Dawn of the New Everything: An Evening Lecture by Jaron Lanier

thanks for having me out I do a lot of things my main job is probably being a computer scientist I spend a lot more time with the algorithms than with writing books or talking to audiences but I I also play music and I love audiences so I I often and I have this crazy addiction to collecting and learning to play exotic musical instruments which I my excuse used to be is that it's cheaper than heroin but actually it isn't but I this is when the when I spoke at Credit Suisse which led to this evenings engagement I played this I thought I'd bring it back was anybody at the Credit Suisse thing okay okay so you this will be your second time you'll have to enter it again this is a mouth or again from Laos it's called a cab and as I'll explain to you in a second I think it's the origin of digital numbers [Applause] [Music] [Applause] [Laughter] [Music] [Applause] [Applause] [Music] [Applause] so I'm gonna sit because I haven't like stand in front of undergraduates and I tower over them as a voice of authority and I prefer to just be on this as much on the same level as you as possible so it's a conversation can I ask a huge favor can you turn down the house lights just a little bit the I mean bring up the house lights or turn down the spots on the stage so I can see better I want to see you guys hi thank you I really appreciate it sorry camera guys the instrument the can is very old we don't know exactly how old we know from archaeology that the Chinese version of it is very old indeed as old as almost anything in China probably older than the abacus it consists of 16 tubes in fixed positions that are either on or off and you play it combinatorially with different combinations on or off so it's a 16-bit number now you might think oh that's too cute that didn't that's not really the origin of computation because it wasn't used for calculation and yet consider this these were traded across the Silk Road and the ancient Greeks and Romans were aware of them the Romans copied them and made a giant musical instrument to accompany the gore in the Colosseum it was operated by steam because it was so big so it was one of these but gigantic steam-powered it was called the hydrolyse one of them survives we actually have one amazingly but I mean it's so big it's not that surprising I guess it was so big that this the slave boys who were supposed to operate it had trouble and so they had to semi-automatic and mate it with a crossbar system that could operate multiple holes at the same time and that level of automation was inherited as a hydrolyse evolved into two familiar things it became the harpsichord than the piano done one stream and it became the pipe organ another stream and in the very earliest examples of the keyed instruments this element of automation persisted from the earliest hydrolyse in the Middle Ages around Mozart's time a player piano was built that was non-deterministic that could improvise randomly in a kind of a simplistic sense but that notion that it didn't have to be do exactly the same thing each time inspired a guy named as a card to build a programmable loom and that inspired a guy named Babbage to make a programmable calculating machine and that inspired a guy named Turing to formalize the theory of computation and so on so it actually is the origin of computation so there it is this is this is the thing that keeps us all gainfully employed it sounds good to I've been doing this stuff for a pretty long time I'm in my world I'm very much an elder especially in Silicon Valley around the turn of the century I was around 2000 I was giving a lecture at Stanford in an undergraduate came up to me and looked up in terracing jaron lanier you're still alive and I was 40 at the time you know so so although it's a little bit less of a youth culture than it used to be there's still I still kind of feel like some surviving dinosaur from the earliest days so I've always had a somewhat quirky take on what we do in computer science my my major mentor when I was a kid was a guy named Marvin Minsky who many of you have probably heard of and Marvin probably more than any other individual articulated and stylized the way we still think about artificial intelligence he didn't make up the term artificial that's another interesting story but he did really kind of create the culture the storytelling around it and when I was quite young I used to argue with him about it because I thought it was really stupid and so and he loved it he loved the arguments and we'd go back and forth and I want to tell you a little bit about what I thought was wrong with it back of then because I still kind of think the same thing but Marvin would win the argument and he would do it in this very interesting way he would say to me okay whatever but we have to make a presentation to DARPA to fund the lab and this is good storytelling we're gonna get funded with it so play along and I'm like okay money talks I'm with you and it worked you know it was really effective storytelling to raise money in the early days still is still is so I want to talk a little bit about what bugged me about the storytelling of artificial intelligence and then I'm gonna kind of more from that into our present situation and talk a little bit about some concerns I have about the way we're deploying machine learning both in the culture and in politics and in in financial products and in other places and then talk a little bit about what I see as a shift that that would make things better which I might even be right about but I'll let you consider that so going back to the early days with Marvin the thing that bugged me about it fundamentally is that we were replacing what I consider to be kind of hard criteria for engineering like this car should be able to go this fast or you know this engine should get this hot without failing or whatever the criteria the criteria might be something you can measure with a meter that you can repeat and and is very precisely defined so you know if you're making progress and I felt like we were replacing that with kind of a science fiction story well we're creating this thing called intelligence well what's intelligence now Alan Turing had faced this question and he came up with an idea what they called the Turing test which he which down right before he committed suicide in the final two weeks of his life and which I think it's kind of significant but that's the topic we can get into if anyone wants to ask me question on my thoughts on Turing but essentially what he said is we're never gonna have a single absolute definition for something like intelligence any more than we're gonna have a single absolute definition for something like health or well-being it's it's a sort of a compound hard to nail down somewhat subjective idea so instead what we should do is define a comparative test where we ask if people can differentiate what a machine can do form what a person can do this is the famous Turing test and you are I'm sure you're all very familiar with it so my my probably the Turing test is I used to say to Marvin but how can you tell if the Machine got smarter if the person got stupider right because people do make themselves stupid in interacting with systems and in those days we didn't have a lot of examples of people getting stupid with computers but we could talk about oh I don't know people who twisted their lives into misery in order to try to beat the IRS or there were examples like that where I thought people were making them so stupid to please an algorithm these days there's so many examples that I mean like basically bring me teenager social media network damn there's your example so people make themselves stupid - please algorithms really easily people teach to the test in a way that actually hurts education people oh I don't know they just some E there's so many things so so so I mean I I move my argument one step further I said you know there's one machine in the Turing test and then there's two people because there's the person that you're comparing the machine to who's supposedly this benchmark of intelligence you know but then there's a judge so the Turing test might be passed either because the Machine got smart the contestant became stupid or the judge became stupid so there's a 66% chance that it's human stupidity not machine intelligence that you're detecting so therefore the the Turing test will tend to detect human stupidity so Marvin was not amused well actually he was amused but I so did this this problem that you're chasing a sort of a science fiction image and I know I want to say some of you might have studied psychology you'll see well actually there is this objective thing intelligence this variable gee that seems to correlate many tests all that even so it's still pretty vague I really have to say like if I told you oh here get on my new jet engine it corresponds statistically to this agglomeration of 75 quality measures but we haven't tested it for any particular thing oh it'll be safe you probably wouldn't get on that I mean you know we the the Vega our criteria is an engineering in the way the easier it is to fall into a fantasy space where you kind of get loose what get loose about your engineering so I think what's happened is this idea that there's this separate intelligence and machines has drawn us into a fantasy space that's allowed us to become sloppy engineers that's my basic hypothesis but then that manifests in a thousand different ways you know I should say I love the algorithms I've done quite a lot of work in machine learning algorithms myself sold a machine vision company to Google in fact I think the first one they have required way back in machine vision so I'm I'm I'm into the algorithms themselves I'm just into characterizing them carefully and not exaggerating what they can do now something started to happen and maybe some of the people in this room were actually right right there when it happened as computers became cheap and fast enough to possibly do things in real time in the real world which was starting to happen kind of in the 70s and a little more so in the 80s they started to be applied to the real-time events in society and we started to have things like flash crashes the fueled in part by automated trading we started to see phenomena I think a really interesting one was the rise of Walmart because Walmart was the first brick-and-mortar concern to use computation to model a part of its business in order to gain supremacy and what they did in their case they modeled their supply chains so that they gradually gathered enough data and had enough of a persistent model of what's going on that they could predict the ultimate positions of people they were negotiating with in their supply chain to their own advantage so they so they turned information supremacy into what was almost like central planning as opposed to the traditional idea of a market where each participant has relative ignorant of what the other participants know and they all have to kind of place bets do you all remember at some point in your education somebody made you guess the number of jellybeans in a jar right do you remember that day it might have been on the first day of an MBA program or the business course we were forced to take if you're computer science major or something like that and so everybody has to guess so what happens is as you all know you get a bunch of people to guess the number of beans in the jar and somehow the aggregate is better than the individual guesses because it captures something of the shared values and tends to cancel out some of this shared misimpressions and you end up with a better number it's a wonderful thing it's one of the main reasons that we like markets right but the thing is if everybody doesn't have individual information positions and doesn't have relative ignorant of each other's ideas it doesn't work anymore and so and that's why we don't like central plan centrally planned communist economies right because they might seem kind of magical at first for the people who are doing them but then they totally fall apart because they don't take advantage of any cumulative intelligence everybody with me on this and this has been proven historically enough times that we don't have to state it as a hypothesis I think we can call it a result so what happened in Amazon's case is that it said you know what we're not going to have all these people that supply chain guessing how many beans are in the jar we're just gonna marble the jar we're gonna model the students we're gonna just take control of this thing and we're gonna own the damn jar and you know what I actually in that case and and I think this will be true for many cases that I bring up I think the initial results of somebody gaining mastery over some aspect of our world with computation can actually be rather positive because for the most part most people are well intentioned so in Amazon's case and this might be a controversial thing to say right now but I'm old enough to remember that one of the greatest fears we had was a violent confrontation with the rising China and basically instead we let them sell us a ton of stuff and get rich and a lot of that was fueled by Amazon creating this extremely efficient sales channel they were responsible for something like 15% of Chinese exports during their main boom in AZ when they rose and they'd brought you know enormous numbers of people out of poverty and created a population that probably doesn't have taste a taste for a major war you know not to keep not too bad so I don't view this as some terrible thing that happened and yet I also don't think it's a sustainable pattern I think once in a while a centrally planned thing can go very well but in the long term it's not what you want in the long term you want a real market with many many information positions and that's in a way a contrarian thing to say in the Information Age because to get the magic of the jelly bean effect to get the magic of markets what that's saying is that there has to be relative ignorance of people to each other so they can form their own theories their own separate information their own approaches their own sensibilities and then those things have to be able to compete if everything is transparent enough for a central planner to emerge and that will always happen because of network effects if you have an entirely transparent system somebody somebody will end up at the top of the pyramid of the network of connections on it and you know I I've been told to to to not get too techie or anything sorry will you tell me if I'm saying something that doesn't seem intuitively clear to you all right so an example of what I mean about how openness can lead to centralization in a surprising way if you have no information boundaries in the earliest days of the internet I was quite involved with a I used to run around with a guy named Al Gore remember Al and there wasn't an internet yet there were a bunch of separate packet switch networks and he said oh let's use government money to bribe them to work together and it actually happened and during that period there was a very interesting debate which is this thing we're gonna call the Internet how much should be built into it should it represent people should there be human identity in it should it have persistence should we be able to deal with money in it should handle transactions should you be able to store information on it and the idea was no it's not gonna do any of those things it's gonna be totally open it's what the government will impose nothing upon it it'll just be this blank slate that has only the minimum bear functionality to support other structures and then we'll let entrepreneurs do it and then of course what happens when you have an open thing and you have a network of different entrepreneurs competing and there's no barriers at all you get Network effect becoming the dominant effect and then you get these monopolies like your Facebook and your Google and so forth and Microsoft yeah so they're they're not there's not just one monopoly there there are a few different niches that each have a different effect they monopolize different kinds of data or different kinds of connections or code or whatever it is but you but total openness gives rise to monopolies what you really want to have an effective information system and this is true whether you're building a machine learning tool for a particular purpose or whether you're building a politics or an economy for a whole society what you really want is a bunch of cells each of which can form their own internal theories have distinct information distinct merit memories distinct ideas and then those things have to compete with each other so that you start to get the process of refinement from the competition top down can give you magical results at first but it doesn't create perhaps I should use the word intelligence and so this brings up another really interesting thing there's I'm sure you've heard a lot of talks about the singularity and how machines are about to become super smart and everything will be different and all that you've heard there right and what I almost like to point out to people is that this has happened before was previously an information system that transformed human affairs that allowed us to do things we couldn't do before that was in some senses perhaps smarter than us and others perhaps not that we personified that we called a creature and that creature was called the invisible hand do you remember the invisible hand so in that case Adam Smith was playing the role of Marvin Minsky and coming up with a little bit of mythology a little bit of storytelling around this thing we call a marketplace and the thing about a marketplace is that it has taught us lessons it has taught us that mark that as it was I promised I'll repeat this again central control is not as good as multiple competitors who are genuinely distinct and so what we need to do is we need to relearn the lesson of the invisible hand but for computer science so what happened around the late 90s or so was that there was this very fervent techie culture that had taken hold in Silicon Valley and I was very much part of it it initially was very hippie and leftist and sort of socialist I'd say and then it kind of took a turn and became more libertarian but at any rate that difference was pretty minor compared to everything else about it we were basically kind of bad boys I guess who planned to take over the world and succeeded you know roughly speaking and we had to tremendously strong sensibilities that were an absolute conflict with each other not quite absolute almost absolute conflict one thing is that we love the hero entrepreneur we loved people like Steve Jobs and Steve Jobs you might recall had this sort of Nietzschean sensibility about entrepreneurs that you dent the universe so you were like you actually you somehow rise up out of the the tediousness of human destiny and you actually create something new you steer you steer the way things go in fate a little bit and a very powerful idea especially to to the young people that we were very intoxicating but then there was this other thing going on which is there'd been a very strong feeling that this came out of the hippie sensibility or the sort of more socialist idea that the internet would finally free humanity from the the horrors of private property and wealth differentials and all of these things that communists hated and on the internet everything would be free everybody would be equal everything would be open and the government would have no power and and everybody would be a cowboy it would be the Wild West forever and this image was incredibly powerful the main cyber rights organization is called the electric the Electronic Frontier Foundation but this cowboy a metaphor in mind and I was at the founding meeting for and I declined to join it because I thought this thing was ridiculous so once again I always seem to find myself on the outs with my visionary friends but this idea that yet it'll be the government won't intervene the government doesn't understand that they're old they don't get it so but here you have this problem you have you love the entrepreneurs you love Steve Jobs but everything should be free and Commerce is evil and money is evil how do you combine them how do you have entrepreneurs without money tricky right so there's only one solution that anyone's ever thought of which is the advertising business model so companies like Google and many others and Facebook now we're born on this idea that you give everything away for free pay for it with advertising and then the the thing about that is it started off really cute it's started by the way if those of you who watch my TED talk from last week will be recognizing I'm covering a little bit of the same stuff I hope that's okay but it was really cute at first it was kind of cool like oh it knows what kind of shoes I like like really into it but then this thing called Moore's lock have been chugging along and Computers got faster and faster and cheaper and cheaper and the cloud got more and more capacious and then the the customers who were using services like Google got more and more sophisticated with experience and then the algorithm started to evolve and we went from neural nets to deep learning and you know there was this progression of the buzz phrases where we basically started to sort out how this stuff can work better when we have faster computers we started to gather big data sets we've never had before to run machine learning on so we could get better results all these things accumulated until the very nature of what we were doing completely transformed and so currently I feel we have to call it behavior modification not advertising because when you have continuous measurement of a person and then you're giving them continuous feedback and you're doing that to create a change in what they do that is behavior modification that's a skinner box right we're one of Pavlov's experiments from long ago that's behaviorism and advertising had never been that before people have always complained about advertising they said oh those television ads are annoying but they're not watching you they might be trying to influence you they might be making a guess about what one fluence influenced you but they're not in a feedback loop with you that's very tight at all whereas this stuff is which is entirely different now interestingly this was foreseen by somebody I never met who is from the very first generation of computer scientists named Norbert Wiener so Norbert Wiener anybody heard of Norbert Wiener I'm just curious so he was I only saw two hands I think three hands two hands three hands okay this is a little shocking to you so Norbert Wiener is one of the Giants of computer science he so let's go back to the very beginning Alan Turing and Venoy Minh had defined a computer in terms of let's say batch processing you start with a certain state then the machine operates then you get another state well if it halts then there was this question of how do you know if it'll ever come to the point of giving you the final state which which was a big theoretical framework for thinking about computer science and it's very earliest phase but what happened is this this other fellow came along and said no wait a second what if there's not just a discrete start state and then the Machine runs and then there's an end state if it halts what if it's in continuous interaction or the real world so this was Norbert Wiener and he created an alternate basis for computer science that you might call the signal processing or the continuous model which is actually more of a description of what we do today in many ways it's not in conflict it's just a different framework that illuminates different ideas he called this approach to computer science cybernetics cyber comes from the Greek and has to do with navigating a boat so it's like you you know you're going the boats going the wrong way you have to adjust the sail on the rudder and then it goes the right way so it's constant adjustment as were the thermostat constant feedback and he said what if we think of the computer as being an elaborate thermometer or an elaborate feedback device with a lot of internal state if we think of it that way we get different results we get interesting interestingly different math and you get you get DSP chips and GPU chips instead of CPUs it just it just gives you a different framework not in conflict but just something different and so he one of his books is called the human use of human beings and I read it when I was 15 and just scared the bejesus out of me because at the end of it there's this there's this final afterward where he says well I've been describing computation as a feedback process just as a thought experiment think for a moment about a hypothetical planet where everybody has some kind of device on their person that can both measure things they're doing that they interact with and can give them stimulus of some kind and all of that connects somehow wirelessly to a big central computer yeah and then it's such a system would naturally turn into a massive global behavior modification scheme and such a society couldn't survive because they'd be under such manipulation that they wouldn't be able to solve real problems and it would be the death but this is only thought experiment it's absolutely infeasible to build such a thing I assure you the scientists that I am but no one could ever build such a global computer nobody could ever build such a capacious wireless network can't happen so it happened right I've thought I've thought often lately of Norbert Wiener and so the concern I have is that we've allowed ourselves to use computation in a way that I think is a little backwards I think what we're doing is we're allowing it to centralize control instead of to facilitate more and more cellular competition between different centers of control and I think that that's true as I mentioned in the algorithmic level I think we're making a mistake in our architectures and I think it's true on a societal level central control is really tempting especially thank you if you think you might be the one to own the center right it's really seductive and I what happens when you have central control of something is sooner or later circumstances shift and it turns out that that system is brittle and then it breaks badly so what happens when you have a centrally controlled financial scheme will eventually hit a more serious collapse than a more diversified one a company that can only make money through one particular scheme will eventually run into a problem and will not and it'll won't be as robust as a diversified company a person who has only one emotional strategy for dealing with life will eventually hit a circumstance in which they collapse diversity is good diversification of strategies is good now I mentioned companies so I want to point out something that is a vintage is there anybody from Google or Facebook here because I I would prefer that there were cuz I don't like insulting people behind their back so I want to be clear I really like Google I sold them a company good friends there have have friends at Facebook too however of the big five tech companies notice they there are three that don't depend on behavior modification loops for pay for their income those three are Apple Microsoft and Amazon they're all they have product they have diversified profit centers if Xbox dies Microsoft is fine if Alexa dies Amazon is fine if iTunes da if the Mac dies Apple is fine you know like diversification is is resilience right however the two companies that do depend on the behavior modification loop do so exclusively Google has tried like crazy to diversify and all they've succeeded in doing is diversifying their cost centers not their profit centers same thing with Facebook they're still basically one-trick ponies no matter how big they get and so sooner or later by the math they'll hit that circumstance where they were they where they pay for it badly and this is good so this is I'm not anti Google or Facebook at all I actually think they need to snap out of it and diversify for the sake of their own investors like you know their play they're playing with you know they're playing a game and sooner or later - the time will run out now and I want them to change I don't want them to fail I want them to succeed but the thing is our whole society is becoming like that too we're losing resilience as a society through too much central control and it's not of a traditional kind it's not some kind of a communist regime that's pulling all the strings instead it's a funny kind of seeding of control - algorithms that are not necessarily any better than we are so what so this has been this has been a difficult thing and I understand that not everyone perceives politics in the same way but if you look at whether you feel yourself to be more to the left or to the right you probably like democracy better than alternatives you probably like societies that aren't run by weird autocrats and than others I but it anyway but I want to say is that prior to fairly recently there seemed to be an unshakable trend in which more and more countries were becoming less autocratic we're becoming more democratic even traditionally hard nuts to crack even China was kind of becoming more democratic and then in particularly you know like if you visit the Reagan Library just to choose a Republican there's this there's this display of all the countries that went Democratic while he was in office which is really an amazing thing isn't it just astonishing and so then rather recently and corresponding rather directly country by country with the arrival of social media on smartphones there's been a reverse trend that had not existed before now we can only run reality once so I can't prove scientifically that there's a causal link here there's only a correlation maybe if we ran reality a second time with everything being the same it would turn out differently but the one sample we have seems to show people turning away from democracy when their social lives become more socially controlled by algorithms so what's going on with that and and I won't talk about the situation the United States we might have different interpretations of that but can we at least talk about countries like Hungary Poland Turkey China that's now got a leader for life and didn't before many other examples can we at least talk about those so I have a theory about what's going on here which I'll share with you and it has to do with the way behavior mod works so behavior modification works through stimulus that's calculated in response to the metric state of the person and their and their surroundings right and so classical behaviorism tended to divide stimuli into positive and negative ones with the classic examples being candy and electric shocks and they're more layers of complexity to this but we can start with that and famously it was discovered by Pavlov that you could use a symbol of a stimuli of stimulus and still get the effect in the the popularized example was he would train a dog to salivate at the ringing of a bell so the Bell came to symbolize the availability of a treat that would make the dog salivate and so in the same way you can do all kinds of tricks with people and this is done in psychology labs all the time some of you probably were talked into participating in Psych experiments when your undergraduates or something and they'll make you do some tasks and watch you through a one-way mirror and in those cases you probably had symbolic reward and Punishment they might have been say oh you'll earn some money or you'll those people don't like you because of what you did or whatever it is now when you have a smartphone in my book I sort of imagine this feature where there's a drone that actually drops real candy when you do the thing that's some benefit but we don't have that yet so it's all symbolic and candy crush you see pictures of candy right which is the same thing as Pavlov ringing the bell and so what you do is you have this world of symbols and the very most powerful ones are social so it turns out that people have a profound profound sensitivity to social stimulus stimuli and the reason for that I think is pretty obvious which is because of our high degree of neoteny so in the Ahtna means that this slowness with which she develop out of infancy how long you retain the traits of a very young version of yourself so in the case of people we're not able to fend for ourselves for years after we're born and that's not characteristic of most creatures on this planet most creatures on this planet are not as dependent for as long as on their parents but we really are we're absolutely helpless for a long time and so therefore being able to get along with others read signals communicate is not just a nice thing it's the skill for survival when we're young period it's skill number one so it's very very very deep in us and when we feel that we're liked that we're respected that we looked up to whatever we get a feeling that's quite powerful when we feel that we were disrespected when we feel that were looked down upon when we feel that we're being made fun of that that can really hurt that's also extremely powerful now what happened was the algorithms that were oh you know I just noticed the timer I didn't even realize was there I'm out of time but I'm gonna keep going is that okay yeah okay so what happened sorry organizers so what would happen was the algorithms that are routing ads to people and it's actually much more fine-grained than that algorithms design the ads any particular ad whether it's for a political candidate or a shoe or anything will go through many many different variations or things like color and font and wording and everything and then and the and those things are not correlated in general as if the whole population is the same but they're correlated not only with you as an individual with people who seem to be like you in other ways based on where you go if you're a woman where you are in your cycle just all these crazy things just all any signal at all that seems to have any positive effect it's just grabbed and agglomerate it onto and so it I do not believe that there was an evil genius sitting in a cubicle at Google or Facebook who said oh yeah we're gonna use these social things to hook people recently some of the founders of Facebook have done public Mayacamas where they said they did but the thing is I knew them at the time and I think they're remembering themselves as having been more Evil Geniuses than they really were because I was kind of there but I might be wrong I'm not sure I'm not but it anyway it doesn't matter the point is that the algorithms discovered that these social positive and negative feelings were incredibly powerful and could drive engagement now engagement is our polite word for addiction and I think at this point everybody acknowledges that that is the more accurate word that's another interesting topic we can go into about what defines addiction what what it is in isn't and all that but in anyway let's say they drive addiction you know here's the thing in in classical behaviorism and even contemporary experiments descended from them positive and negative feedback can be compared you can say is it more effective to use positive reinforcement with your work force or negative reinforcement and any of you who've been who have been involved in HRM management won't be aware that there are books their whole libraries of books about this topic and you can hire many a speaker to talk to you about it and the answer is a little complicated there's some circumstances where positive seems to be better and earn others where negative seems to be better it's very very much a patchwork however here's the interesting thing which has not been part of the classical science because nobody had any reason to ask it that the timeline profile of the two kinds of emotions if you make a bin of negative ones and a bit of positive ones is very different the negative ones rise faster and fall slower and the positive ones were eyes slower and fall faster right so to give you some examples it takes longer to build trust than to lose trust it takes longer to fall in love than to fall out of love it takes it takes longer to build respect than to lose respect many examples that kind of know because the feedback loop for these systems for behavior modification and addiction run by companies like Google with YouTube and Facebook and Twitter and so forth because these run on a fast feedback loop because the advertisers are a little like high frequency traders if you like they're trying to they're getting a better result if they move quickly with the most recent data it tends to naturally amplify the negative emotions that was not planned so even somebody who's just selling soap is gonna tend to amplify the most cranky and ugly and obnoxious people on YouTube or Facebook or Twitter or whatever it is and that is where the problem started in my opinion and I'm pretty sure because I was there that nobody planned that I think this is not an evil plan this not evil intent this was not it might maybe carelessness but I think this is mostly just a surprising result that nobody foresaw so I don't I think this is a mistake we need to back out of not something that we need to lash out at people about and I'm actually a little concerned depending on how things go that there there's already more of an anti tech sentiment than there had been before and I'm a little concerned that could get out of hand and that's not justified I really think that this core of the problem was not foreseen so we currently have a situation where you just get massively more bang for your buck if are stimulated by whatever you do and you're advertising online and that's terrible but that is the truth right now so we have a kind of an driven economy at the moment we have we amplify just some weird kids doing obnoxious things can-can they don't necessarily become the biggest YouTube's for YouTube stars but cumulatively they do become the ballast of the whole system and that is disturbing so how do we get out of it so I'll very quickly just describe the solution and I since I've taken so much time we might not have time for questions but maybe they'll let me take them anyway I don't know anyway so I think the solution is we is we have to stop pretending that we can have capitalism and everything's free we can't it's really it was a lunatic idea at the time it's still a lunatic idea we're artificially failing to measure the degree to which our economy has grown because of the beautiful energy and problem-solving that can go on online we've turned that all into this null free thing we've created this wave of anxiety that robots are gonna put people out of work whereas in fact most of the robots are gonna do 10 depend on data sets that still come from people doing stuff online so people will still be needed just in a new way I can get into that in a little bit more detail there is there's a degree of Auto Learning systems now but they're mostly still dependent on large data sets from people so what I propose is very simple it would be a new economy in which you pay for Facebook a little bit a little bit like you pay for Netflix if you're really poor and you can't afford it there's some sort of accommodation and the same way that you get subsidies for your heating oil or there's a public library and so forth if you contribute a lot and a ton of people look at your stuff on Facebook and you're really driving it you get paid you get paid a royalty now you might say oh that's horrible I'd never pay for Facebook I would never pay for that junk well here's here's what I want to point out to you during the same period of everything must be free mania Silicon Valley fervently believed that it was a terrible sin that anybody ever paid for movie or television that instead there would be armies of unpaid volunteers who would create superior movies and television and the model was the Wikipedia they said the Wikipedia is better than the Encyclopedia Britannica people do it for free we'll do the same thing with movies and TV and they tried their resilient startups and then this amazing thing happened companies like Netflix and Amazon and HBO got people to pay for their damn video and the result was not the world's worst television but what's called peak TV it got better like the amazing thing is if you pay for something it can get better like this is this new discovery of the last ten years right nobody ever knew this before so so customer choice can actually matter you can't have customers if everything's free so we can imagine a world of peak social media my wife's had cancers had health problems and she's she's doing much better now but just trying to get decent information online is impossible because it's completely filled with cranks and crazies and you have to dig through mountains of total junk to get anything useful and you basically can't in peak social media you'd be able to get good information that's worth paying for now you can't even pay for it because there's not enough scale because the free stuff wipes out the possibility to create a good plugin paid service you might ask how would we get from here to there how would it work exactly these are big questions the transition has to be done gradually I call it the move slowly and fix things approach and so you what would have to happen is companies like Google and Facebook would have to very slowly start to expand their Google already has a little bit of a paid area in YouTube and they'd have to start expanding that they'd have to do it with Google can't quite bring itself to let the paid part of YouTube be a real market they still want to centrally control it and pick and choose they have to gradually let go that and interest in the market the and it's very hard to do once you have control trusting in capitalism is very hard but they have to relearn that old lesson that's been learned so many times by so many earlier regimes so I'm not going to go into the details now unless we do questions and I get away with talking about it more in response but there has to be a gradual transition there are a lot of different ways to structure the way you might pay for it in the future there might be a subscription as with Netflix there might be a pay-as-you-go sort of a thing with micro payments for what you use the same thing for how you're paid I work now with economists and with economic students and we've been modeling these things using both analytic techniques and the less the lesson fashion agent-based modeling and in all cases we really believe this can work I believe that this would make companies like Facebook and Google grow I think they do better I think their volume would increase more than their profit but I think overall they become more robust more profitable bigger better companies I think we would create employment for a large portion of the the current user base for these systems not obviously it can't be all of them were then the whole system would collapse but I think of my a large minority would earn money out of it and I think that this would help allay this potential future of massive unemployment because of AI and the need for a gigantic basic income that we actually would not be able to pay for I think just paying people would make better quality stuff more robust companies but what really speaks to my heart is would make better computer science because instead of chasing the goal of modifying people to some stupid end we would actually suddenly be serving people and we'd have goals that we can state we'd understand what we were doing which is my old complaint with Marvin from so long ago all right that is the talk and what I'd like to do is destroy your schedule totally by taking questions is that okay all right when you were talking about the your experiences in the early days of the internet so if that foundation and the political ideals maybe more specifically that early internet pioneers hoped the internet would bring about sounds very familiar to what a lot of folks early in the crypto space sort of evangelized that the crypto currencies might bring about in our society do you see a repeat of the Internet where it is folks think this technology might bring that about but as it has becomes open it will become more centralized or do you think something else might happen yeah that's really interesting so there's definitely I think a cultural bleed from the earliest days so in the earliest days so this started with people like Richard Stallman and open software who was I have stories about that in one of my books if you want to read about me and Richard arguing in the old days and you're not a gadget I told that story but also people like John Perry Barlow had just passed away with this and in the early days the fantasy was a cryptography would be the new privacy and it would protect people from the government in the government was the only potential problem there was no other problem but an intrusive government everything that people would take care of their own affairs and everything would be fine if the government would just butt out that was the idea so so with cryptocurrencies I feel like the culture of crypto current craze for crypto currencies kind of a fusion of that idea that if it's if everything's encrypted that nobody can touch us we're safe and it's also got a bit of a in my opinion a bit of a tulip craze feeling about it and I think it's a it actually what it reminds me of mostly is the fine art market of like those of us who can we'll create our own rarefied objects to be apparent see among us through our own trust and and we'll just won't just make sure they're valuable to each other and you know I is it okay if I say slightly SmartKey thinks about cryptocurrencies why offend anybody alright so um a few years ago I think the question was why are they all Ponzi schemes ok so then just for the hell of it I had some students design a few cryptocurrencies that wouldn't be Ponzi schemes and just we never release them but just kind of to do it as an exercise so we looked at them and then we looked at those that why are these so boring and so then I said no no the right question is how can you create a cryptocurrency that looks like it's a Ponzi scheme but actually isn't right because when you could get people excited about it but then it could grow into something that a society could really use in the long term and so you have to play this little trick and so I feel like what if cryptocurrencies gonna jump out of the sort of tulip thing of a sort of fanaticism into something that society can use the long term I think that's what has to happen they have to be designed so they look like tulip craze is they look like Ponzi schemes so if people jump in and they're all greedy and excited and they think the world's gonna change but then it turns out Oh lo and behold we did this mature thing wow this is like this stable currency so we have to trick them into stability right and I guess in a way every successful economy has done that trick you know the US more or less did that the Wild West did that you know we kind of tricked everybody so there's this infinite gold up there it's a and actually it was no we're just building cities and row this is just gonna be normal eventually but you know so we have to do that trick again with cryptocurrencies and I don't think there's so so then the question is so so far so my prediction is the current generation are all Ponzi schemes that I know there might be some that I have the ones I've investigated or all Ponzi schemes maybe there's some that aren't then there's gonna be a next wave that won't be Ponzi scheme so they'll be really boring oh these are so boring I missed the good old days when they were when they were you know dangerous and then and then people will will come up with this eh and I'm talking about the Ponzi cryptocurrency that will turn normal eventually and I there's a few really hard nuts to swallow I'm kind of a Keynesian after a lot of battles with these ideas and looking at the evidence and I kind of think that the mature cryptocurrency is gonna have to have some kind of human structure guiding it you know I think there is gonna have to be some kind of fiscal authority of some kind there's gonna have to be some sort of a something there's gonna have to be a crypto fit or something because I really don't think you can have a long term general currency without without having interventions once in a while you know I just I maybe am wrong you know but that that's my impression and I and I see that not from an idealistic or from a political or an emotional point of view but just from a mathematical point of view I don't I don't think these things can be long-term stable enough totally on their own I don't think you can automate that and that gets back to the AI thing that I you know I think we kind of are asking too much of our algorithms so so I would say like you know invent this tricky thing invent this tricky thing that that looks risky and immature but actually is really mature and can lead to stability that's I think that's the next thing that has to happen here it's like discussion at this table earlier and I just wanted in your opinion if cryptocurrency user contribute to economic democracy your economic tyranny well I mean you know each this they're they're all a little different now there's so many of them but typically the early people especially the founders get more than the later people and they are they are kind of formally Ponzi schemes as I was saying so in order for that not to create some sort of profound economic centralization the various cryptocurrencies would have to be very disjoint so that the peak of ownership for each of them was totally the strength from the others so that way every group could have their own cryptocurrency or so I don't know but I it that sounds impossible and unstable to me I think network effects would create a super winners and a massive number of losers I so I think what you what you have to have to have a growing economy and you have to have fiscal policies that a certain kind of conservative or libertarian might view as kind of fishy or corrupt or something but I just don't think there's any other way because otherwise you do really end up with people from some earlier stage having too much influence when they no longer know what they're doing and the new people can't get financed you know and I think that's really you know you you need to you need to dilute the old in an economy you know and and you can't have a Ponzi scheme run everything just doesn't it doesn't last this is getting back to my earlier argument about central control and you know something like Bitcoin as you know the first example is essentially controlled currency there's a small number of people who were there at the start who talked to each other and kind of run it and it doesn't expand it just divides further and further so it it's a it's fundamentally I'm not sure I mean I think it does formally fit the criteria of a Ponzi scheme but I think you might also call it just some kind of Pluto currency you know it doesn't seem that way because it's still relatively new but if the same thing is true for say 100 years or 200 years imagine then so I think that's so they simply have to evolve you need to have like I say you need to have a cryptocurrency that looks Ponzi at first but then turns Keynesian later that and and that can be designed I don't think there's any mathematical impediment and and that's where we have to go with it is that clear a dollar is that clear yeah see I I don't think that's true I think what would happen if Argentina had been on Bitcoin then the same thing would have happened but it would have been the descendants of the bit point the Bitcoin founders who would have done it to them do you see I mean that's the problem is that Bitcoin is not what it claims to be it's actually just a different centre of control but it's still it's actually even more centrally controlled yeah I'd be interested to hear your thoughts on the attack surface of cryptocurrency state again the attack surface of cryptocurrency is because everyone always talks about how secure they are yet the exchanges have been so many exchanges have been hacked it's almost like you've got a deadbolt on a cart but with the cardboard door yeah well I kind of think I've had a minor career as a red teamer for different things where I try to pretend to be a criminal and break stuff break into things and whatnot and the thing is whenever you put all your effort into absolute security in one thing it tends to create a pretty steep drop-off into the adjacent thing that you then neglect there's a similar thing going on right now with like quantum secure communication lines or something like that like oh nobody can get between here and here but then just right over here oops you know oh who hired that janitorial service what you know that's a real example by the way so I I do think that there's a danger in trying to go for this absolute and I actually have always thought that that I think the purpose of to get back to your deadbolt the purpose of a deadbolt isn't to stop somebody who's really good at dealing with locks you know like your average MIT undergraduates student is gonna laugh at your deadbolt right there but what it does is it it it creates an a kind of a memorialization of a social contract what it's saying is look you know that if you go through this thing you're doing something that some that the social expectation is you shouldn't do and and I think if you make the lock super secure it tends to then under emphasize the value of that social contract which is ultimately the only security you really have because ultimately at the end of any of these things are gonna be human beings ultimately cryptocurrency exchanges are run by humans with computers there's other people around those humans and somehow something is gonna get screwed up if none of the people kind of believe in some sort of social contract some sort of social norm some sort of ethical system some sort of shared value some sort of shared stake ultimately you need to have that social level or any level of security is useless we've seen exactly the same thing by the way in the intelligence establishment where the more computers computery they got the more they had these massive leagues right so you had you had sudden catastrophic leaks from people who weren't even master spies again and again and again as you started to harden the cyber side of these things and now they kind of have the problem that they're so hard they can't talk to each other they can't talk to themselves right so you really like there's no substitute for trust and for a trusting Society that is security everything else is just a memorialization and a reminder and I so so I'm not arguing that there's no utility to crypto I'm just saying that if you expect more from it then it can give you're fooling yourself it's all ultimately based on society that might be UPS I'm you mentioned singularity earlier on how far isn't the singularity would be the point at which people have descended into fantasies about computers to such agree that they're too stupid to deal with their own affairs and they commit mass suicide through idiocy and the singularity either will not happen at all or will happen in five years all right wait I'm sorry was there somebody else this is trying to suss this question because it was something you talked about at the end I'm trying to suss this question it but basically you said something about you're talking about how negative reinforcement from advertising isn't necessarily something that was kind of constructed by the evil masterminds at Facebook and Google whatever it's almost something that we selected against ourselves because we're susceptible to it by our nature and I guess I'm paraphrasing that but that's kind of a disturbing realization I don't I'm not gonna challenge it it's probably correct but it's something that is very hard to fix if that's true in an open communication yeah how do you yeah sure the way you've been deleting all my social media see you you you create this horrible Taxation on any spending related to Internet content at all that goes away the longer you wait to place it based on any data you get that might influence it so in other words in order to if I want to post something based on any information I get from the world I have to pay an extra I don't know a thousand percent and every day I wait it starts to go down until after maybe a month it's down to almost nothing so what it does is it incentivizes longer cycles and then and then what you do is you start to catch those positive emotions that run on longer cycles and the negative ones so it's very similar to those people who think that high frequency trading is a terrible thing and they propose transaction tax or some kind of thing it's a very similar problem I'm not high-frequency trading is different though and I don't want to bundle them together because the argument about whether it's good or bad or the different versions of it it's real it should be kept separate but but the the mechanism is similar and so the the the competitive compensatory mechanism would also be similar any other questions oh I saw a hand hoop go over there there we go so for someone who spent their entire career in computers you seem to have come to a place where you're very focused on humans and the societal structures we have so I'll ask you a more philosophical question which is what do you think the the meaning of life is or the the purpose of humanity you know it's it's I think this is where we get into faith and we get into I hope you won't see this as evading your question but as answering it but I I kind of often find myself stuck between two groups of people and I find myself disagreeing with both of them and so I'm kind of in the middle and it's a tough place to be one group are this or my sort of technical nerdy friends who think that religious people are full of it there's no such thing as consciousness it's just a coffee it's just some effective computation and this whole level of spiritualizing things is is just the source of a bunch of superstition and manipulation and trouble and we should just give it up it's just it's just a flaw of weak Minds and then there are the religious people who think that you know we've fallen into these times of distress over a lack of faith and a lack of belief in that where we're turning ourselves into robots and all that and I actually think I have sort of partial sympathy with each side but I think the way I used to put it when I was younger it still kind of works for me that you have to walk this tightrope where if you fall to the left you become superstitious and if you fall to the right you become really duction issed so there's just like tightrope and on the tightrope you recognize that internal experience exists the consciousness is actually something in that it's kind of hard to fit in with empirical science because it's not something that we know empirically we can't make a meter for it and we don't really know what it is I've never seen any convincing I mean they're interesting things to be said and they're interesting speculations but ultimately it's mysterious and you have to in order to be a realist you have to be able to accept the degree of ignorance that you possess you can't be a realist by pretending you know everything that is not realistic and so and that's where people get mixed up what they say is no I'm scientific I'm rational and I you know we already basically know everything we don't know all the details but we know that math runs the universe we know that the basic physical theory is we know that it's all molecules we know it's all computation we can learn and invent details but we basically know it and if you think you know that that much you're not being rational you have to accept a limitation to yourself in order to be honest and in order to be rational you have to be able to see ignorant to be a scientist or otherwise how could you even identify unknowns you have to accept that you're ignorant in order to be a good engineer for otherwise how could you recognize that there might be solutions yet to be discovered and so I I don't have a good word for this this middle path this this type this tightrope it's it's something it's like a rationalist spiritualist or something like that a rat's spirit I don't know but it's it's a it's this in-between thing and so from that perspective we have to say that we're not privileged in our position to know a lot about ultimate questions we don't know what death is we don't know what's going on with this overall reality that we're in we don't know if it's even possible to speak of such things we are free to have different ideas about it however so for instance let's suppose that somebody perceives a sense of grace and reality which is to say that it's not quite entirely random but somehow there's a a sensibility or an aesthetic to the universe or something and this is a kind of interesting thing in nature you can do evolutionary storytelling you can say well this butterfly is beautiful because it's signaling or whatever but the thing is there's always kind of a little more extravagance and really would be needed to meet the criteria of any of those theories there's like nature's just a little bit more extravagant than it needs to be and and that's something similar to this idea of Grey's and I don't think it makes you into an irrational boob to notice such a thing and I think that that is actually not too far from the thing that religious people perceive if it might have a slightly different language or slightly different storytelling behind it so I think there's some kind of consensus of all these things that's kind of in the middle and probably pleases no one okay well thank you very much sure [Applause]

Channel: Bridge Alternatives

Views: 29,627

Rating: 4.9018407 out of 5

Keywords: Jaron Lanier, time Summit, Bridge Alternatives, Social Media, Machine Learning, Artificial Intelligence

Id: fhT1mseRLTs

Length: 70min 42sec (4242 seconds)

Published: Wed Jun 27 2018