And I think when you're in the right technological paradigms usually it's good luck not bad luck that kind of grabs you.

[00:15:19] Have you ever heard about this idea of like serendipity versus luck.

[00:15:23] No it's interesting it plays into exactly what you're saying it's something Jason's why talked about.

[00:15:27] Like on Patrick O'Shawnessy's podcast like a long long time ago.

[00:15:30] But the idea is like luck is it if I just find a loyal in my backyard or something like that's luck.

[00:15:35] But if I'm out meeting with people and you know learning about my industry and doing all the right things and I just happen to come across the right person at the right time that advances my career.

[00:15:44] Then that's serendipity because I put myself in position to get the luck versus just sitting here and hoping the luck would come to me.

[00:15:50] So I think that fits pretty well with what you're talking about.

[00:15:53] I think it's a great a great way to put it you know and I think the more back to the idea of curiosity too.

[00:15:59] I'll pull in that idea as well.

[00:16:00] I think luck can also be somewhat a byproduct of curiosity or maybe serendipity is the better word for that.

[00:16:08] But I think the more curious you are as an investor and the more you play with new things in the world and try new products whether it's tech products or retail products or soft drinks.

[00:16:19] I mean we're invested in Celsius it's a company and our core tight and fund.

[00:16:24] Just go out and be curious and try new things and I think the more things you try and the better of a barometer you get of like okay this is really different.

[00:16:32] This is special this is real.

[00:16:34] I think that can help make you a better investor.

[00:16:37] Yeah especially in a world where we all get pigeonholed you know I'm either a value investor or growth investor.

[00:16:41] I'm on this side of the political spectrum or the other like we tend to I think try new things less or go outside of our comfort zone less.

[00:16:48] So I think that's really really important in the world we're in today.

[00:16:51] Absolutely 100% agree.

[00:16:53] So this idea of the difference between bull markets and bubbles you have a whole blog post on this and this is something that people have kind of been talking about with what's going on they are like are we in a bull market are we in a bubble how do we know the difference.

[00:17:04] And you came up with four criteria that that outline what the differences are and I wanted to work through those individually and talk about them and the first thing you had here was this idea of evolution.

[00:17:14] So how is the evolution different on a bubble than on a bull market.

[00:17:18] Yeah.

[00:17:19] The most striking way is when you're in a bull market I think people all they talk about all the way up is valuation you climb the wall of worry as the saying goes and then when you get to a bubble when you evolve to that because all bubbles have to start as bull markets.

[00:17:36] When you get to the bubble you just stop talking about valuation altogether or you just start using preposterous metrics I mean people joke about you know user value or you know the value of a minute spent online or something like that like you extract so far away from free cashflow which I think is ultimately probably the best way to value any business.

[00:17:56] But you know I think that's what happens in terms of the evolution you worry about valuation forever and then you just say who cares like let's go it's bubble time that's when things have gone pretty far.

[00:18:09] How about timing that was your second one what's the difference with timing between a bull market and a bubble.

[00:18:14] For a bull market I think they can last years and years and years so you know you can probably pulling in that idea of valuation you're going to feel like you're too late pretty often pretty frequently and the reality is the bull market is probably just getting started and oh we've also been in this period where obviously we've had generally down sloping interest rates for like 40 years plus a lot of people have pointed that out.

[00:18:38] And so we've had a few strings here of very long bull markets but even historically I think bull markets 10 to last years bubbles don't I mean bubbles don't last for years and years and years like if you look at the dot com bubble I would argue the bubble was only like 18 months you know maybe two years 1998 to when it burst in 2000.

[00:19:01] But for you know 94 all the way up to 2000 that was the bull market that led into the bubble and then ultimately the bust.

[00:19:10] How about psychology this is something that's interesting because I would definitely think there's major differences in psychology between a bull market and a bubble but if you ask me to explain it I probably couldn't explain it that well so what are the psychology differences.

[00:19:21] I think that ties actually to kind of the first point of like the evolution I think the psychology in bull markets ironically is a psychology of worry you know it's like people are always thinking that things are too expensive.

[00:19:38] You know and like nothing's cheap right now and it's hard to put money to work even though the market just keeps going up and so it just keeps pulling more people in the psychology of bubbles is yellow I mean it's it's not that.

[00:19:51] It's 2020 it's 2021 it's crypto and it's this feeling that like things will never go down and it really pulls the broader market in and I think that that is the defining kind of difference in psychology between bull markets and bubbles is that bubbles are a very sort of broad almost like global phenomena they pulled in every investor you could possibly get interested in whatever the thing is that's really hot.

[00:20:18] Which is definition is like okay you've got the last incremental buyer and then all you can do is kind of go down there and then more buyers run as asset prices decline so I think that's the psychological thing is that transition from worry to just euphoria that's a pretty strong telltale sign you've gone from bull bubble yeah when everybody else stops worrying that's sometimes the time to start worrying.

[00:20:40] The whole be fearful when others are you know greedy when others appear for one other's greedy the last one is in I could kind of guess this to the ending is probably very different for bubbles and that it's a lot worse that would assume but what did you learn around that like how do they the endings differ between bull markets and bubbles.

[00:20:58] If you look at it what you generally get is like bull markets can kind of end in these like longer like painful more painful rundowns I think or at least they feel drawn out and maybe more painful and ultimately like they get to a point that I think offers sort of like modest opportunities you know you can finally buy assets for a reasonable price when you get through a traditional let's say bear market.

[00:21:25] And so it's just kind of like this long right up and then I a long ride down least will feel long if you're a bull on the other side with bubbles I think it's much more cute up and then much more cute down so you look at the correction or the bust really from dot com 2000 to 2000 late 2000 to I guess 2003 is roughly where the trough was if I remember I think peaky trough was like 70%

[00:21:52] and so when you have that large of a magnitude of a drawdown the thing that's different in my opinion is that a lot of companies get thrown out with the trash where you end up getting generational buying opportunities.

[00:22:08] We I don't think you get the same level of generational buying opportunity just in a turn from a bull market to a bear market so if you go back to calm I mean Amazon writers is probably one of the most generational ones and obviously they had things they did along the way they went from just selling books to selling everything they added AWS.

[00:22:28] So it's been an evolution as a company but and you throw Netflix in there too again another one that was public in the dot com era and now is one of the biggest companies in the world so trying to find when you do have the bubble burst.

[00:22:41] The real gems like the truly magnificent companies with magnificent operators both in the case of Netflix and Amazon I think that's where you get incredible generational buying opportunities and that's when you really need to be on the lookout and start buying stuff when it probably feels terrible.

[00:22:58] Because stocks just going down every day you touched on two things that I wanted to ask about next and one of this idea that the acute move up in the acute moved down it makes these things very very difficult to invest in because the timing can be very hard.

[00:23:10] You've got these huge moves very quickly and figuring out like how to invest and if it's even worth attempting to time it seems like it's a challenge so like as someone who invests in technology how do you think about investing if we do get one of these bubble type environments.

[00:23:23] I think it's a portfolio management question really in sort of a risk management question depending on how strongly you feel about where you are in the cycle.

[00:23:31] You know how much capital you would like to allocate to that type of risk giving your portfolio I think those are the two dynamics that you really have to consider and then obviously there there is a chance you're going to be wrong you might be too early you might be a little bit late you might get out a little late.

[00:23:48] But the goal in my opinion you know it's not to it's not to kind of call the top or call the bottom I think it is still within the context of this massive move like this instead of maybe a slow up and a slow down.

[00:24:00] You still want to find high quality companies you know that you think are persistent winners in whatever the new technology is so maybe some people feel differently about this but like I wouldn't go and just invest in the the S codes as they as they call them on X.

[00:24:17] And sometimes those have some of the most parabolic moves up but they also have the most parabolic moves down find really high quality companies we mentioned Amazon and Netflix quality companies with good leadership that you really think are going to be leaders for the long run and then I think if you get to a point where you found a portfolio and you can build some of that allocation of your capital into those kinds of companies.

[00:24:40] And you're right about a bubble and things start to get a little crazy you say hey like I can't really justify this price on this company anymore then you trim a little bit or you pull out entirely and say I'll just all get back in at a later date when I think it's a little bit more sane but I think that's the only way to do it is you really have to look at it in aggregate of the entire portfolio and think about how much risk you want to apply to that game.

[00:25:05] Yeah, I would think that would be one of the more challenging things like you probably have certain names you love the most and in a bubble you'd expect those names even the ones you love the most to probably get above what you might consider fair value and I would assume like handling that is probably one of the more challenging things but correctly if I'm wrong about that.

[00:25:19] The discipline to sell and things get crazy is super hard and it's like like we talked about in that framework of kind of like bull markets versus bubbles.

[00:25:29] You almost like literally have to write a note on your monitor every day to just say like remember at some point you're going to stop caring about valuation.

[00:25:37] It's like when you start to say, oh you know what? Like I thought that was expensive a couple months ago and now it's like, you know, I don't even care.

[00:25:44] That might be a sign that you should take some risk off the table and rethink some of those favorite names because it might be it might be a local top or the top.

[00:25:54] Another thing you alluded to is the idea that Amazon was one of the big winners from the previous bust but there were a lot of companies that were not.

[00:26:00] You know, I remember back in the day we had pets calm and all that kind of stuff.

[00:26:04] How do you think like investing today if we're going to get into a bubble? How do you think about like figuring out what the eventual winners would be and I'm sure that I'm sure you can't do that exactly but what types of characteristics are you looking for in terms of companies that maybe the eventual winners that end up being great companies on the backside of it.

[00:26:22] A couple things and I would also offer frameworks like there's there's usually kind of three phases, I think when you think about new technological paradigms emerging into the market.

[00:26:32] The first phase is usually some sort of an infrastructure or CAPEX build so obviously we've been living through that within video.

[00:26:39] We've been working on hyper scalars or spending tens of billions a dollars a year now on chips and infrastructure so that's kind of phase one phase two I think is sort of like a data phase and it can happen concurrently with phase one but you're sort of organizing the I mean in this case data is a component of intelligence for AI.

[00:27:00] So we're sort of organizing like the pieces and the tools in the information age to be able to extract value from that data.

[00:27:07] And then the third piece is the application layer so things that hit and consumers whether that be B to C B to B that's usually the final part and like that I think we're at the earliest parts right now of in AI you know Chatchy BT is the best example but still like we talked about earlier not that many people use it in the grand scheme of things.

[00:27:28] Within that framework what we try to do is think about okay you should construct a portfolio that has some pieces of each of those stacks where you have the most confidence and it might change a little bit depending on where you think we are in the cycle so if you were early and you had a good crystal ball you probably should have been pretty heavy and infrastructure maybe a little bit in data not a ton of applications super early on.

[00:27:51] And then I think as the trend progresses you're going to find that the data companies and then lastly the application companies will be the ones that sort of get the Nvidia bid and hopefully they have fundamentals that are starting to support that bid to.

[00:28:05] But we kind of look at it in that framework in terms of what kinds of companies and then below that then I think you have you know do they have good quality management teams do they have products that customers love and they can't get anywhere else that's a question we like to ask a lot.

[00:28:20] And when you start asking those two questions like do I really trust the management team and is this product truly unique you get through a pretty small number of companies pretty quickly and those tend to be the ones that that do pretty well over the long.

[00:28:33] Do you think a lot of the biggest winners are companies that aren't public yet I mean I know you guys do both private and public investing do you think a lot of the biggest winners will come from the private world.

[00:28:41] Right now I am I would say more bullish about what I'm seeing in terms of opportunities in the private market versus the public market particularly on sort of that data and application side so I've been spending a growing amount of time in those two places and a little bit less on infrastructure.

[00:28:58] I still think there's there's probably a good amount to go on the infrastructure side but trying to get ahead of kind of that data and application layer.

[00:29:07] I am really more excited at this point for the private companies versus public just one quick thing on that I'm wondering if like in some weird way like I think back to like the late 90s where you had all these internet companies kind of coming public and using the public markets for financing.

[00:29:27] And you know didn't end well for a lot of those firms but we've been in like this IPO drought so i'm just wondering if like from a like overall company development maturity standpoint.

[00:29:39] If that's actually going to be a like positives sort of in this environment versus what like what we went through maybe in the late 90s I mean I guess it depends on like what anymore and it's it's eventually that bubble sort of forms but i'm just you know that that idea of maybe companies being able to mature more.

[00:29:58] Because of this IPO like drought maybe maybe that's a good thing.

[00:30:03] I think it can be I think there's actually something they're just in that's interesting and it ties to 2021 oddly where we just went through like the everything bubble we went through specmania we went through crypto and meme coins meme stocks.

[00:30:23] And we just had like this period of insanity I've sort of had a thesis to your point that the next period of insanity whenever that comes and whatever drives it potentially AI.

[00:30:35] I actually wouldn't be surprised if more investors focus on higher quality names period where they say look like I got blown up by game stop you know and I got blown up by whatever random spec that you know I bought it went up 200% and then it went down 550%.

[00:30:51] Or it's not possible as negative but 98% I'll say adjust my numbers they probably felt like 550%.

[00:30:59] But I think you could see investors you could see people say I actually just want investing companies that I'm pretty sure are really high quality companies they're not these weird crazy spec things or meme stocks.

[00:31:13] And you might see and maybe you're already starting to see it a little bit you could argue with the mag 7 like people kind of piling into known quality you know they're going to have benefit from AI you know their high quality management teams that have great track records.

[00:31:29] And so I kind of think when we when we think about that IPO window opening I wouldn't be surprised if maybe we get fewer companies coming out but they do have real businesses attached to them and they have really high quality you know sort of forward prospects as early leaders or early winners in their space.

[00:31:49] I want to ask you about the counter argument to this you know I've kind of become a big believer in AI and what it's going to be so I always want to challenge that by trying to find different opinions and I'll put up this tweet I was just looking on Twitter for something and I found.

[00:32:01] This Ram al-Awalia I think is the proper pronunciation of his name but he was talking about this idea that in the dot com bubble you had a bunch of killer apps whether be email e commerce the browser you know digital music internet gaming like a lot of different things and he was kind of saying those aren't there yet.

[00:32:17] In this space it and I wonder what do you think about that he was saying I can do fine without chat GBT it's not indispensable so he's saying there's not as many indispensable things right now is there were like when the internet came how do you think about that.

[00:32:30] I think there's a couple things one is I think is a little bit of revisionist history because in 1995 you probably could have done fine without hot mail but like it did make your life better it made communication better and then by virtue and literally the nature of networks which was the real innovation of the internet those products all follow the sort of met cash law right whereas more people got on the network email became more valuable chat became more valuable a lot of products that the Internet.

[00:33:00] Made possible only mildly valuable in the beginning when like 1% of the world is using internet super valuable and 90% of the world is using the Internet.

[00:33:10] I think we're in a similar early stage now with AI and I would argue you know like AI chat GBT Claude which just put out a new model that I think is really impressive and Gemini Google I mean kind of the big three leaders on the model side the chat side they are kind of like the early net scape you know and like it wasn't like net scape

[00:33:30] was this thing that the second you use net scape for the first time it was completely indispensable and you're like I can I can never imagine life without net scape like if you had to go back to just calling people and faxing people you would have been able to figure it out.

[00:33:44] But I think with chat GBT now people are just starting to see that initial potential you know and depending on how deeply you've experimented with it and gotten creative with it.

[00:33:55] I think it might color how much you say yeah like I couldn't imagine a world without that versus kind of taking the skeptics view and say well yeah I'm sure we'd be fine without it.

[00:34:05] You probably won't be saying that in 3 to 5 years I would be my guess.

[00:34:08] I think that's such an interesting point because if we think 10 years in the future and we think what we're going to be doing with this technology and if I had to guess that right now I'm going to be completely wrong.

[00:34:15] It's going to be there's going to be some mind blowing things were doing that I never thought were possible and email was probably the same exact thing like when it came on the scene I was probably go email that that's great.

[00:34:24] And now is like an indispensable part of life so I think that's so important to put yourself back like in your shoes at the time and what did you think about these things versus you know what we know now they ended up being.

[00:34:35] Arthur C. Clark has this quote and I'm going to paraphrase it because I don't know the exact words but he basically says like if we could actually predict the future the future would probably disappoint us like if we could predict it actually because you're one jack like we are always wrong usually we underestimate like how impact it.

[00:34:53] How amazing this stuff ends up being and I think that just like you know with the smartphone you had your first blackberry I don't think anybody imagined having iPhones and then vision pros.

[00:35:06] And I think just like today with chat you be T is like crude and early as that might feel like imagine what chat you be T 10 looks like 10 years from now it's going to be like you know the original iPhone versus the iPhone 10 they won't be close.

[00:35:21] Just one more for me before we switch to the AI stock picking stuff and I was just wondering like we did podcast episode recently about the top 10 companies in the S&P 500 and when you go back historically and look at them they always changed decade to decade you know what the indispensable companies from 40 years ago you know quickly fell off that list many of them still existed but they weren't the biggest companies anymore.

[00:35:41] And I'm just wondering like part of me thinks that maybe changing now because these the Googles and the Amazon's these types of companies are so dominant.

[00:35:49] That maybe they will be you know how powerful for decades but then the other part of he says well I probably felt that way about to punt and all that stuff if I was there back in the day so I'm just wondering since you're in this text base like what do you think about that in terms of these companies ability to continue to dominate and continue to be the biggest companies in the world.

[00:36:05] I love Templeton's quote about the foremost dangerous words and investing or this time is different and then he adds that 20% of the time it actually is so actually think when we have a question like this the questions are like okay well what what is potentially different this time such that it's not just a reversion to the mean kind of story.

[00:36:25] I think what's different is we actually look and we did this calculation recently you look at the concentration of the S&P 500 the top five stocks in 2000 the top five were almost 20% of the index they were about 9% of the total earnings in the top five were the top of the end it was Microsoft Intel exon and Cisco I'm pretty sure that's right now top five they are 27% of the time.

[00:36:55] So I think that's the total index last time I checked and they're about 20% of the trailing earnings so it was almost double in 2000 when you think about the concentration earnings relative to weight in the index versus right now it's not even 0.5% versus the double versus the 2x or so 0.5.

[00:37:15] So what's different with those numbers tell me is different is that these huge mega cap tech companies have captured so much value so much more value.

[00:37:25] That's what's put them there it hasn't necessarily been the decay of those other companies I mean Intel exon Cisco these are all still really big companies multiple hundreds of billions of dollars in market cap and their businesses are bigger today than they were in 2000.

[00:37:39] So what's different is that Apple and Microsoft and Google and Amazon have created these just gigantic businesses that I think if you fast forward 10 years or 20 years from now you kind of have to ask yourself well is somebody going to create a business that's even bigger than what these guys have created because that's how they would get ahead or is the business of some of these big winners going to decay.

[00:38:00] And I would say it doesn't seem likely in most cases that the businesses are going to decay I don't really see a case where Microsoft now kind of being the lead for AI cloud compute.

[00:38:10] I don't know why they would decay Amazon is Amazon right Google with search I know there's a lot of debate there and controversy that ones maybe the one where you can say maybe open AI comes out and kind of displaces them but I think it's just hard to imagine a world where these guys just they create so much value in the world.

[00:38:28] That they would be displaced or that someone else is going to come out and create significantly more value in the way they did versus kind of the old guard 20 years ago.

[00:38:38] For the back half of the conversation we want to switch to your AI powered models because they're really cool and they're really interesting but before we get into what you're doing with them I just wanted to ask you at a high level if you have to compare AI as an investor to humans as an investor, what do you think the big advantages AI will have?

[00:38:55] The biggest advantage I see is the lack of emotion. You know we talk about and just like studying Buffett studying Munger I feel like the corpus of their writings is partly about investing but a lot of it is actually about human nature.

[00:39:12] You know and I think that as an investor learning how to do a DCF learning how to do basic fundamental investment calculation is pretty easy anybody can figure that out pretty quickly.

[00:39:27] Learning how to apply that and manage what your mind tells you to do is actually the whole job for your entire life but that's basically it and I think that's why Buffett and Munger I mean Munger to his dying day I think

[00:39:40] it's sort of consistently writing about understanding human nature of others and also yourself to be a great investor.

[00:39:47] And so what AI does that I think humans don't is they don't have to battle with that like inner dialogue that emotion that you know when the market's going up and you're not fully invested your frustrated you hope there's a pullback when the market's blowing up you're not concerned right if it's AI.

[00:40:05] And I think that's something that when I think about AI participating more and more in markets I think that's an advantage it will always have like it should always be able to weather storms whether that is continuing to stay long when things are our bull market into froth and staying the course when everybody else wants to sell I think AI will always be better at that than humans.

[00:40:30] I've been thinking a lot about what it will mean for the asset management ministry and in prepping for this I was reading a blog post you did which I didn't even know this thing about Elon Musk's process here but you kind of compared what Elon Musk did to build Tesla and the process he went through and you had like a big goal of saying like the same type of thing could be done to disrupt the black rocks of the world with AI can you talk a little bit about what must process was and how that kind of applies to that vision.

[00:40:56] Yeah, Elon put out this master plan he called it in 2006 and he kind of tongue and cheek said just between you and me but it's a blog post for everybody to read.

[00:41:07] And so what he said in the post was sort of Tesla's mission the founding mission was to accelerate the world's transition to renewable energy and I think he's been pretty consistent about that for last almost 20 years when I think about the next 20 years.

[00:41:24] And how the investment world is going to change I think AI powered investing is going to be a huge influence on that so like last 20 years it's been all about the shift from active to passive the emergence of the ETF and so next 20 years I think kind of active and passive flows will actually go to AI powered strategies because I actually think they can they can beat markets and they can beat humans.

[00:41:49] When I think about the master plan Elon's master plan was basically launch the roadster high price car prove the market then launch the model S more mass market and then launched a model three right the big mass market car.

[00:42:04] And what I think about for this idea of a black rock for AI powered investing how do you kind of inspired this transition from passive and active traditional management to AI powered management.

[00:42:15] I think you have to go out you have to prove the product just like the test the roadster did for Tesla show people that AI can persistently beat markets with smart structure around it and then start to put products into the world and use that money to launch more products and ultimately I think you will see at some point in an AI.

[00:42:33] I think you will see at some point in an AI powered version of black rock so can you talk us through you have a number of strategies that you're tracking in real time by the way these are on.

[00:42:45] Thematic or go thematic dot com and you're calling these the intelligent alpha strategies there's a number of them but just at a high level can you just talk us through what.

[00:42:57] I can't see on this page I can so and I'll give a little bit of brief history this whole intelligence alpha project started when I was playing curiosity experimented with chat GPT last summer so last June as to sort of just started to wonder what can chat GPT beat the SMB 500 right pretty simple simple question simple task and it's turned in now to I think I will say 50 total strategies that I'm tracking and so I have large cap focused.

[00:43:26] Large cap focused strategies mid cap small cap different sectors different factors you can kind of get AI to think like any kind of investor that you want.

[00:43:35] And the process I've developed is I use a three AI investment committee so it's GPT it is a clawed from an anthropic and then it's Gemini from Google and what I do is I'll say for example like let's create a large cap US equity focused strategy we want to be the SMP 500.

[00:43:55] Super basic I will get some basic data fundamental historical data about earnings revenue growth things like that and also pull some consensus data I will feed that into the AI with a prompt that describes exactly what kind of investor I want them to think like.

[00:44:15] And that's the really important part you have to give the investor a philosophy that's actually what the prompt the purpose of the prompt is because if you just say you know give me a portfolio of large cap stocks to have high growth like it basically is just a glorified screener and that's not very compelling but if you say I want you to think like war and Buffett right and here's some things that Warren Buffett thinks here's some characteristics of companies that he really likes could be qualitative or quantitative.

[00:44:42] And then you have the AI choose portfolios then you start to get varied outputs between the three a eyes that I use number one so they kind of do think differently when you give them even if it's the same prompt and data.

[00:44:54] And then I'll take those outputs and I'll put them into a single portfolio and that's the portfolio I then test against whatever benchmark I'm trying to be and so are you.

[00:45:04] Are you looking it's not a consensus based approach me you're not looking for overlap you're actually just kind of taking these and then putting them together in one portfolio.

[00:45:14] And I guess removing the duplicates obviously or maybe yeah basically way works is each of the three committee members has a third of the weight of the portfolio you know so if there isn't overlap you know if each of them have apple at 1% then apple end up at 1% in that case not.

[00:45:33] You know third of it right so you're actually the a eyes actually giving you an output it's giving you weights and then you can probably even go so far as you know give it sector constraints risk controls mean you can.

[00:45:47] You can build all that in which is which is pretty pretty amazing and how often do you are you rebalancing like monthly or squarely or what's the rebalancing frequency depends on the strategy so I have some that are more I would say.

[00:46:01] Somewhere between indexes and active strategies where they're intended to be kind of long term static more static portfolios they might only review twice a year or quarterly.

[00:46:13] And then they have other strategies that are intended to be a little bit more active that review monthly but what I don't want to do is you know like a maybe more of a traditional

[00:46:22] or a lot of people want approach where you're trading every day you're looking at stocks every day I think that game is really probably hard to play with generative AI models I think generative AI models are really good as investors rather than traders and so that's what I'm really trying to tune them as so you don't really and shouldn't have to look at the portfolios as frequently.

[00:46:42] So do you give it guard rails so do you say like I don't want a position to be more than this percent or less than this percent of the portfolio or I don't want any less than this number of stocks or more than this number of stocks.

[00:46:50] Do you let that let it figure that out itself or do you give it guard rails around what you want to do.

[00:46:55] I do give it guard rails and they sort of vary all the time to paying on the portfolio usually they pay attention to the guard rails sometimes they don't pay attention to the guard rails and that's one of the things I think that's so interesting and sort of one of the things you learn working with AI is that even when you're like clear and explicit with what you want it to do.

[00:47:14] Sometimes it will vary from that and sometimes when it varies and has what we call a hallucination sometimes that's like the most interesting thing because you can ask it about it where you just take it for what it for what it is worth.

[00:47:26] And like sometimes it will bring in stocks that don't fit for some reason but they actually end up being like the most interesting ideas you see like why would it put that in there then you start doing your own work on it and you're like, I've never heard of this stock it's really interesting for these three reasons.

[00:47:41] So the hallucinations and when it kind of disobeys you is actually the most fun part.

[00:47:46] I think the fact that you started tracking these real time was important because me on the one hand like I would want to know like if you could back test this historically what that would look like but then on the other hand it's like I would wonder would that be just a big exercise in data money.

[00:48:06] You know and so because the AI will figure out what the optimal strategy is like if you got a historical data set of fundamentals going back to like whatever 63 which is the form of French you know which is when their original data set started.

[00:48:20] But but I don't know have you thought about doing that on historical data and seeing what the numbers look like or is that not really that interesting to you so actually made a great point just like.

[00:48:31] The way that I think the way anybody would create stock strategies with generative AI I think it would actually be impossible to back test because part of the.

[00:48:45] The output is giving it the data that you have as of right now so you'd have to kind of manipulate manipulate your data back to a certain point time and depending how far back in time you'd want to go.

[00:48:58] The models the underlying models right like so Gemini etc etc have data into the future that would sort of mess up the exercise.

[00:49:09] So if you wanted to like back test back to like okay what if I submitted this data to Gemini in me know 2019 or something like that.

[00:49:19] You'd have to find a way to essentially pull out all of the training data that Google used from 2019 to 23 I think is what Gemini's current to and then have that model do it that be the only way to do it authentically so I think the answer to your question is all of these have to be done sort of bespoke in real time looking at the data.

[00:49:38] I think the models are so dynamic they're changing so much.

[00:49:43] And that is the exciting part and the moat in my opinion is like all these models are changing so much they do think differently I think at least I perceive them to be thinking differently as investors when I use them.

[00:49:56] And so different models actually are better in my view for certain kinds of investing applications than others so finding the right model for your investment use case is is part of whether you can find alpha and the long run.

[00:50:08] Are these are they all fundamental based are they doing things like reading articles about the company reading annual reports reading interviews with the CEO are they doing stuff like that to her is it pretty much based on fundamentals and numbers.

[00:50:19] Fundamentals and qualitative data that they would have in there in their training data set to the extent right again I think a chat you be T or GBT and caught I don't remember where that one's current to but I think they're all mostly current to 2023 or like mid 2023 now.

[00:50:37] So they should have some general understanding or probably have some general understanding in their training data set about you know companies reports and perhaps some CEO interviews and things like that that were publicly available that they pull in.

[00:50:52] But I'm not giving them like immediately current qualitative data I only give them immediately current quantitative data.

[00:51:00] Do you know anyone else that's doing this I don't and every time I ask other people have they heard of a whales trying this I haven't heard of anyone else who's really experimented to this depth.

[00:51:10] I've definitely seen other people try to get into pick stocks but really constructing the broader architecture to create I think products because these models are scalable to create a full suite of investment products.

[00:51:24] I haven't seen anybody else do that so yeah so your vision for this is I mean this could be the early stages odds some type of investment firm or set of actual investable products that investors could someday have access to potentially it's something we're definitely exploring.

[00:51:40] And I think it's been fun just to run the experiment again for for learning purposes I feel like I've learned so much about just AI and the potential for it back to our earlier conversation about like how do you be an optimist.

[00:51:53] That's the biggest learning so far is that we should be optimistic about this technology because it's very capable and it's only going to get better from here but yeah I think there is I think there is an audience from this I do get a lot of people who email me and just say hey like you know have you launched a product yet and we're exploring that.

[00:52:09] And do you know I'm just curious it's that you know there's not tons of performance history here obviously because you know the longest one would be coming up on a year I guess in June or something like that and some of them were launched you know later but

[00:52:22] on percentage like that and percentage like winning percentage versus on a strategy basis over its respect of benchmark how many strategies are being there are benchmarks.

[00:52:34] So if I exclude a few of the really new ones that are just like less than two months old the thing is mostly just noisy data it's about 73% last time I calculated it the strategy is ahead of its benchmark on average the strategies are winning by about 350%.

[00:52:51] Yeah that's pretty high batting average I mean you know even if you were like 50 50 with that you know it'd be pretty pretty solid so 70 well that's going to be interesting to see sort of where that takes you and deep water and you know how investors

[00:53:08] sort of embrace that the one thing you know we talk a lot about like behavior and emotions and I think that's on the one hand like it's great AI and these types of strategies they remove the emotional components I think the other side of it is you know as investors eventually invest in this stuff when they don't produce out performance having the ability to sort of understand why or

[00:53:37] like that'll be important to you know it'll be important like because you know right now a human analyst might be able to articulate what happened in a portfolio what happened in the strategy but to be honest with you probably in the future AI probably will be able to just tell you exactly why like the chat

[00:53:54] bobble be able to tell you why it underperformed and some investors you know might accept that yeah you're hitting on the big question I get when I talk to potential customers so I've done some customer discovery on this and particularly on the institutional side you know endowments they want to know the answer to that question

[00:54:14] like can I still trust this machine to perform and so you need to have the explainability behind it I think explainability is this a really important concept in AI in general but to your point Justin like it will be even more so in AI when you need to make the decision do I still trust the AI to do the right thing with my money or has it lost its edge.

[00:54:35] This has been great dog two more just quick questions one market related and then another about the future so I'm just curious what your thoughts are do you think that there's a possibility here that AI as more more companies embrace it it's sort of like a just profit margins higher like that's one thing I've kind of been thinking about like if you if you got a bump of an I don't know I'm just pulling this number out of the hat I mean margins are historically high but you know if

[00:55:05] through the through AI like you get half percentage point increase in margins and then you think about maybe a full percent I don't know but whatever it is you you know you discount all those future cash flows way in the future that's a huge amount of wealth creation

[00:55:23] sort of that can happen at the aggregate level and I'm just wondering your thoughts on that I mean could that be the upside here with the overall market that AI helps make companies more profitable and investors just we're not seeing it yet because it's hard to it's hard to believe that or see that I'm just wondering your thoughts on that.

[00:55:43] I think it will make companies more profitable and I think we have one example recently with Klarna there's some news out that they are using GPT to handle a good portion of their customer service inquiries and I think the statistics they put out or something to the effect of like G.P.T. is doing the work of almost 700 human agents so you can do some math to figure out how much how much money that might be saving a year

[00:56:10] and they were reducing times from call it several minutes to like a minute and a half in terms of resolving a query.

[00:56:18] So it's like not only could you save money but like even better I think you can make the customer experience that much more impactful that was more profound so you kind of get a double whammy and so I think we're going to see more things like that if we go back this idea of infrastructure data application

[00:56:34] we're still early on that kind of infrastructure maybe getting a little more into the data side it might still be a year or two before we see some of these applications more widespread and really impact the bottom line numbers but I do think it's coming I think it's something that investors will probably start to bake into numbers maybe more next year when we see it happening in the lot in the wild.

[00:56:54] When you think out over the next 10 years what parts of the world do you think will be most impacted by AI geographically what parts?

[00:57:01] I just mean just over in general like what not geographically like from our living our live standpoint.

[00:57:09] You know, I mean I could there's so many different you could go somewhere places so wherever you want to take it.

[00:57:15] I think I think it will be yeah 10 years out I think it'll be pervasive you know and I think what's funny about technology I don't remember who said this but like great technology it's actually like magic where you don't even realize you're using it.

[00:57:29] And so it's an odd way to answer your question but I think in 10 years like we don't think about pulling out our phone and the fact that like we're connected to the entire world we've got all the information in the world that our fingertips.

[00:57:40] And that's kind of magical like it's not this crazy experience is just now part of our lives.

[00:57:45] And so when we zoom out 10 years I know that AI will be interwoven into like everything we do right like our scheduling on this we probably would have never even emailed each other.

[00:57:56] It'll just be our AI is handling our calendar sending links to get the podcast on editing it right like we probably just literally step on camera and then that's it.

[00:58:05] That's the end of the whole interaction and it gets published to the web so I think AI will it will do so many things but we won't even realize it in 10 years will be spoiled.

[00:58:16] But I think we'll hopefully remember this moment and say that was pretty cool AI just did all that stuff for us and it's only going to get even better from there.

[00:58:25] Well Justin is the editor of this podcast I like that vision.

[00:58:28] You're already out here.

[00:58:30] Doug I think that was an excellent way to answer that question so thank you very much it's always great to have you on such good thoughts we really appreciate it.

[00:58:38] All the best thank you.

[00:58:41] Thanks guys thanks guys good to see you.

[00:58:43] This is Justin again thanks so much for tuning into this episode of excess returns you can follow Jack on Twitter at practical quant and follow me on Twitter at JJ Carbano.

[00:58:54] If you found this discussion interesting and valuable please subscribe in either iTunes or on YouTube or leave a review or a comment we appreciate.

[00:59:02] Justin Carbano and Jack forehand are principles of the capital management the opinions expressed in this podcast do not necessarily reflect the opinions of the

[00:59:09] capital no information on this podcast should be construed as investment advice securities discussed in the podcast may be holding the clients of