[Music] hello and welcome everyone to another in our series of investing talks we have a very special guest with us here today please join me in welcoming professor Andrew Lowe thank you so I want to start by thanking Saurabh for inviting me to speak to all of you today it's a real pleasure and an honor to be here at Google one of the largest and most successful companies in the world and really the home office of technology and I'm particularly grateful to all of you for being here because I know this is the the last week in August most people are on vacation so this is more people than I expected was show up I'm delighted to talk with you about my new book because it's been a labor of love for a very long time and I have at least one former student in the audience who can attest to that I've been talking about these ideas for a while but it's taken me a while to get it down on paper and just to give you a little bit of background let me just tell you you know what this book took so this book was actually due at the publishers in on April 15 2008 that's when I promised to deliver the manuscript and I'm a bit late part of the reason of course was something interesting happened that year the financial crisis hit and I thought that it might make sense to sit through that and try to understand exactly how the ideas that I wrote about in the book relate to financial markets and you know before you know it you know eight or nine years past I also want to explain that originally the book was supposed to be 80,000 words and and just to give you a sense of how many more words there are in the book it's actually now at a hundred and fifty thousand words and fortunately the publisher deserve you know they reserve the right to reject the manuscript but they didn't and I'm very grateful and there's a reason why the book took so long it's because I really wanted to reach out to a broader audience with these ideas I that it applied much more than just to financial markets this notion of adaptation of course and so I was told by my publisher that if I wanted to reach the audience that I was hoping I really couldn't use equations because he said that for every equation you put in the book you can reduce the number of readers by half and so as a case in point the previous book that I published with the same publisher a non random walk down Wall Street this is a typical page from that book and so if you use the publishers algorithm that means that the only person who bought the book was my mother because they are there tons and tons of equations so because I couldn't use equations I ended up having to use more words and that's why it was 150,000 words or 500 pages why I apologize in advance but the reason is because I really wanted to try to communicate these ideas in a non-technical fashion now the basic idea sir I mentioned had to do with this notion of market deficiency the notion that prices particularly financial market prices fully reflect all available information and so that's it one under the extreme but over the course of the last couple of decades a number of economists psychologists neuroscientists have challenged that idea that people are rational and at markets accurately reflect all available information and at the opposite end of the extreme is that people behave irrationally we all act emotionally and so this divide has been really at the heart of a lot of academic controversies for 20 or 30 years so much so that many of us who were brought up in the tradition of efficient markets and have gravitated more towards this notion of irrationality we're sort of sick of the fighting back and forth it's kind of like if you've ever experienced your parents having an argument you know and you're stuck in the middle you love both of them you just wish they would stop fighting and get along and that's really the idea behind the adaptive markets hypothesis it's a way to basically reconcile these two opposing schools of thought and so by the end of this half hour or 45 minutes I'm gonna show you that actually markets have this dual nature that they are sometimes rational sometimes emotional and the dynamics of these two kinds of competing forces are really what we ought to focus on and that's really where the adaptive market hypothesis comes in so let me just quickly summarize what adaptive market hypothesis is about basically the point is that the traditional investment thesis the framework that we use for making investments is flawed it's not wrong but it's incomplete in some very important ways and what I'm going to show you through way of some simple examples is that in stable environments the traditional investment framework makes perfect sense so 60/40 stocks and bonds investing in equities for the long run diversifying across these different asset classes the traditional ways of thinking about investments do make sense but only in that kind of an environment in an unstable environment these dynamic policies are going to have to replace the static investment policies and I'm gonna argue that the current environment that we're in is highly highly dynamic and so the traditional investment framework is going to have a lot of trouble dealing with those kinds of situations you know when politicians started talking about what really drives an election they came up with a phrase it's the economy stupid I actually think that biologists should be telling us economists that it's the environment that drives various kinds of human behavior and so I'm gonna show you how the adaptive markets hypothesis is really ideally suited to address those questions not just in finance but in a much broader framework and because I know that this is a technology audience and not necessarily a finance audience I'm gonna keep the finance actually to a minimum and talk a little bit more broadly about how different disciplines actually contribute to how we think about adaptive markets so let me first start with the basic statement of what adaptive market says and bring you through it the basic idea is that individuals react in our own self-interest law economists could tell you that right we all agree on that but at the same time humans we make mistakes we learn from those mistakes we adapt to those changing kinds of situations and ultimately it's the way that we adapt and the way that we interact with each other that determines market dynamics and so it's really evolution that drives financial market systems so that's the basic nutshell of what adaptive markets is about and you know the one the very famous biologist theodosius dobzhansky once said that nothing makes sense in biology except in the light of evolution and I'm gonna argue rather immodestly that nothing makes sense in financial markets except in the light of adaptive markets so I'm gonna try to illustrate that to you through a couple of examples and then hopefully leave plenty of time for Q&A since I suspect that'll be the most enlightening part of the discussion before I do that I want to just mention one other thing about my book which is that part of the reason it's 500 pages and 150 thousand words is that it really chronicles a personal journey that I took in arriving at adaptive markets so the first five or six chapters is really kind of a travel log of that journey so I started out as a very strong efficient markets disciple that's how I was raised that's how I was taught in graduate school that's what I believed and it took probably in my case ten years to have the combination of data statistics and market experience to beat that out of me and what I'm gonna try to reflect in my discussion and what I reflect in the book is how I got there it was a very tortuous pathway that went through evolutionary biology ecology neuroscience psychology artificial intelligence and various aspects of computer science all of those elements are part of this book which is again why it's as long as it is so I'm gonna try to borrow from different pieces of it and then show you what I think might be most relevant to this audience a technology audience because technology both plays an important role in evolution and is also a good illustration of adaptation at work so let me first start with a simple example of adaptive markets at work and to do that I'm gonna take you to a particular state in India called Kerala those of you who are from India you probably know that this is on the southwest coast of India that's about 600 kilometres of beachfront that borders the Arabian Sea and not surprisingly the fishing industry is a very important part of the Kerala economy a typical example of how fish fish are fishermen work and how the fish markets work in Kerala is that there are a number of relatively small fisherman collectives they leave in the morning to go fishing probably you know 5 or 6 a.m. before the Sun rises and so they're out in the morning and they fish for a few hours and by the time they're done they come back to the beach markets and they set up the markets at that Beach where people will come and buy fish so there are literally thousands of these fisherman boats and fish markets dotting the the coastline of Kerala and it's a very important part of the local economy very fresh at the time in the 1990s and early 2000s where this flow these photos were taken there were not a lot of refrigeration so fish had to be caught purchased and eaten that very same day now this kind of a fish market has a particular challenge and the challenge is that when you're out at sea and you catch a certain amount of fish you really can't determine which beach market to go to back in the 1990s there was no way to communicate easily with the shore and the reason that that's an important feature is that if you have a lot of fish and you bring it to one beach market and it turns out that there are lots of other people with a lot of fish then the price will be very low you won't be able to sell all your fish and because it's not possible to go to another fish market during the same day because remember there's no there's no refrigeration here a lot of the fish will spoil so when you're out at sea back in the 1990s you had a decision to make where are you going to go in terms of being able to sell the most amount of fish for the highest price and that was a challenge so in 2007 an economist by the name of Robert Jensen published a remarkable study about the kerala fish markets what he did was to collect data on the prices and quantity of fish sold every week for a period of about five years from the 1990s to the early 2000s now why did he do this he did this because he actually caught the period when this part of India started introducing cell phone coverage for its residence so cell towers and so in this article that he published in 2007 he looked at the efficiency of the kerala fish markets during that period of time before during and after cellphones were introduced why is that important it's important because if you have a cell phone it turns out that the range of cell phone coverage actually goes several kilometres into the ocean and so while you're out at sea if you've just caught a huge amount of fish and you need to figure out where to go that will value that fish most highly you can talk to one of your compatriots on shore and they can say don't come to this market we've already got a lot of fish go to the one that's one kilometer down the beach and so over the course of 1997 to 2000 he looked at three different regions of Kerala where they introduced cell phone coverage first in 1997 here and then in 2000 here and then later on that year here and what you see in this graph is the percentage of the population that actually has cell phones it starts out as zero of course and once cell phone coverage is introduced the number of people that have cell phones grows and grows until many people in the population do and the same thing in region two and the same thing in Region three so you can see that as cell phone towers were built people in the population started getting cell phones and they started using them and now what Jensen did was to survey the prices and the quantity of fish sold in each of these markets during this period of time and let me show you what he found you found something remarkable this is the average price in rupees per kilogram of sardines sold in various different fish markets about 20 different fish markets large ones and small ones on the Kerala coast from 1997 to 2002 and these black lines indicate when cell phone coverage was introduced and what you see is something remarkable or maybe not that surprising given that you know where I'm going with this the prices of fish were way way more volatile before cell phone coverage why because before cell phones you show up at the beach market with your fish if it turns out there's a lot of fish there the prices would crash and if it turns out that you show up and there's nobody else around and you're the only fish seller the price would shoot up there would be scarcity or an overabundance and so the prices would go crazy after cellphone coverage was introduced you can see that the prices were much less volatile there's still some movement but they're much much narrower in range from the point of view of economic efficiency the point of view of prices fully reflecting all of available information these prices are much more efficient than these prices why because of technology technology allows you to adapt to different circumstances and markets become more efficient not surprisingly what's interesting about this example is not only did prices become less volatile but the consumers ended up paying less for their fish on average and there was less waste on average so everybody benefited from this technology now it's not just cellphones they also introduced refrigeration they introduced other methods of communication so it was a combination of things but I think this illustrates the point markets are not either efficient or inefficient they adapt to changing market conditions so that's the positive side of Finance but let me now show you a negative side which is that we still nevertheless despite the adaptation in some cases have not the wisdom of crowds but the madness of mobs that is we tend to sometimes overreact to the information when we adapt we sometimes adapt in constructive ways so a herd of elephants is a clear example of an adaptation that can be very useful in certain cases in warding off predators but it can also create great havoc and you can get trampled to death in the midst of a kind of a stampede the Crusades religious wars you know the Salem witch trials bank runs those are all examples of these kinds of adaptations gone awry and if you need any more examples well the financial crisis in 2007 and 2008 was a pretty good example of the fact that sometimes our adaptations can lead to panic and counterproductive dynamics I'm going to give you a couple more examples of this to be very specific and to do that I'm going to give you an example of adaptive risk perception the fact that we all humans react in some very predictable ways to different kinds of risks so to do that I'm gonna ask you to make an investment decision right now I'm gonna show you four different financial investments I'm not gonna tell you what they are or even over what time period they span I'm gonna just show you the cumulative return of a one dollar investment in each of these four assets over an unspecified period of time and I'm gonna ask you to pick the one that feels right for you if you're gonna invest your entire wealth in this asset so asset the green line this asset turns a dollar into about two dollars not very interesting but not very risky the red asset turns a dollar into about four and a half dollars a little bit more interesting but more risky the blue asset is more rewarding but even more risky in the black asset is somewhere in the middle all right so these are the four choices and you can only have one out of the four you can't mix and match them so by a show of hands how many of you would pick the green asset nobody one person okay what entire wealth absolutely your entire well right green asset okay you know that's that's a safe asset how about the red asset how many of you would pick the red asset to invest your entire wealth no I want you to make a mental note that nobody picked that because I'm gonna come back to it later on and ask you about it again the blue asset how many of you would pick the blue asset all right we got a few techies here that are willing to take risk and then finally the black asset how many of you would pick that yeah that's the most popular one in all of the audience's that I've pulled with this example well so so so let me tell you what these assets are before you give it away so first of all the for ask the the time period is from 1990 to 2008 so we're talking you know a very long time you know almost 20 years and the green asset that that most of you didn't pick is US Treasury bills and it's a very safe asset but you don't earn much with it so it's not a great vehicle for your retire the red asset that none of you picked and I asked you to keep that in mind that is the S&P 500 which most of you already have in your retirement fund so you really better rethink that decision because none of you picked it what about the blue asset the blue asset is the single stock pfizer very risky but very rewarding and clearly it's not for everybody and what about the black asset the one that's the most popular of all it's got the best trade-off between risk and reward well this is the feeder fund for the Fairfield century hedge fund structure that was the Bernie Madoff Ponzi scheme that's why I had to stop it in 2008 it's because I you know that's when they blew up now you know how Bernie Madoff got as big as he did it's because all of us are drawn like a moth to a flame to high returns low risk in finance we call that a high Sharpe ratio Sharpe ratio is the ratio of excess returns divided by standard deviation and this particular black line here has a very high Sharpe ratio but it's not real and so we have to be very careful of that that's human nature that's how we think about risk now let me give you a different perspective on adapting to different kinds of risks this has to do with an article that was written in 1975 by the famous Chicago economist Sam peltzman peltzman did a very interesting study back in the 70s that has a very boring sounding title the effects of automobile safety regulation what peltzman did was to ask the very simple question given that we the government have imposed all sorts of different requirements on automobile makers they had to put a lot of safety features into the automobiles in the 1970s things like seat belts lap belt shoulder belts reinforced steering columns padded dashboards safety glass all of these things cost money and if you require automakers to put them in that's gonna raise the price of auto so consumers are gonna have to pay for it so here's the question how many lives did those feet safe was it worth it so he looked at the data and he found something absolutely shocking what he found was that in most cases there were no savings in terms of the number of deaths from highway traffic accidents no benefits at all in fact that's not exactly what he found what he found was that after the safety regulation was introduced in the first two or three years the number of highway deaths declined and then after two more three more years it went right back up to where it was before in fact there was only one incident where he found that the number of deaths from automobile achtung occupants declined more or less permanently but in that case the number of deaths of pedestrians actually increased and offset the benefits of that safety regulation and his interpretation was that what was going on is that once you tell people that automobiles are safer they'll dry faster and more recklessly they'll adapt and as a result you don't get any benefits his argument is that if you want people to drive safer what you ought to do is to remove all the safety features and install sharp spikes on dashboards pointing at the driver then they'll be very safe so since peltzman published his paper and it was not surprisingly very controversial right because this whole safety program was a really big deal a lot of money was being spent on it many many other economists weighed in they argue that no no you didn't do this correctly you didn't analyze the kind of driving was it city driving urban suburban driving you didn't analyze the context of driving where people commuting were they on vacation you have to control for all the factors driver education driver skill so on and so forth so this so-called peltzman effect has been studied ad nauseam over many many different papers and in some cases supporting it in all cases refuting it there was only until it wasn't only until 2007 or eight that there was a paper published that actually controlled for all of these factors and put this debate to rest this is a context where the authors were able to control all of these different features and focus just on being able to get to their destination a little bit sooner and does anybody know what that context is where you can control for these other factors and look for the palace min effect what would that be what would you think any any guesses yeah right that's right so that's another example of adaptation but what is the context where you can control for these various different factors for example in the previous case maybe you have conscientious drivers that are much more focused on mileage as well as you know how they drive and so the Prius is a bad example maybe people who drive other kinds of cars don't react the same way how do you how do you deal with that kind of an argument yeah rush hour bumper-to-bumper traffic would be a good example people commuting if you could focus just on the commuters the problem is it's hard to collect data just on commuters right again a great experiment because you could look at new drivers although then you have to worry about if they're really bad new drivers versus good new drivers how do you separate that out so it's hard right because you have to collect the right data and you of all people you know here at Google know that data is the key to a lot of these issues getting the right kind of data and being able to stratify it well so let me tell you what the context is you'll recognize immediately why this is such a cool result in 2007 these two economists Sobel and Nesbitt studied NASCAR races and it turns out that what they discovered was absolutely shocking it turns out that NASCAR has a serious problem and the problem is that every time NASCAR introduces a safety feature into its cars the number of accidents and deaths increase they don't just stay the same they increase and not surprisingly it's because if you're trying to win a race and I tell you that you've now got reinforced struts and roll protected bumpers you will push your car and that much harder won't you and that's a problem so this tells us that adaptation can actually lead to some very bad results we sometimes adapt in unproductive ways depending on the nature of the data and so I want to now turn to a subject that all of you I suspect know at least in passing if not great in very deeply which is how it is that people make decisions and how we model that and the difference between artificial intelligence and natural intelligence because this goes to the very heart of the matter of adaptation as part of the book I tried to understand how to explain this kind of behavior the peltzman effect and not only how to explain it but how to model it so that we can start thinking about modeling financial dynamics so to talk about artificial intelligence I want to take you back to search and understanding how technology has played such an incredible role recently I started getting interested in biotech so I decided to order a book from Amazon on one of the most successful biotech companies in the history of the industry Genentech so I went to Amazon and and searched then I found this book so I ordered it and as soon as I ordered it you know Amazon does the most obnoxious thing they post five other books that other people who bought my book also bought and you know this feature I'm sure that it's happened to you before and of course there are at least two of these five that I actually wanted as well so I ordered that too and this kind of a feature the ability to search data and come up with interesting patterns is an incredible change from how we actually think about expert systems now most of you I don't think our old enough to remember what artificial intelligence was like in the 1970s and 80s but but I was I'm old enough I could tell you and in those days expert systems has a very different philosophical approach to what we now think of as you know machine learning tools and let me just explain to you briefly what it was in those days in the beginning of AI an expert system was a piece of software that was supposed to capture a particular aspect of human decision-making you know picking a a book or choosing a car that matched certain consumer preferences or something physical like controlling a robotic arm to catch a wall in order to program an expert system in the 1970s 80s the typical approach was to learn everything you can about the particular context and in input all of the intelligence as a bunch of rules and to try to exhaustively specify all the possible use cases and then develop these rules to match and deal with those use cases so typically the code in those days was you know hundreds of thousands if not millions of line very very big code base and relatively small amounts of data being used because back in those days data was actually very difficult to come by not because we didn't generate it but because storage was really expensive so along with computing power and Moore's law one of the real revolutions of technology is storage capacity right and so in those days expensive storage led to small uses of data but very very complex algorithms the algorithms for anticipating all of the various different use cases in an expert system was really really big what happened today as of today what we have is exactly the opposite now what we do is to focus on large data sets and look for patterns in those data so the Amazon feature that I find so for us Trading frustratingly effective is actually relatively simple the algorithm is relatively simple right what's complicated is the data and how the data is structured how you access it how you organize it how you draw from it but the pattern matching is not that complicated by certainly by your standards it's not that complicated well this turns out to be a big philosophical shift in AI and the philosophical shift actually brings us closer to human intelligence human intelligence is much more like this pattern matching on large data sets than it is coming up with every possible use case and developing a response to it and so I want to give you just a couple of examples of that and bring to bear this notion of narrative because that's really what differentiates human intelligence from other kinds of species it's that we construct very complex narratives we are designed to store retrieve and interpret and react to narratives by narratives what I mean is a story that's different from a fact there are lots of facts people can memorize facts all day long that's not what we think of as intelligence what intelligence is is in being able to weave a cause-and-effect story so I'm gonna give you an illustration of that that has to do with a very very simple challenge that all animals seem to deal with which is the identification of threats friend or foe and I wanted to show you a picture because the visual cortex is actually very critical for identifying friend or foe are you in danger or are you not in danger so can anybody tell me what this picture is some nicely colored squares abstract art maybe you don't know what this is right well it is a picture but I've just given you very poor resolution there are very few pixels here here's a more higher resolution picture anybody tell me what this is does this look like anything to you yet what men with a gun could be and you would think that's a pretty important thing to figure out right you know that's a an important fact it turns out it's not you know a businessman with a tie excellent that's exactly right what's that thing on the upper right though a man chasing the businessman with that time maybe that's the man with a gun so you can you clearly having blurred vision is not a good thing higher resolution gives you a better chance of a day of adapting and figuring out friend or foe so let me tell you this is a picture of yours truly it's a selfie and behind me is a ninja stalking me now it's not a real ninja I took this picture in the Washington DC spy museum and so there's no threat here that's why I'm smiling but it's important for me to have the resolution to be able to develop that narrative so this notion of friend or foe requires us to have a certain kind of resolution and then to identify patterns based on the resolution this pattern I can tell you right now is not scary but if it were a man with a gun or a knife or a real ninja that would be a threat so I'm gonna give you another example a friend or foe this is something that I think all of you do all the time I want to imagine that you're at a cocktail party and maybe this is for new employees so you're meeting each other for the very first time and during the course of the evening you end up finding out different bits of information of people at that cocktail party so the kind of things that you might talk about or you might find out about are different facts about themselves like gender or sexual orientation there are two genders roughly and two sexual orientations roughly so there are four possibilities for that characteristic race race ethnicity let's say roughly there are four major races or categorizations for different age groups you know young middle-aged older millennium current home state fifty home states religious affiliation so on and so forth suppose that these are the different features that you are collecting when you're talking to these people at a cocktail party okay so I want to introduce you to two people in particular that you might have met at this party Jose and Julia Jose is a Latino gay male young professional from California who has no religious affiliation he's a Democrat middle class with an MBA that's Jose Julia on the other hand is a heterosexual female white middle-aged from Texas Christian Republican affluent with a Bachelor of Arts okay so now I'm going to ask you a few questions about Jose and Julia first of all if you were going to be launching your own tech startup who do you think would be most likely to help out and be productive Jose or Julia how many people think Jose how many people think Julia okay Jose gets the job if you are planning a fundraiser for a cancer event who would you hire to help organize the fundraiser Jose or Julia how many people will hire Jose for the fundraiser how many people would hire Julia for the fundraiser okay if you were looking for somebody who was cheating on his or her tax returns who do you think would be most likely cheater when you think Jose would be a cheater or you think Julia Jose or Julia wow that's a boat that's enough that's amazing what's amazing is how judgmental you people are I can't believe that you haven't met these folks I've just given you a few little facts and you're ready to you know Hall went off to get an audit and hire one what happened here is human nature in the sense that all of us make judgments all the time we are evolutionarily trained to make judgments even though we're talking about a relatively small amount of data right I mean you know these are very very minor bits of superficial information and yet you're ready to make decisions based on them now it turns out that they're actually not that little information as it might seem because if you do the combinatoric s-- and ask how many distinct possibilities are there of individuals how many different pixels are there if this is your view of society it turns out there are three hundred and forty five thousand six hundred different distinct types based upon just these characteristics that's more pixels in a six hundred by eight hundred photo so just from a casual conversation you can actually find a very very big set of little boxes and figure out where julia fits and where Jose fits in those different boxes and based upon your historical experience of those three hundred and forty five thousand six hundred boxes you will make a prediction you predict that Julia will be good at a fundraiser because based upon what your accumulated dataset has shown you your pattern-matching algorithm suggests that people within the vicinity of Julia's feature vector will actually be good at doing a fundraiser what we do is what the Google search engine does every day we do it to and it's really effective but it's not perfect and the reason it's not perfect is because with three hundred and forty five thousand six hundred pixels and with a finite amount of lifetime of experience we have a very sparse data set that we're making inferences of but the point of this experiment is to tell you that we do this all the time we are hard-wired to make those inferences and in some cases those inferences can be disastrously wrong but nevertheless from an evolutionary perspective what you don't do can kill you what don't react to can kill you so we are all designed to make these inferences things like racial bias gender bias these are all things that are based upon very very sparse data sets and human learning algorithms that are not unlike the machine learning algorithms in fact they're probably less sophisticated than the machine aren't learning algorithms that all of you develop so a couple of things that I'm going to wrap up it turns out that because these kinds of inferences based upon very very little bits of information can force us into making decisions and developing narratives it turns out those narratives actually have a real effect not only on our behavior but on reality so I want to give you an illustration of that effect and that has to do with this example that I like to give to my MBA students about Harvard Square how many of you have been to Cambridge and Harvard Square you visited ok so you know that Harvard Square is pretty congested and not unlike downtown San Francisco and you know that it's often hard to get parking in Harvard Square so what narrative that I was taught early on is if you're gonna go to Harvard Square do not drive because you can't get parking so this is what I thought until I learned an algorithm for getting parking in Harvard Square it's an incredibly sophisticated and subtle algorithm and it really works and I'm gonna share it with you now because most of you don't live and work in Cambridge so I don't think you'll compete with me for parking spaces but please keep this to yourself so here's the algorithm if you're looking for parking in Harvard Square before you enter the square while you're driving either in a stoplight or if you pull over before you get into the square close your eyes and say the following words three times Rabbi Mahoney rabbi Mahoney rabbi Mahoney if you utter that and can-can tation before you enter Harvard Square magically you will increase the chances of getting a parking now you think I'm joking I know you're smiling but try it and it will work in fact you can try it here I think it'll work here too so if you're going to downtown Palo Alto and you know the parking is a pain there too try this and if you're too embarrassed to try it tell a friend okay and then have them try it and then they'll come back to you I'll guarantee they'll come back to you and say wow this is amazing I don't understand why it works so you know I tried this and you know I got a parking space right here let me tell you why it works this really does work I'm not joking it works but it works not for the reason that most people think they think it's some kind of magical superstitious supernatural kind of phenomena here's why it works so people that I tell this to that have tried it their typical response is not to go to Harvard Square with a car because they just assume that you can't get a parking space when you utter the incantation part of your brain in my friend's case is not a very big part of their brain but part of your brain believes that it might be possible that it could actually work and so if you believe even partly that it could work it turns out that that belief affects your behavior I noticed this myself I measure this so what I found myself doing is I tried this out so now when I Drive through Harvard Square what do I do I Drive slower I look more carefully along the sides to see which cars have their taillights come on I look at people and see which ones are walking to their cars because I expect that there might be a space because of that narrative that has changed in my brain from you'll never get a space forget about it too I might get a space that small change changes my behavior and the change in my behavior actually gets me a space not always but as I said the probability goes up it goes up so this is a really important lesson what it says is that our human learning algorithms based on sparse data generates all sorts of narratives and those narratives have a life of their own so gender bias a case that obviously you dealt with recently here at Google is this situation I'm not going to debate whether the facts are true or false I don't know I haven't studied the data but that's why it's important to look at the data but more importantly not just looking at the data the reason the gender bias is such a hot-button issue is that that that narrative itself can actually affect behavior so you've got to be very careful in thinking about how people interpret what you say regarding these kinds of issues so the last thing I want to tell you about wrap up has to do with an example of just how powerful narrative can be this notion of coming up with an interpretation on various sparse data and the story I'm going to tell you has to do with the individual by the name of Erin Lee Ralston Erin Lee Ralston is a hiker who in 2003 was hiking in a remote part of Utah when he slipped down a crevasse and a boulder fell on top of him and pinned his right arm to the boulder wall above his elbow and you may have heard of him because he wrote a book that ultimately became a movie 127 hours and in this book Ralston describes his ordeal where for 127 hours he was trapped with very little food very little water no way to communicate no way to free his arm and after five days in this crevasse he cuts off his right arm above the elbow with a multi-tool now the question that I had was how not how in terms of the actual bloody process which is an awful awful thing kept comes the flesh break the bone I mean it was terrible what I want to know is how does somebody come to the decision after being stuck for five days with little food and water how does he come to the conclusion yes now at this point I will cut my arm off so he tells us in this book so let me tell you what he says a blonde three year old boy in a red polo shirt comes running across a sunlit hardwood floor and what I somehow know is my future home by the same intuitive perception I know the boy is my own I been to scoop him into my left arm using my handless right arm to balance him and we laugh together as I swing him up to my shoulder then with a shock the vision blinks out I'm back in the canyon echoes of his joyful sounds resonating in my mind creating a subconscious reassurance that somehow I will survive this entrapment despite having already come to accept that I will die where I stand before help arrives now I believe I will live that belief that boy changes everything for me what's amazing about the story is that in 2003 Aaron Lee Ralston had no boy he wasn't married he had no children he didn't have a girlfriend at the time he didn't have a girlfriend or a wife until six years later and at that point he had his little boy leo but it was because of that narrative that came to him that allowed him to cut his own right arm off to be able to realize this other vision the point is that humans are capable of incredible things like the founders of this organization at one point this notion of a search engine was an idea in the minds of two graduate students that narrative has enormous power in fact all narrative can have an enormous power and I want to conclude by pointing out that in finance we need a new narrative we need these kinds of alternate realities to understand what we can achieve finance is capable of amazing things but it's also capable of terrible things and so the purpose of adaptive markets is to try to differentiate between those two so let me wrap up by saying that it takes a theory to beat a theory the efficient markets is a great theory it's not wrong but it's incomplete and the behavioral is don't have a theory so the adaptive markets is an attempt to try to deal with that deal with the fact that like the parable of the elephant first encountered by five monks who are blind from birth the monk that grabs the elephant's trunk said an elephant is just like a snake the month that grabs the elephant's leg says an elephant it's just like a tree they're not wrong but they don't have the complete picture so we have to understand the human behavior has not changed a whole lot in the last 60,000 years but our environment has changed thanks to technology all sorts of technology and so what we need to do is to adapt to that kind of technology using the principles that Charles Darwin taught us and the bottom line is how adaptive are you to that kind of changing environment this was brought home to me in a very personal way years ago when my younger son was turning 5 years old at the time my older son was 10 so he was very jealous of his younger sons of mine younger sons or the attention and the presence that he was getting so he told his younger brother Wesley today you're 5 but I'm 10 I'm twice as old as you I'm twice as tall as you I'm twice as strong as you and basically just twice as good as you and that got his younger brother very annoyed for obvious reasons and it wasn't until dinnertime that his younger brother came up with a response and you know he said Derrick you're twice as old as me right Derrick said yes I am and then when youngers hasn't said well that means you're gonna die first and if you understand that you understand the adaptive market hypothesis thank you thank you this was a great talk folks were open to questions now yeah professor law I'm actually a fan of your adaptive mark hypothesis I follow you through like a 2008 when you first mentioned that thank you so but I want to have some different stuff like say you have three assumption like the human the first of three like you make certain assumptions a human actor follows self-interest and human like say there are but there are lots of a counter example like a most of a Googler here they actually vote against it a there I think financial interests like most other support to a high tax Yeah right there are lots of like just like a human rational the kind of thing I wonder if have you thought about maybe when when academia think about a financial market that they are put too much emphasis how the human behavior yes actually though those are maybe not you know not that relevant the reason is we are human so we know how the human works yes but that's probably not relevant using a whole say and the Julia example like I do like ads like pzt our motto like we have a billions of Lego features yeah try to flick the predictor up human will only click ads or not right but we don't make an assumption how the human behavior right we just observe that fact just like a physicist observe the photon will go to which is later we go to right but if you like to pick up a humans there I want to observe you with like a clickers or not the human behavior changes yes right yeah so there are many interesting things that you raised in that question so let me try to address just two or three of them and give other people a chance as well so you're absolutely right that when economists talk about self-interest almost always they're talking about financial self-interest that's not what I'm talking about when I talk about self-interest I'm talking about survival and we know that survival is related to financial self-interest but there are many other aspects of that self-interest and so from the adaptive markets hypothesis it's not at all irrational to vote for higher taxes even if you're going to be paying for higher taxes because you know that those taxes benefit other aspects of society that you would enjoy living in for example as a social animal I suffer when other people suffer this is human nature too so if I take a sharp knife and I start cutting my arm off right now all of you believe it or not you will feel pain the parts of your brain that trigger for pain will get triggered when you watch me cut my arm the reason that we have that feature is part of our social heritage we are social creatures and in order for us to survive together as pacts I need to feel your pain and so I don't want to live in a society where I'm the only rich person everybody's poor that does not feel good to me and so you're absolutely right economists are thinking too narrowly about self-interest when you broaden the notion of self-interest though it turns out that these theories of evolution actually have very very strong explanatory power the second point that I want to comment on is what you said about what you do here at Google and looking at data without making any assumptions that was part of the really big wave of AI changing the fact that by looking at large amounts of data and you need large amount you agree that you need large amounts of data right with a small amount of data making no assumptions and having a small amount of data tells you nothing it's garbage large amounts of data do not require assumptions small amounts of data require many many assumptions so I would argue that there's a trade-off between the two so given that Google has access to tremendous amounts of information you can do a lot of inference without making any assumptions but imagine if you're a macro economist trying to figure out whether or not you should stimulate the economy or you should engage in tight monetary policy how much data do you have then well as of the 1990s and the early 2000s we we didn't have the massive amounts of data that we have today so we had maybe 20 years 30 years of quarterly GDP data not as many observations so for that kind of decision you need a layer on top of it many many assumptions the problem is that those assumptions don't work very well so we need to get smarter about how to make that kind of a trade-off but with more and more data agree with you fewer and fewer assumptions for someone who's read the book who's watched your lecture you know they're wondering you know what do I do with this for an individual investor you know do they index their money like how do they practically put their teas at you know lessons into practice yeah that's a great question there are many different answers to that and it depends on the nature of the individual and how much time and interest and effort you want to put so let me give you a few examples one obvious implication is that all of us need to understand ourselves better before we can invest effectively in other words investment has a very very critical part to it which is you the human and so if you're investing in a portfolio that could lose 20 or 30 percent over the course of a week are you ready for that is that something you're comfortable with if it's not then you need to change the way you're thinking about investments if you are the type of person that losing 20 or 30 percent is no big deal then that should change the way you invest and so before you can decide what's right for you you have to first ask yourself Who am I how much losses can I afford what my goals are over the various different periods and what are the various different time constraints that I have for being able to monitor my portfolio so in certain cases investing in index funds no load low load you know kind of passive vehicles is the right answer because you actually don't have time to focus on making more sophisticated decision but for some of you studying financial markets as a hobby in which case you can actually engage in more high turnover strategies in that and more exotic securities and so on so that's one example a more sophisticated example what do you do with this is once I tell you that people adapt to changing market conditions you can now start monitoring market conditions and ask the question what are the species of investors doing in this particular ecosystem to give you a concrete example right now we have very low volatility and in low volatility periods the peltzman effect starts to take hold which means people start taking more and more risk they don't think they're taking more risk they think that risk has gone down so they're putting back the risk that has disappeared from their portfolio that can be a very dangerous thing because if you think the risk has gone away and you've taken on more leverage or higher volatility assets you might find yourself taking huge amounts of risk in another six or 12 months when volatility spikes so that's an example where if you know that markets are dynamic you need to be equally dynamic in watching your portfolio and if you can't do that you need to make allowances for that fact fantastic thank you so much once again and thank you all for being such a wonderful [Applause]
More from this creator:
Half of all Americans have money in the stock market, yet economists can’t agree on whether investors and markets are rational and efficient, as modern financial theory assumes, or irrational and inefficient, as behavioral economists believe—and as financial bubbles, crashes, and crises suggest. This is one of the biggest debates in economics, and the value or futility of investment management and financial regulation hang on the outcome.
In his new book, Andrew Lo cuts through this debate with a new framework, the Adaptive Markets Hypothesis, in which rationality and irrationality coexist.
Drawing on evolutionary biology, neuroscience, artificial intelligence, and other fields, Adaptive Markets shows that the theory of market efficiency isn’t wrong, but merely incomplete. Taking several examples from his book, Prof. Lo will provide an overview of his new theory of financial markets and what it means for financial crises, how we invest, and the future of financial technology.
Get the book here: https://goo.gl/5W3py6