Jaron Lanier (Scientist, Author, Musician and Artist) | Blockstack Summit 2019

Hey hi humans how many of you saw the New York Times videos I did in the last couple weeks so not all of you I'm going to talk about a concept called dated dignity this comes from I wrote a book called who owns the future about a decade ago now I think and the original term for the original title for that was gonna be digital dignity and it's kind of evolved into dated dignity it's a term that's getting used a lot now the candidate Andrew Yang's using it there's some little cluster of researchers into it at Microsoft so what is it it's basically the idea that people should have ownership and control of their own data now in order to explain that I'm going to first explain to you why I think we're screwing ourselves over by not doing things that way and I'm going to do that with a parable about machine language translators that I have found to be clarifying then I'm going to introduce the concept itself it has two parts one is called data as labor and the other is called mids M IDs and then after that I'm going to talk about some of the new problems that would come up in a world like this and what we'd have to do to mitigate those because nothing ever is perfect alright so I'm going to start with the parable of the translators my very sweet and generous mentor when I was just a teenager was a guy named Marvin Minsky how many heard have you heard of Marvin probably minority Marvin's the principal author of how we think about AI he's also one of the foundational computer scientists he figured out a lot of the basic stuff about computer science and he was MIT professor he passed away a few years ago and he was this most generous and wonderful mentor ever ever and if you trace back the lines of people who've done a lot in the computer world they'll often go to a mentorship with Marvin so Marvin was really into AI and I've never liked the idea very much we used to argue about it all the time anyway in the late 50s Marvin assigned a couple of graduate students the project of making natural language translation work over the summer which sounds comical now because it turns out it's harder than we initially thought but in the late 50s it was plausible to think well if you took two dictionaries through the two languages and you had translating dictionaries between them and you had grammar models and translations between the grammar models the whole thing should click together of course that only made gibberish it never worked language translation didn't start to work until much later despite many many many many attempts and a lot of investment it started to work in the 90s and initial are initially at IBM's labs and of course that technique introduced the idea of big data there was a very large corpus of translations that had been done at the United Nations and then these reused for to source phrase translations the phrase translations could be matched up with a few rules and suddenly there was readable translation kind of an amazing thing these techniques have been refined for the last 20-something years and now as we all know you can get memos and whatnot translated for free and Google or Microsoft or other places the Chinese service is and they're pretty good now I love these services I'm so happy they're there they've saved me a lot of time and trouble I think it's great it's an advance but here's what I want to look at before the service came along there was a whole world of people who made a living translating memos and corporations the language translators these people have seen a decline in their career prospects that mirrors the decline of other people whose work can be conveyed online like recording musicians investigative journalists photographers what happens to people is a little different than the cliche there's usually a public argument between some people who say oh my god automation is going to put all these people out of work think about the truck drivers giant crisis we need a universal basic income and then somebody else will come along and say oh no you're a scare monger if you have automation it just makes more jobs you know and so this conversation bounces back and forth between these positions but if we actually look at what happens neither scenario is exactly what happens what happens instead is that the overall number of people employed doesn't necessarily go down but the distribution of rewards to them changes so before everything's turned into something organized by a central digital hub there's a bell curve distribution more or less which is what you should expect a market is a measurement device and a measurement of phenomena in nature generally should create a bell curve right and so you should expect that in economic outcomes however once things are organized around a central digital hub you see instead a zip curve you see a small number of people do super well and then it falls off into a long tail so for instance musicians before made a bell curve in the united states around the turn of the century over 500,000 recording musicians making over $100,000 a year in the dollars of the time so a large middle class of recording musicians today it's a much smaller number just in the thousands but of those some of them do pretty well the measurement of how you how well you do tends to be how closely integrated you are with some big computer so it's a different that the computational architecture overwhelms the previous dynamic of market demand so different things being measured which results in a different curve so that's what happens so the and there's another way to think about it which is that the people who are out in the long tail are taking the risk for the people who are closer to the big computer so a way to think about this is that all the YouTube producers take risks to fund their videos but only a small number are deigned to be worth being pay by YouTube who acts like a communist central planner if you like or another example in the old system which had its own problems particularly with corruption a taxi driver could purchase a medallion that would give them a kind of exclusive entree to be a driver particularly in cities like New York the advantage of that wasn't necessarily that they made a lot of money as compared to today's uber driver but what they got was lowered risk they got predictability so they could say hey I can take 20 years and put a kid through college because I know this thing's gonna be good the uber driver doesn't really know what's gonna happen so the way I put this is that the big computers radiate risk out to the periphery and the way they make money is by making everybody else take their risks and that's fundamentally the reason why companies like Google and Facebook are valuable but getting ahead of myself this the same transformation of a bell curve to a zip curve has happened for the language translators now you might say to yourself if there's some language translators who are living less plantable lives if they're more nervous about their positions if things are harder for them that's sad but it's part of creative destruction they after all are obsolete because there's now this AI program that those the translations instead so they should be grateful they have anything okay the problem with that is that it's not honest about the nature of the service human languages are alive the English we speak today is not the same English we spoke five years ago the English we speak today is not the same as the language we spoke yesterday and the reason why is because there are new memes there's news there's culture there's science how many of you were using the term quantum supremacy before today or yesterday probably not too many but now it's a thing it's got to get translated I mean it was out there but it was really obscure and there's a zillion examples like that every single day okay so what makes the corpus of phrase examples keep up so that the things still work well I need to be blunt about this those of us from Microsoft Google and other companies who operate these things have found people who happen to be translating as volunteers anyway and we steal phrase translations from them there's no other way to put it there are people who have multilingual family members on social media they are amateur clubs that do captioning of YouTube videos for their own their own language group all kinds of things like that we go and we steal tens of millions of phrase examples every single night when we scrape their web incorporate them into the model and then we have this magic service so I hope you can see where there's something screwy about this and I'm going to go through a few levels of it that strike me as being terribly wrong and then I'll go into the alternative that I'm sure will at least be better although certainly not perfect okay the first thing that's wrong is that it's a lie and it's the worst kind of like cuz it's the kind of lie that removes dignity and humanity from another person we're telling the people who translate you know what we don't need you anymore we have our big electronic brain you're obsolete oh but don't go anywhere because we need to steal from you to make our big our big electronic brain work do you see what's wrong with that there's something just spiritually wrong with sending a message that oai it's gotten so good it's doing better than people when the people are still needed there's something like a spiritual crime in telling people you're going to be replaced when it's based on a really screwy and dishonest and self-serving set of definitions so that makes me ill I really hate it but then let's go to another level of what's wrong fundamentally our job and for running a language translation service is to make it work as well as possible wouldn't it be awfully nice if we could go to the community of people who are good at translation and say hey we really need the latest phrase translations we need to have the phrase translations from the popular videos from overnight from the popular posts from the popular tweets we need to have translations for whatever the hell the president just said we need this stuff because other people are gonna want this translated and we need a bunch of you to do it because we don't trust any small group of people to do it why can't we tell them that why can't we ask for what we need to make the program work better well the reason we can't is that even though we know how to find their work we don't actually have a relationship with them this is all done through this veil of anonymity so in a sense their privacy is protected but the protection of their privacy involves stealing from them in a sense we're getting the data we need because of the amount of scale but it's a really inefficient very quirky unreliable and loose way of getting what we need when obviously just being able to have a conversation would be better so the engineering actually suffers in order to create this illusion of the giant electronic brain and the total arms-length volunteer nature of the internet but then there's a third level of stupidity here which is why are these people even cooperating with us well the reason why is that they're addicted and I kind of a formal sense what started as a cute idea which is to give people the feeling of socialism and everything's free and everything's sharing while still being able to run businesses resulted in this weird setup where any time if you're a big alert well not for all of them but let's say for Google and Facebook and Twitter anyway we'd love to all you guys I sold the company to Google still like the people there but still the business plan is evil he is a certain term because 90 something percent of revenues come from third parties who want to influence I need to people who connect over these hubs so you want to do social networking or you want to accumulate search results somebody else can can benefit from any of those things they're all financed because somebody else wants to benefit all the parties who are involved so the people who are providing the free phase phrase translations are doing our finance to do so - somebody else thinks that they can be influenced in some way now this started out cute this started out as just a referral service here's your local dentist that it was like that 20 years ago and that was cool but everybody in the system gradually got more sophisticated the computers got bigger because of Moore's law the network got higher bandwidth the algorithms improved because of science and math but all the players got more sophisticated and in particular the distasteful and and ill-will the players the psyops people and the organized crime people and all that and so what you've ended up with is this giant manipulation machine where everybody's in a partial skinner box all the time do you know what I mean when I say Skinner box you all know that BF Skinner somebody said no BF Skinner was one of the famous behaviorists he would put animals or people in these boxes and control exactly what they experienced and then they get rewards and Punishment like candy and electric shocks and he studied how that changed behavior he's kind of a creepy guy he thought all of society should be turned into a network of Skinner boxes so I have another question for you who designed the very first actually implemented experience of a digital network that people could actually use an experience and then observe them for a few years this was from before the ARPANET who did it oh come on BF Skinner the very first implementation of a digital network was for the purpose of fooling people and tricking them and changing their behavior Skinner would have loved the world we created for the movie Minority Report he wanted a world where there was some central authority that designed society and influenced people's thoughts and desires I think it's incredibly evil and creepy but some people like it I'm looking at you Chyna but that's another tale all right so we're putting all these people in mild distributed Skinner own another thing another reason the Skinner boxes are stupid I guess I don't remember how many reasons we're up to you but maybe this will be the fourth one was it three so far so the fourth one is that Skinner boxes require a feedback loop where you observe people's behavior and then you reinforce things based on the behavior well the only thing we can observe through smartphones is how much you click how long you stay on something once in a while we can watch your eyes and your listen to your voice tone but basically the kind of stuff that we can really catch are the fight-or-flight responses so these are the emotions when you're scared or angry so those are the ones that create get reinforced so whatever content is put up there like just as a kind of example that really bugs me there's some reform group that puts up a ton of content like let's say black lives matter or the Arab Spring at first they get this magic carpet ride for free the social networking services connect them with the world all these stuffs happening but meanwhile behind the scenes the algorithms have to make money and they can only do that by increasing engagement so they can go to customers and say hey we can persuade these people that are engaged so they look around for who clicks who stays on well it's the people who are scared or angry so what you do is you detect the people who hate the stuff you introduce them to each other you reinforce them and suddenly black lives matter ends up funding essentially a rise of the Ku Klux Klan which had been dormant in the American Nazi Party suddenly the Arab Spring funds the propaganda rise of Isis everything turns to I'm sorry I don't know if this will be broadcast or something pardon me everything turns to crap and so that's so now we're up to four reasons why this thing's stupid I'll just do another reason I can go on and on another thing that's stupid about this is that in order to make this plan work we have to make manipulation the only product for these giant companies some of the most valuable in all history the the manipulation income called the advertising revenue but I'm not anti advertising I'm anti feedback loop advertising I'm anti Skinner box very easy to draw a red line between those things because almost all the revenues and the companies like Google and Facebook are manipulation revenues you end up having to pretend that all this other stuff has no economic value the machine translations this search all the videos on YouTube it's unpaid it's all volunteer it's all socialist sharing and what that does is it artificially dampens our measurements of how much the economy is benefiting from digital advancements and that to me it's just not so we should be showing I mean in my opinion a company like Facebook should be worth a lot more than it is but so should all the people using it like the whole tide that's supposed to rise raise boats has instead been lowered to the mud just so a couple of boats can sit there on the mud it's really a dumb macroeconomic idea all right so I think it's enough reasons to criticize the whole thing I could go on but that's enough I haven't mentioned politics six reason I I've visited places like Turkey and Brazil when they had elections recently and the UK and there's this really weird thing that's going on we're all over the world all at once there's this shift to a certain style and feeling of politics sometimes it's described as right-wing populism or populism but I think that's actually an incorrect designation I don't think there's any particular reason why it should accrue to the right or to the left at all and and I don't think it's populism because truth is a lot of these people aren't really that popular this isn't like the air Mussolini this is something different this is an error where everybody's just so damn cranky so paranoid and so removed from the truth that that anybody who's also like that resonates I used to know Trump not well but I met him a few times over a period from the late 80s into the 90s I think there were many reasons to criticize him for corruption and racism at that time but what he wasn't was paranoid and irritable and unhappy all the time Twitter addiction made him that way as it did his voters and they relate to each other it's not populism it's something else it's Skinner ISM it's a different thing okay let's get to the solution space so there's a lot of people who want to reform the tech companies Elizabeth Warren wants to break them up there's privacy legislation I played a small role on that I co-founded the advisory board that preceded a thing called the gdpr in Europe tried that tried other kinds of regulation I really skeptical of using regulation to fight this problem because I don't think you can capture it in words I think what if you just say it'll be illegal for programmers to trick people well what does that mean programmers are pretty nimble whatever the court definition becomes some people be able to work around it so privacy coordination between companies whatever it is that you're trying to prevent I don't think you can define it I think code wins when code meets law so how do you do anything you do it by reforming incentives incentives rule the world right now under the current way we're thinking the only incentive is to trick everybody more and more and so that's what happens if you have to pay somebody for the data you need from them to trick them that incentive with the undone but meanwhile you could value add to that data to create a service that they'd want to buy and then a different incentive would be created to make more and more valuable services I think that's an ironclad argument getting from that argument to actual implementation is not necessarily easy in fact it's almost necessarily really hard and tricky but I want to portray it to you so this is going to be the first part of the two-part definition of data dignity and this is the part called data as labor which is a term I tribute to my economics colleague Glenn while you can search it you can search for it and find some of the papers we've written so one of the first questions there are many questions that come up but one of the first ones is well if people are getting paid for their data what kind of a thing is it is it a product or is it a service a or B to which I say it's neither choice C is better and choice C is it's art there's an interesting legal tradition for art which is the creative output of people art doesn't get sold at least not in most parts of the world particularly in Europe there's a concept of moral rights where when somebody makes art it's inseparable from them all they can do is license it all right this is a crucial concept because your data becomes you the more advanced technology gets so if you could really sell it it would create a new world of so loss and slavery and all kinds of things so you have moral rights to your data in a data dignity world but you can license licenses are revocable you can have multiple licenses they're not masters that necessarily exclusive but they follow roughly along the lines of the laws that have been created to protect artists and musicians and so forth but obviously at a much finer level of granularity and universally another question what kind of data are we talking about if you sometime wonder what you're gonna do overnight and you just can't sleep and you know some other people who also feel that way get together and try to come up with a categorization of different types of data I assure you it'll take all night and you won't get done it's a really tricky thing to do people talk about exhaust data and creative data and live data and organic data I think the easiest divide that everybody can agree on is data that is destined to be read by an algorithm versus data that is destined to be read by a person the same data might go in both directions which should be treated as to co-present cases now in the example so some examples data that's going to an algorithm would be for instance let's say your professional gardener and you're known for a particular topiary or something but then somebody comes out with these great gardening robots and all of a sudden gardeners aren't being hired and yet people need data to run their robots well you might be contributing some of your data to make the robot do beautiful topiary and you should get a royalty for that this seems simple enough to me in the case of the language translators what starts out as data that is just going to the algorithm become something that's a little bit more planned where people are thinking about the end user and what they'll need to in a dated dignity word world instead of it just being randomly gathered in secret and stolen instead the purchaser and you think wow what phrase translations does somebody need so it's starting to shift from one category to the another to the other although it exists in both but anyway that's that's some of the kinds of data the kind of online data that I've been most excited about that actually exists recently is at Microsoft there's a thing called Skype video chat thing that's used a lot around the world and in researching it it appears that there are very large numbers of people reaching into the hundreds of millions in the developing world who figured out ways to top up their income and by the tens of millions create full-time jobs out of offering advice and services over Skype things like music lessons and yoga and legal advice language lessons all kinds of things like that totally excluding sex work by the way which is morally complex just in these other categories and that to me is a great example of data dignity nobody's helping them do it they have to cobble together a way to get hey they have to cobble together way to find each other and schedule it's a mess they have to climb a mountain to do this each and every one of them and yet they do it which is one of the reasons why I'm optimistic that creating tools for people to better themselves create it creatively will work that people will rise to it okay so I'm running out of time so I'm gonna skip ahead to the next half of the definition which is mids mid stands for mediator of individual data mids okay a mid is a collection of people who decide to work together now why is this important I'm going to give you a series of reasons why mids are essential and whether they're there at least as important as just paying people for data the first thing is that if it's each person offering data as an individual the competition of each to each will drive the prices down to approach zero this is exactly what happens on Mechanical Turk for instance so people have to be able to collectively bargain now I want you to notice something about what I'm doing here I'm drawing ideas both from the left and from the right both from libertarians and from socialists because I think all of these different traditions have value to offer I don't really like them all and I hate them all what I want to do is just mind them all for pieces that can be part of a solution so I just took a lefty thing and put it in the middle of variety thing we have to be willing to do this we can't be ideologues based on our parents and grandparents ideological divisions it's just too stupid like too pathetic so we have to just get rid of that just use ideas where we can find them okay so collective bargaining is job number one job number two right now we're in this very bizarre period where in order to enforce the centrality and overarching power of the central servers like the Google and Facebook cloud we've radiated risk outward as I said before and one of the ways that happens is that there's nobody who's responsible when when somebody says something that does gap they're shattering in a crowded theater thing so if somebody says go here's the real address of this person go kill this person and it happens who's responsible it was anonymous you can't find the person and things like this kind of happened to varying degrees all the time so we've petitioned the big companies Facebook Google Twitter and so on we've petitioned them to monitor our own speech who said oh please big corporation tell us what we can and cannot say and it's I don't like where that leads if you follow that down the road if you say sadism and lies and all these things they're they're terrible well of course they're terrible but how can you get rid of those things without creating some kind of central speech authority how do you do it there's an ancient answer and the two best writers on this predate tech culture and I would have mentioned Hannah Arendt and de Tocqueville any of you heard of either of those people okay if you don't read them okay seriously so both of them were interested in observing societies that seem to create a sense of quality and decency without central control and in a context where people are free the Tocqueville was looking at the United States khana rent was trying to understand how Nazi Germany went wrong and why some countries do better they both came to the same conclusion as have many other observers which is that what makes a society decent is it's not a playground of each person against each other person when you have that you end up with bullies instead it's groups of people who work together now I want to give you another example of this which might sound very different to you but it's actually similar I just didn't event with a guy named Muhammad Yunus in Italy a couple weeks ago any any of you heard of him it's kind of interesting different hands rise for each of my questions about what you've heard of so he won a Nobel Prize for starting the microlending movement which is where instead of a bank having to figure out which poor person deserves a micro loan they form together into clubs and look out for each other so the responsibility and the benef are both distributed and suddenly it works so mids serve that function amid might be like the New York Times it might be like UC Berkeley it might be like maybe the 49ers it might be like a great band it might be like a confederation of gardeners who are offering data to a gardening robot it might be like a Federation of phrase translations who are offering data to a translation program people join them freely they govern themselves they can inject people who don't do well they distribute benefits and I'm almost out of time oh god this stuff's so important I'm trying to eat as fast as I can they are a little like unions a little like journals a little like clubs but the key thing is that you can join amid and look out for each other and the mid develops brand quality that people benefit from upholding so the reason the New York Times doesn't publish a big picture of feces on the front every day is because there's different editors and writers they're saying now if we do that it hurts us all so having mids gives you a way to promote decency and quality and standards without because it's all by free association without imposing control from the center it's magic and it works it's worked historically again and again so mids mids solve the speech problem I'm gonna very quickly go through some of the other things I'm not gonna have time to talk about the problems this brings up unfortunately but next next solution that mids offer is a solution to complexity right now whenever you do anything online you click through an agreement and of course nobody knows what the agreements mean I was in sin Geneva with a conference of internet law people who wrote these things and I in 2,000 of them and I said can I see hands for anybody who's read a EULA that you didn't write yourself not a single one then I asked have you really read through the ones you did write yourself and there were only a few hands I don't know what that means the the thing is the outcome of the gdpr was a lot of regulations that change what the Kliq means what you still know what it means and so if you're part of a mid the mid can look out for you in a sense it becomes like a law firm that you've hired it becomes like a club you've joined like the triple a four cars or something like that there'll be lots of different mids they'll be mids that are hard to get into they'll be mid that everybody gets into like a mid for eye motion you can join freely and there might be twenty million people in it and it gives you a buck a week or something like that something little the mid for gardening robots might give you a lot of money it might be something you specialize in that really becomes a career you'll have a whole bunch of them they'll pay royalties these royalty streams will accumulate as you do more and more in your life turning into retirement funds so it creates a reasonable outcome a reasonable arc for an actual biological human life rather than a hypothetical robot life there are many other reasons mids are important oh god I have to mention the fiducial the idea of a fiducial is it goes way back to medieval common-law it's a tea of somebody if you hire somebody who has more information than you do that person has to represent your interests they can't work against you so a lawyer can't represent an opposing party the doctor supposed to represent your interests not the pharmaceuticals etc we can't ask Facebook to be a fiducial because people have different interests and you can't have a universal fiducial but mids can be fiducials so what it does is it creates an advocacy agent that actually is positioned to be effective okay so that's me it's a very brief introduction there's so much more this world I've outlined oh very quickly in any learning system that's complex you need to have what we call coarse-grained accumulators which are elements in the system that generate results based on persistent combinations of elements not each element by itself in biology we call this a species so instead of just a but a bunch of little microorganisms trading genes they build up a wall and whole species can compete against each other so that you can see the evolution of complexity something related but in a different plane happens in machine learning algorithms where if you just connect input neurons to output neurons it doesn't learn you need to have intermediate layers which are the coarse-grained accumulators then the thing can learn mids become the coarse-grained accumulators for society that's what the societal institutions that made society improve before did that was one of their functions mids can last longer than an individual life they can form value chains with each other they fame they form a coarse grained valuable ethical Society and they decentralize alright I think that's enough I'm 4 minutes and 23 seconds over all right cool thank you [Applause]

Channel: Stacks

Views: 3,928

Rating: 5 out of 5

Keywords: Blockstack, Blockstack PBC, Blockstack Summit, decentralized, decentralized web, blockchain, web 3, crypto, bitcoin, Jaron, Jaron Lanier, transactions

Id: wTbi-TPeFPE

Length: 35min 47sec (2147 seconds)

Published: Wed Dec 18 2019