Podcast - The 6 Laws of Technology with Nathan Furr Part 3

Welcome to Beyond the Benchmark, the EFG podcast with Moz Afzal.

Moz Afzal:

We continue with our three part series. This is number three. If you haven't listened to the first two, please do on the six laws of Technology. With me again, I have Nathan. So Nathan, welcome back.

Nathan Furr:

Thanks for having me, Moz.

Moz Afzal:

So we've covered the four laws out of the six in the previous series, and let's go straight into number five and to get your thoughts as we wrap into the last two.

Nathan Furr:

Yeah, so just as a reminder that what are these laws? It's really the synthesis of just decades of research on the patterns by which technology changes the world. And it's try to answer this question about where and how does a leader invest in technology because it's so confusing and there's so many mixed signals out there and a lot of myths. And one of the myths that I see out there is this kind of naive and or optimistic belief that the best technology wins. That if we create better technology it will win. And if we invent technology and if we invent better technology, that we'll make a lot of money and be rewarded and create a lot of value and capture that value. And unfortunately, that is definitely a myth over and over. We see that the best technology doesn't win. It's usually something a little different than that.

Sometimes it does, but it's usually a compromise that wins. And more important, it's not the inventor who makes the money, it's somebody else. So the law we propose is that control beats invention. And this comes out of a framework we have in my field called the Profiting from Innovation Framework. And it basically asks, okay, so why is it that inventors, the people who create the technology and introduce technology often aren't the ones who benefit from it? So some silly examples from the past, but RC Cola creates the first diet soda, but what are you drinking? You're drinking Diet Coke or Diet Pepsi or something, or EMI creates the first kind of MRI machine, sorry, they create the first CAT scan. Let me be accurate. And within three years they're squeezed out of the market.

If we go back to the early automobile industry, even electric cars have been around since the beginning. It was electric, steam and gas. In fact, steam was the leader. More people had steam based cars, but a lot of people predicted that electric cars were going to be the winning tech. In fact, there was a big moment, made a lot of headlines where they built the biggest automobile factory in the world. And it wasn't a gas powered car factory, it was an electric car factory. It's called Baker Electric. And they were going to just kind of capture the whole industry with this massive investment. And of course they failed significantly. So the question was why if you invent something new, why don't you get the value of it? And there are really three core reasons or three principles to keep in mind. Number one, and perhaps the most intuitive is there intellectual property, the people who have intellectual property.

If the intellectual property regime is enforceable, then the people who hold the intellectual property tend to capture the value, but that may not be enough alone. The second factor is something we nerdy academics like to call the dominant design. And that's just to acknowledge that early on in a technology, there's usually what we call this era of ferment where everybody's experimenting with different versions of the technology. So like I talked about the automobile, is it electric? Is it steam? Is it gas? Actually, there were all kinds of experiments. If you're a automobile car buff, does a car have three doors? Does it have two doors? Does it have an open top? Does it have a closed body? Does it have three wheels? Does it have four wheels? I mean, does it steer as a lever? Does it steer with a steel? So people are making all these things and it's not the best technology that wins.

It's what we call the dominant design, which is usually the best compromise among things. And so one of the reasons why the gasoline powered car emerges dominant is because a gentleman whose name everybody knows now, but his name was Henry Ford, everybody credits Henry Ford as inventing the kind of assembly line and driving down the cost of the vehicle. And that's true, but he actually did something else that gets overlooked. And that is he invented a really robust suspension system so that his cars could go anywhere. And back then there weren't a lot of roads. So actually the reason why electric cars were so popular is because, yeah, the range wasn't great, but the roads you could use weren't great either. So the electric car was good enough for the roads you could use in town, but once you got out of town, you had these roads that were terrible, you'd had horses on 'em and buggies and they were just gets wet.

It's a mess, right? So a car just was really bad outside the city. And what the Model T introduced was this incredible suspension system that essentially made the range of this vehicle infinite. And so the dominant design that emerges, whoever creates the dominant design, they are the ones who capture a lot of value. And if you think about televisions, it's a lot of work on televisions today, and we all have televisions, but there's been a lot of competing technologies like is it LED? Is it OLED? Is it LCD? And this is a battle of dominant designs. And so number two is who has the dominant design? And then number three, and this is probably the most important, this economist at Berkeley named David Teece, he said, who has the complimentary assets? Such a long word basically means who has everything else that's needed to commercialise it?

And the person who owns the complimentary assets, they capture the value. And I think the best example of this that I love is Apple launches the iPhone. It's an incredible success. And all the other kind of manufacturers out there say, oh my gosh, we got to create something to compete with Apple. And they create this open handset initiative that Google's a part of, create this kind of new operating software, which we call Android today. And when they finish developing this new OS system, all the handset manufacturers start running off and making their own version of Android with their own store. And at that moment, Google, who was a member of this consortium, pulls their hidden card out of their pocket, their Trump card, out of their pocket and says, oh, by the way, if you want maps, Google Maps, you have to use the Google version of Android and the Google Play store.

Now to you and me, you might think like, oh, wait a minute, big deal. But actually maps are incredibly hard to make an essential killer app for the smartphone. And so all the other hand sub manufacturers, the hands were tied because Google controlled this complimentary asset that you needed to get value out of the smartphone. And if you don't believe me that this is so important, if you're an iPhone user or around iPhone users, when Apple tried to launch their own maps, just remember how people talked about it. And if you don't remember, go search on the internet. How did people like Apple Maps say, here's one of the most profitable cash rich companies in the world that introduced the smartphone that changed it, a generation of technology and Apple maps sucked. It was not very good because it's so hard. So again, what are the three factors?

It's is there IP and is it defensible? So a defensible IP. Number two, are you making the dominant design or are you able to make your design the dominant design? And we see this, by the way, think about the battle for charging infrastructure and EVs absolutely a battle for dominant design. And number three, and most importantly the big one that everybody overlooks is do you own the complimentary assets? And it's the owners of that who make the money. So the reason why, by the way, the EMI goes out of business is they don't have the sales force to really commercialise the CAT scan, but the other competitors do. And the reason why RC Cola lost control of their incredible formula for a diet of soda is they didn't have the bottling and distribution. And the reason why Google everybody uses the Google Play Store and the Google Android is because Google had the complimentary asset that everybody needed.

 

And complimentary assets can even be things like owning the customer. And I watched a lot of telcos the past decade that in telecoms there was this obsession. They said, we don't want to become a dumb pipe and we don't just want to be this infrastructure provider. And they spent tens, hundreds of millions of dollars, maybe billions, and none of them succeeded. But I think they've kind of backed down from that because they realised, oh wait, we have this really important set of complimentary assets, which is one, we control the whole infrastructure by which connectivity happens and we own the end customer. I'm still waiting for them to say what they can do with owning the end customer. I think they could probably do more, but those are really important complimentary assets. So I would say look for those three things, and of course move that into the AI world. Moz you and I have been talking about these Meta being able to do many hundreds of thousands of new processors. I mean, that's just a gargantuan complimentary asset that gives 'em a real stranglehold on the development of the next generation of AI.

Moz Afzal:

I guess media also comes to mind when you're thinking about that, and we've obviously seen that in companies like I guess Netflix, Disney, they've all tried. We've seen the telco companies obviously also try into median and horribly failed, certainly over the last, in fact, the last 20 years actually, they've tried and failed many times and they're even failing more recently, but especially around cable and media. How do you think Netflix just again, is moving far ahead compared to many of the other companies? Even Disney I guess had the complimentary asset, they're successful but haven't quite cracked it yet.

Nathan Furr:

Yeah, so I think it was really interesting to watch the emergence of streaming because initially you may remember, everybody thought it was just this kind of add-on that was just get a little bit of money. I mean, MySpace, way back in the day platform, you could play any song for free because the major music labels that, oh, who cares? And the same thing happened with the early movie streaming movie studios just gave away the streaming rights. Like, oh, great, little extra money. Who cares about that? But what has really evolved to be is a battle of who controls the content. It's all that. And a lot of players woke up to that too late. So if they didn't have a kind of critical mass of content, it wasn't as interesting. But let's just take Disney and Netflix and compare them. Listen, there's smart people at both companies.

Disney obviously has this amazing legacy, but I think one thing that characterises Netflix, we got to give Netflix credit and we talk all the time about companies being disrupted. Netflix is a company that has gone through generation after generation of technology. I mean, they used to mail DVDs. If you think about what are the complimentary assets for doing a DVD business, well, it's like logistics and shipping centres and it's trucks and it's physical analogue stuff. And now they're totally different company. They're like they do to serve you any content they have to be able to serve any content. They have to be able to instantly repurpose content to thousands of formats, whether you're on your phone, your iPad, your tv, the technical sophistication of that company is incredible. And one of the things, and this is why I love it, because one of the things that makes many of the digital first companies really challenging to compete with is they were built for scale.

So we take something and financial in China, the reason why they're so hard to compete with is they were built to use algorithms to make decisions without people. And so a traditional bank doesn't find the margins that financial gets attractive enough. It doesn't pay for the buildings and the salaries and all that kind of stuff, but it does for them because it's algorithms making decisions. I think what makes Netflix really a particularly hard competitor to beat is they've kept in mind that mindset of data, algorithms, AI, I mean, they've used AI from the beginning to understand so that now they can actually predict better than maybe anyone else what's wanted, what will be successful, what's the cost trade off. I even think since we're talking about revolutions, here's another revolution that's happened that many of us who are raised in the world of gap accounting.

So what I mean by that, financial statements, so quarterly income, profit statement, balance sheet, all that, we are all raised in that world of accounting. And how you make money is like quarterly profits, EBITDA. And by contrast, what happens in a lot of these digital startups that are so powerful, they have what I call the digital growth engine. And that is they're looking at a totally different calculation. They're looking at what's the cost to acquire a customer and how valuable is that customer to me over their lifetime? And this is the early Amazon, when people said, oh, they don't make any money, they're going to flame out. No, no, no, no, no. They knew that they could spend X amount to acquire a customer and they could multiply that X amount by 3, 4, 5 over their lifetime value. So why would I return any money to shareholders when I can take that money and turn it into one pound, turn it into three pounds, take one euro, turn it into three euros.

But why now? Why is that happening now? It's because we can measure that. We could actually measure what it costs to acquire a customer and actually measure what the lifetime value is. And so that's what Netflix has going. They can actually using algorithms say, is this show worth doing? What's our lifetime value? Customer payback? So by the way, you may wonder why is my Netflix so full of standup comedians? Well, standup comedians have incredible payback. They're really cheap to produce. A lot of people watch it. And anyway, but again, this is what we were saying, technology has come to the heart of the organisation. And so it is of course about making great content. It is of course about telling great stories. All those human things remain, but technology lets you do it better. Netflix, my guess is that, and I don't know the inside story at Disney, and maybe somebody's listening to this and call me up and tell me your beautiful story, but my guess is they're playing catch up on how do we, and they're pretty good already, but they're playing catch up to Netflix, which has been doing this from the beginning to understand what do people want, what factors predict the impact? How do I serve that to you most? So you watch that next show and you don't cancel.

Moz Afzal:

No, absolutely. I keep on hitting the like button whenever I like something. So let's move to the final law of tech number six, Nathan.

Nathan Furr:

Yeah, so the last one is I think one of the myths we have out there for lack of saying it better is that we know what technology will do, that we can predict technology, and actually we can't. It's incredibly unpredictable and we never know how it's really going to evolve and what's really going to happen. And my favourite example of this, and it may be well known to some of your audience members, but not to most people, which is the great manure crisis of London of the 1890s. So there was this big conference that they had because really London had become very congested and there were 50,000 horses on the street of London every day, and they left about 2 million pounds of poop on the street every day. And they predicted as the city grows where you're going to need more horses, but they weren't going to need more horses to remove these 2 million pounds of manure becomes 3 million pounds, it becomes 5 million pounds.

But those horses are also leading manure. And so it becomes like they predicted in 50 years the streets of London would be covered in nine feet of horse poop. So how are they going to solve this untractable problem? And of course, we all know what happened, which is the automobile, but the automobile itself as a technology really changed the world in ways we couldn't predict. For example, it created suburbs. So the value proposition of a city had to fundamentally change in some ways. And a great example of that is the city of Melbourne, the city of Melbourne was this really in Australia, was this really vibrant city in the 1860s, seventies, eighties, nineties. By 1970 because of the automobile, it became an empty husk downtown. I mean, there were less than 600 occupied apartments in downtown Melbourne. There was a lot of crime people just fled to the suburbs, and it was just a massive problem of urban decline.

And what changed, by the way, if you go to Melbourne, it's one of the always rated, one of the most livable cities in the world is they had to reinvent what the value proposition of a city. And so they said, let's create mixed use. So shops on the bottom, apartments on top, let's beautify it. Anyway, that's a whole story. I could tell you for the sake of time, I won't. But let me just say technology evolves. And so the law here, it's its survival of the agile. Now we like Darwin's survival of the fittest, but many people think fittest means strongest. It was actually the most adaptable. So we're trying to highlight the principle because technology evolves in ways we can't expect, but there are some predictable pieces in it. And one of the most predictable pieces is that there are these bottlenecks that occur as the technology evolves.

And there are different kinds of bottlenecks as the technology evolves and they require different kinds of solutions. If you know what they are, you can actually make smart decisions and they tend to go in an order. So the first kind is just, we call it a technology bottleneck. And I've written about this and it's, is it good enough? That's the core question. Is it good enough? Can we even make it? And then it tends to become a capacity bottleneck, which is there enough of good enough, and then it becomes a control bottleneck who owns good enough? And I think the best illustration I could give of this is the solar photovoltaic industry. So I was at Sanford when everybody in the investing world, especially venture capital was saying that Cleantech would be the third pillar of venture capital. So it was it, and then it was biotech, and then it would be Cleantech.

And really smart venture capitalists with smart LPs were pouring tonnes of money into these thin film companies that promised to produce radically better solar panels that would basically give us electricity for free. And then I watched all that money evaporate and those companies die. And it was sorry about, I watched all those companies evaporate and those companies die. And it was a question about why did that happen? And it actually can be very easily explained and obviously explained by bottlenecks. So basically what happened is when the solar industry starts to take off in the two thousands, suddenly there's a massive demand for the silicon you use to make it. And the price of raw silicon goes up nine x. And in that time, of course, that makes a silicon based panel really expensive, and it seems like it's staying high. And so all these startups get founded based on this alternate thin film technology that promises all these wonderful benefits.

And the price of silicon is high and high, and it looks like it'll stay that way forever, but it doesn't take half a brain to realise what's going to happen. Number one, raw silicon is expensive. Oh, a bunch of people enter to produce raw silicon. Well, why isn't the price dropping immediately? Because it takes seven years to get the factory online. And by the way, in the meantime, every inventor in that whole ecosystem is saying, how do we make silicon thinner? How do we make it faster? How do we use less? So like clockwork? Seven years later, five to seven years later, these polys silicon manufacturing facilities start to come in online. The price of raw silicon drops. The price of crystalline based silicon modules just collapses to this incredible level, both because silicon is cheaper, but also because they've gotten so good at making it thinner and faster and all that.

And all that venture capital money just evaporates, and most of those 10 film companies die. And it was so sad. And again, it was thin film is a better technology in the long run, but it never had a chance to really commercialise and get to scale. And it was just a story of bottlenecks. And it may sound like an old story, but it's exactly what's happening right now in AI. So one of the reasons why these large language models that are underneath gen AI, and sometimes we call them foundational models, is such a big deal. I mean, it is a big deal that everybody's downloading it and using it and all that. But I think the really big deal, the really big deal is that it resolves a really crucial bottleneck. So in the past, if you wanted to use AI, again, we've had three waves of AI.

This is the third wave. You had to do a lot of work to develop the algorithms to predict accurately. So for example, I know a gentleman who led the transformation of one of Europe's largest hospitals, and they had 26,000 different algorithms that they had trained up to predict which patients were in trouble and to save their lives. It was beautiful, amazing, but why did they need 26,000 algorithms? Because they needed different algorithms to predict all the little corner cases and all the exceptions and all the things. And so that was, as he said, it was so much work to make those 26,000 algorithms work. And he said, what these foundational models promise is to move away from these more expert system type approaches where we essentially make rules if then rules, to instead take all the knowledge that we have, tokenize it, make it essentially correlations so that that model builds itself.

And so if you look at, there's been this fundamental shift inside of Tesla and how they do AI from a kind of clunky approach, but it was still advanced of, well, how do we build the rules to make a good self-driving car? If you see a stop sign, stop, if you see a child stop to instead using this now foundational model approach, just give it all the footage we have and let it infer what are the right rules. And so what you see is in this new system, all of a sudden cars slow down when they get near pedestrians, but they slow down more when they near children. Why? Because you and I, when we drive, we slow down when there's pedestrian, but we slow down even more when there's a pedestrian. So it's just inferring the rules. And that is just, that's an algorithm approach, the neural network approach being developed in these foundational models. And so that's why I really think the deeper transformational moment we're having, and I don't want to oversell it, it's still a journey to develop this, but that's the promise of what's so exciting about this newest development.

Moz Afzal:

So just taking that example. So obviously with self-driving, obviously Waymo's one of the first has been out actually for a very long time. I was looking watching a movie the other day, and they were showing it, gosh, it must have been like 10 years ago or 12 years ago even. But in terms of that evolution starting later in this case, was that better than starting the beginning and persevering with a old rules-based system versus a naturally evolving system?

Nathan Furr:

Moz, what a great question, and this is why I love it, because it points to the law survival of the actual. So in general, the first mover in an industry does not usually become the winner, and it's because they bear all these pioneering costs and those who come in later can benefit from all the lessons of those who have gone before. But the exception to that is those who are incredibly agile and are willing to kind of, okay, because by the way, we have this concept in my nerdy world we call absorbative capacity. It's that no individual organisation can absorb all the lessons and information at once. There's this, again, time compression, diseconomies of learning. It just means you can't learn. There's a limit to how fast you can learn. So if you can take all you learned, and then when you see this bottleneck resolve, that's what the foundational models did is they resolve this fundamental bottleneck around expert systems that it takes an army of people labelling data, creating these if then rules, even if you use some great machine learning algorithms, it still takes a lot of effort.

It has the promise to break that bottleneck and make it just correlational, teach itself kind of approach. So yeah, I would say in general, first movers don't always win, but if they're agile, and this is why, again, I like everyone else fine Musk and incredibly, he is a polarising individual, but there's definitely a great deal of intelligence in saying, okay, we invested all this effort into how we did AI at Tesla, now we're going to flip it and we're going to use this new modelling approach that's agile. And so it's the survival, the agile and Waymo uses neural networks as well to do that. And they've been doing that for a long time. So again, neural networks aren't new. I want to be clear. They've been around since the sixties. It's just that we're now just at this tipping point where we can consume these vast quantities of data to really create these kind of really deep correlational networks. So that's the power.

Moz Afzal:

It reminds me of actually my undergraduate thesis a very long time ago, I won’t say which year, but a long time ago, we actually built a neural network to predict FX movements. And it didn't really work, but it certainly gave me some foundation in exactly what neural network actually could do. And that, oh gosh, that was a long time ago.

Nathan Furr:

Yeah, but I mean, let's be honest, like what data did you have to use? I mean, you had some data, but if you had, and this is also by the way, the limit to foundational models is so you've consumed all the texts that humans created that's on the internet, great, but what's next? And so that's why you see them pushing for video for photos. And so there are limits to this too, but my guess is if you could do that today, you probably would've had a model that would've done something kind of interesting.

Moz Afzal:

Yeah, yeah, absolutely. So thanks Nathan for taking us those laws of tech. So thanks very much for listening into this series. I hope you've enjoyed it. Please do send us feedback, and certainly we'll be talking to Nathan again soon. So thank you very much.

Nathan Furr:

Great to spend time with you.

 

This podcast is provided for general information only and assumes a certain level of knowledge of financial markets. It's provided for informational purposes only and should not be considered as an offer investment recommendation or solicitation to deal in any of the investments or products mentioned herein and does not constitute investment research. The views in this podcast are those of the contributors at the time of publication and do not necessarily reflect those of EFG, international or new capital. The companies discussed in this podcast have been selected for illustrative purposes only, or to demonstrate our investment management style and not as an investment recommendation or indication of their future performance. The value of investments and the income from them can go down as well as up, and investors may get back less than the amount invested. Past performance is not a guide to future returns, return projections or estimates, and provides no guarantee of future results.