 I'm Joel Garrow, I'm a co-director of this initiative. Future Tense is a partnership among the New America Foundation, Slate Magazine, and Arizona State University. Our goal is to look at not just at emerging technologies, but to ask what do they mean for our culture, values, and society, and what it means to be human, which is how we end up with this question about our killer robots, job killer robots. I want to give a special bow to our partner Slate today, especially I'd like to ask Tori Bosch to stand up and take a bow. She's the editor of our New Future Tense channel, being launched on Slate in the coming weeks. That's a space to watch. On Slate this week you're going to find a terrific series by Farhad called Will Robot Steal Your Jobs. I strongly recommend it to you. It's a terrific piece of work. Is Josh Levine here yet? No, I guess not. He will be here later, I guess, the executive editor of Slate and the editor of this series. Don't we have robots that can do that? No. Well, he's not here, he's a robot. The reason we're here today is that when robots came for the automobile assemblers I was not an automobile assembler. But now they're coming after the doctors and the lawyers and the journalists. So these are the white collar eating robots. And of course by robots we don't just mean WALL-E. We mean artificial intelligence writ large that is actually now arriving. And if this freaks you out too much remember that we'll be having adult refreshments at the reception afterwards, so stick around. A few housekeeping thoughts here. Remember this event is being webcast. So everything is on the record. If you choose to speak up please wait for the microphone. And please introduce yourself and please be concise. So now let me hand this over to the star of the show. A columnist for Slate who writes about the technology industry, new products and social media. He is the author of True Enough, Learning to Live in a Post-Fact Society. Ladies and gentlemen Farhad Manju. Hi, thanks and thank you all for coming. So I want to talk a little bit about the series. It's focused on this question that I've been thinking about for a while because many members of my family, my dad and my wife, my dad's a pharmacist, my wife's a doctor, she's a pathologist and she spends her days looking at slides that have images that may contain cancer in them and may not. And this has always seemed to me like a problem that's not well suited for humans because it's sort of like TSA agents, right? They have to, you have to sort of be very focused and also spot something rare. It seems like a thing that a machine would be better at. And in fact when I talk to my wife, many of her colleagues are kind of afraid of this and this was the spark of the series, which is a look at robots and machines that will be replacing people who are paid a lot to do their work but whose work is in some sense vulnerable to machines. And so the people on this panel I think speak to some of this and we wanted them here because they each come at this from a different sort of standpoint. So Michael Schmidt, he is, as a graduate student at Cornell, he worked on a project called Eureka which was a shorthand way of saying it is it's a robotic scientist. And what it does is it looks at data from the physical world and comes up with fundamental laws of science, of nature from that data. And he can talk more about it in a little bit. And now he's the president of Newtonian Inc which is a business that's focused on expanding this graduate project into a business and he can talk about that too. So Sarah Kramer is the next person. She is a doctor, a family practice doctor. And the reason I wanted her here is because I write about her in the series as someone who is actually not vulnerable to machines. Her job I think is one of the few jobs I found that maybe in 50 years or more we'll see. She could be replaced by a machine but before then I think it's going to be difficult and we can talk about why. And then Robbie Allen is trying to get my job because he is the founder of Automated Insights which is a company that writes automatically produces news stories from data. So their first application was to look at all of the stats from various sporting events baseball, basketball and you can look at all the stats in the game and come up with a pretty good story that is not written by humans. So I think I wanted to start with Robbie because you have the most sort of impressive demo of like if we want to see whose jobs are vulnerable next we would look to your technology. Tell us about how the robotics sports writer works and then I wonder if you can talk about what it can do beyond writing sports. Sure. And real quick you compared your wife to a TSA age, does she know about that? Yeah I'm lucky she's not watching this website. Yeah so you know conceptually our technology is pretty straightforward. We take a large amount of data which actually is not very easy to compile. We write algorithms and create a database of phrases that we then use to determine based on a particular situation in this case a sporting event which of those phrases apply to that game and then we do various replacements in the data that will include the stats, player names, venues, that type of thing and then when you piece it all together it's a fairly coherent three to four paragraphs worth of content describing the important events and things that occurred in a particular sporting event. So in order to make this happen we needed sort of a combination of skill sets. I have a background in computer science but I've also written several technical books. A couple years ago I had the thought that I wanted to create the next great sports blog network for those that are familiar with SB Nation which is a local startup that is probably the most popular collection of sports blogs. They have about 400 writers on staff that author all of their blogs and I thought well I'd like to do something like that but I wanted to automate it and so that's kind of where this notion came along that I thought if I could marry my expertise programming and writing that if I had the data I could just create technology that would automatically produce narratives. Can I just because that seems a peculiar thing why you know you wanted to create a sports book why did you want to automate it like who thinks of that? Well so stat sheet when I originally started a stat sheet in 2007 it was just me you know before I raised any money and so in order for me to scale I had to automate everything that I could do you know I could just sort of a manpower. It was a manpower thing there's no way I could have a 400 writer staff within a year or two and so I thought well if I could automate it in some way then that allows me to scale more quickly. I mean one of the things I found interesting is that you're doing a lot of work you're creating a lot of sports stories that compete with what human writers are creating but you're also creating sports stories that human writers could never do because they're focused on a team so small that you wouldn't be able to pay a human to do that work and now a robot can. I think this is and I found this in many different fields it's a it's a situation where the robots are giving us things that you know they're benefiting society at large there are some people who want to read about those sports who just you know those teams who just can't because there's no human writing about them and now we're getting that from from robots. So do you think of it then you know your technology is replacing humans or do you think of your technology is supplementing humans? So this is a question we get all the time mostly from journalists. When I started this you know the initial sort of blog post I wrote about it on our company website automatically got like a half dozen reporters emailing me within 30 minutes to know are they gonna be automated out of a job. I personally don't see it that way I don't think that what we're doing is gonna automate journalists out of a job at least yet. So I think there's a couple things one is we can have broader coverage than what a typical person can do as far I mentioned you know there's a lot of teams that just don't get adequate coverage in the sports space and so we can do that very inexpensively. The other thing is I will say while it's potentially not right now that our content is as good as a human writer it is gonna it's getting better every day and within the next three to four years it will be better than what a human can produce and the reason for that is pretty much the foundations of computation. We can analyze and access significantly more data than what one person could ever do on their own so if you think about writing it's a solitary event one person sits down and writes about something with our technology we can incorporate multiple people's understanding and logic into our software and then over time we can improve it in such a way that it can it can be you know improved at a greater rate than what one person can possibly improve on their own and so we get to take advantage of all the benefits that you do in traditional software but in writing words. A similar I found a similar thing in what you're doing Michael because well maybe you can start by talking about how Eureka works and then talk about Newtonian because I think what you're trying to do with Newtonian is also allow people to get scientific insights in a place where they wouldn't normally be able to hire scientists to do it. Sure so I've spent the past five years or so working on this sort of robotic science project and the goal has been to accelerate scientific discovery and make it easier to understand data and to figure out relationships and the meaning behind phenomena you might deserve in the laboratory and what we're able to do is design an algorithm that can look at a set of data so it could look at say a swinging pendulum and just think about that data for a while and then in the end it comes back and says aha f equals ma and this will tell you how all pendulums work and many many related physics and things like this. But explain how it does that. So how does it do that? So it's this is a type of artificial intelligence where we're searching over a space of equations so f equals ma is an equation f equals m divided by a is an equation there's actually an infinite number of possible equations but only a few actually make sense and the algorithm uses a evolutionary approach to compete multiple equations together so that the right equation emerges and the filters out only the the most meaningful expressions until it can narrow down to kind of the single fundamental salient property of the systems so like f equals ma or e equals mc squared and so we had a lot of success with this we got a lot of press about it people got very excited so we released the software so people do it at home and the software is called Eureka you can actually download it it's really free to try out let's go to Eureka.com and you can throw your data in there you can see what the hidden mathematical properties are in the data. And then other scientists have used it and you were telling me about you know in all fields of science so you did it on physics first of all on a set of physics equations and we knew the equations you were looking for there these are things that Newton found but now people are using it for to break new scientific ground. Yeah the really interesting thing is when once people have things that they don't didn't already know about they don't understand it but they have the answer. We're kind of turning the tables on how science is going to be done because you'll have solutions you'll have answers but not necessarily the accompanying explanation that goes along with them. So like you said a lot of people have tried this out and use it actually you can go to Google Scholar and see it was around 50 papers now that have come out that have used results directly from the software and it's really exciting to see. These are things where ordinarily people might struggle for years figuring out the math behind their data or they might not be able to afford a PhD student to analyze their system or investigate their product or wherever you have. So we're trying to empower people to make them understand things more quickly and to cross the gap from being able to get the data and manage the data and understand the system to fill in in the details what is the fundamental physics behind the system and that's where the software comes in in Newtonian. Just one more question before I get to Sarah. So what is so it sounds like many scientists have responded positively to this. Have you heard people who you know are worried about this kind of software replacing the human humans in. Oh yes absolutely. So after the our paper came out in science I think in 2000 the end of 2009 and we got a lot of interesting emails and letters and phone calls from people that are very concerned about conspiracies and saying that too but like I mean were they crazy people or were they real scientists. See I can't always tell. I'm not an expert in this. What's the difference. We have one man's crazies and one man's genius. So yeah so some people are surprised but I think once people actually try it out and use it and see what it can actually do for them they actually get quite excited about it. And you think of it as a kind of a supplement to humans. Oh absolutely so we're trying to make everyone into a Newton. We want to make everyone strong enough in their math and their physics to be able to make those leaps that ordinarily they couldn't afford to do. So the interesting thing about what Robbie does and what Michael does is they both are fields that involve a lot of data. It's a lot of structured data that they can analyze and kind of get an answer. What Sarah does is interesting because she's so she's a primary care doctor. People come to her every day with problems. And it seems like you could have a computer that would just you know I could submit you know it hurts my leg hurts when I do this and then it would give me an answer. Tell me why. And in fact I've you know there are diagnosis software out there and IBM is trying to turn Watson into something like that. Tell me why we need you. So a little bit of context I am actually as we discussed before I'm actually an user of a lot of data that is derived. I do use an electronic software. Certain elements of medical diagnosis have actually been automated for a long time and the easiest one is EKG. EKGs have had automated algorithm based interpretations for I want to say close to 20 years and they work pretty well and hasn't put cardiology out of business though. So to answer the question about why diagnosis is handy and actually the faster you get the diagnosis it's kind of like having you know I would imagine having something analogous to Michael. His software using maybe like having this like a team of really super smart medical students kind of looking everything up in the journals and telling me what's going on and that would be super handy. But in primary care especially so much of what we do is both contextual and also translational because in primary care we say you know the patient is the nurse they have to walk out of there knowing what to do how to manage it. I mean you may come to me and you already know you've got an ankle sprain. I mean you know the diagnosis is uncontroversial it's a matter of what are your goals what's your past history of ankle sprains. Do you want to run in a marathon six weeks from now what is your you know what is going to be the plan around that or one of the stories I was sharing at Robbie outside in the hallway is I saw a young man recently came in and he had been having a cold for two days. The diagnosis was uncontroversial if you plugged it into a Watson it would clearly be a viral cold there he was healthy as a horse didn't smoke doesn't have asthma there was like nothing wrong with him. The context of his concern was that his girlfriend had just been diagnosed with Hodgkin's lymphoma and he came in and said I can't get her sick what can you do to keep me from getting my girlfriend sick because she's going through you know she's gonna have her immune system wiped out and have a bone marrow transplant. So that's where you know the diagnostic algorithms only get you so far and then it's the context and then the translational piece and what is the patient really there for there was no other reason why this otherwise healthy guy was in my office because he had a cold for two days. I mean one of the things that I found in in reporting on this series is that a lot of people in their work do kind of mindless work a lot of the time and it's the mindless portion of your job that is most vulnerable to robots but it sounds like a lot of your work you were telling me that the moment a patient comes in you're sort of always kind of watching what they say their body language how they say it trying to figure out if they're lying or yeah and lying is very rare but that usually you're trying to sniff out what is the real question you know in this case the example of the young man who had you know the girlfriend who was diagnosed with cancer that was pretty straightforward he was willing to share it people don't always want to share about it people can be victims of child abuse domestic abuse substance abuse all sorts of things that maybe surprisingly enough if I actually ask them the question they'll tell me but I have to think to ask that or recently I saw a woman who was you know grieving the loss of very tragic loss of her 26 year old son and she was complaining about a lot of insomnia and she was going on and on and finally I told her and I said you know if you get some sleep tonight you know Jason won't hold it against you and I had to say that and she looked at me and said I needed to hear that just now the barrier for her getting the sleep was that she was she was using that insomnia as a means of it was like a token of her grief and her loss if I could interject real quick I think one underlying fallacy of the whole will robots replace humans is this notion that there's only one right answer you know like if you have you know these set of issues that'll lead to this diagnosis I mean she's pointed out that that's not the case it's the same with writing I kind of chuckle every time a journalist asked me are you gonna automate me out of a job so what you're saying is if I wrote if one of our robots wrote a perfect story on about a particular event that means nobody else needs to write anything else about that event you know same with science I think there's no better case than you know the situation where E equals MC squared may not be valid you know so what if a robot had computed that and does that mean we should no longer look into that so science should just stop no more investigation into that because it a robot said it therefore it's true so I think you know there's a little bit of a fallacy that if a robot can do it that means humans and or anybody else should just stop looking at it well but it's not a fallacy if you consider that your software could do it at a you know fraction of the price that any human can so maybe we'll have so that I mean that sort of replaces all avenues for humans to write right I mean if the software can do it extremely well as well as a human can maybe you'll have three different flavors of the software writing three different stories but why would you want a human to do it and that maybe I'm just saying where we're at now you know I don't think we're in any danger of automating humans out of a job for most of the things we're talking about that's probably the next sort of step down the path once we've had you know several evolutions of all of the technologies we're talking about before that would even be realistic but also I'm thinking like in the case of like a biography writer you know a biography writer would write a very different story than an autobiography writer so even if you were type of software could empower someone like okay in the future everyone will write their own autobiography it would still be a different story that was told and look at all that number of biographers that have written so many different stories about you know famous people because they take particular aspects of that and look at their life through a different filter I think the the greater danger I think is when these these sort of approaches break away from us entirely for example say your algorithm starts writing articles only for other computers for example right the similar thing in for science is the the algorithms could break away and we they start doing science that we can you understand we can't keep up with but for they find it useful they continue on they they break away from us the real danger here is that you you think that's a real danger explain why that's not science fiction so we're we're getting to this point where we can generate results that we don't immediately understand we're analyzing things that are are are more complicated than we can even possibly appreciate right and we so a famous mathematician was actually on my committee Steven Sorgat said that we're actually in a very special time in the whole evolution of mankind and science technology so we're in a special window where we can actually understand the things around us most of the science all the science that's been done in the past thousand years we're fortunate that it's simple enough for us to actually understand and appreciate but you can see now especially when we start analyzing social networks and a lot of the cutting-edge technology in in physics and things like this the systems are too high-dimensional for us to understand or wrap our hands around that we're fundamentally limited and if we want to step beyond that we're gonna need these sort of automated approaches right and then it gets us into that the situation where we find results that we cannot understand and we fundamentally cannot understand because they're too high-dimensional for us that's what you mean by break away break away from us yes that's what i mean by break away and i don't think that's too far away i mean i you mentioned watson i mean the thing that was just very interesting to me was last February watching the watson jeopardy challenge where IBM software and hardware you know beat two of the best jeopardy challenge champions at a very humanistic game i mean jeopardy you know ten years ago would have thought that software could beat jeopardy champions at a very humanistic game it's already happened and it was wasn't even the close right and so that's a humanistic game that they can already software can beat humans at much less going beyond that in questions that humans can't even understand but but computers can potentially process and understand i mean i think talking to you michael and your advisor hod hodlipson the the kind of the the thing you guys struggle with is that the computers will give us the answer we won't know what it means and this sort of philosophical question that arises out of there is like whether that's an actual discovery for humans whether that's actually science if the computers understand everything about a certain system a certain set of data but the humans don't and and you guys are working on ways to have the computers kind of explain it to the humans explain what they find the humans can you describe that and whether that's possible like are we just are the computer is going to get so smart that we're just going to be like very dumb compared to them yes i think the it's crucial that that we we we focus a lot of energy and attention on not just the technology to make discoveries but also the technology to explain discoveries and explain them in a way that we can appreciate and understand so being able to connect it to our existing knowledge base and being able to say this is similar to this but it's much more complicated than you could possibly understand right it'd be certain things could be like trying to explain quantum mechanics to your dog or something like that but we still want to benefit from that so we need some way of crossing that gap and having these discoveries map back to a a demand that we can still understand and make sense of and i think that's kind of the the long-term future of science i think we'll be explaining the connections from what we know and that's the role the humans would play um i think that's where they will be focusing on over the next hundred years mostly does that mean we need Eureka for humans well i think there's a long way to go but we need uh yeah this is something that's that maybe Watson could help out with but it's something that's fundamentally human that they will need to understand us as well as we understand each other right in order to uh put something in a frame reference that we can understand so um i think that's possible but i think that's still a ways off that's that's kind of an area of air that's very slow moving at the moment so Sarah i was thinking about what you do and you know so the reason i think that part of what you do will be difficult to automate what you just talked about is um and and other jobs like yours which have this quality of face-to-face interaction and a conversation which computers aren't very good at now but one of the things that i find about technology is you know they're all it's always getting better at a rate that we can't really appreciate and it's it's hard for us to predict the future because um because it's sort of always getting faster and we don't really understand that um do you and so you know maybe 10 years ago a travel agent might have been in your position and said uh no one is ever going to use a computer to book their trip because there's too many variables it's too complicated and i won't and you know they'll always need me and they and now we don't um can you sort of respond to that argument and think and and you know do you do you hold it in your head as at least being possible that we might not need doctors like you in five years time so let me try to answer this by outrunning my headlight slightly because this is going to get outside my my area of strength but you know as as you know in a lot of other areas of you know very high cognitive specialties we're starting to see much more but what you could simplify is high of intelligence right certainly in the scientific community you know you don't really see a lot of solo scientists it's very collaborative and i think one of the when we see scientific innovation in medicine one of the things that's really hampering is all that although the scientific aspects of medicine are very much benefiting from this collaborative sort of high of intelligence there's a lot of regulatory barriers that prevent us from implementing that and so then the delivery of medicine we have two problems one is the delivery of medicine is still for a lot of regulatory purposes has to be done in this very you know one-on-one individualistic style the other sort of fundamental barrier we have is we're good at aggregating data from clinical trials and from sort of hands-off medicine so you know you could query all the pathologic specimens in your wife's lab for example and you could you could get data from that we don't have a good way of really querying a lot of clinical practices to really aggregate a lot of data points and some of that is very some of that's just kind of hardwired into the culture of medicine but a lot of it is regulatory there's a lot of you know naturally you know if you did an article about you know scientists being put out of business or or you know having to fundamentally change the way that they share scientific data you know there would be a select group of people that would in the public that would care about that and would react to it if anything hits the newspaper about aggregating patient data and sharing it in the cloud somewhere it's you know people's worries about privacy and regulation exactly in a nutshell yes exactly in the near term that is one of the things that's that's I mean that's why it's going to it has often been a very late arrival to this is because we're not able to share data so much and there's so many barriers to to aggregating no I that's an interesting point that I touched on in various parts of the series is that you know there are laws that humans have designed that protect humans you know pharmacists you in order to dispense medicine you need to have someone who's an accredited pharmacist in a location to for anybody to give out medicine there you know even like the the pharmacy tech person who makes minimum wage and and so that's you know that's a regulation that's going to protect pharmacists even if we have technology that might be able to replace pharmacist and you know give you your medicine from something like a vending machine you can't really do that because there are laws about that I wonder if you but you know I saw some evidence that might that might change in pharmacy because the economics of having machines do it are so compelling that maybe the regulations might go away do you think that you know at some point in medicine we may see people may say you know it's really could be good for patients if we aggregate some of this data we're seeing that in an emerging economy certainly in Africa and in China you know these large populated countries where the the medical and you know infrastructure is very skinny there there's a lot more of a compelling argument and you know significant reductions in regulatory barriers to implementing a lot more telemedicine and and and sort of the common practice for delivering a lot of health care in African villages is you know tribal based it's or you know village based medicine it's it is population based medicine that's the model that they're used to I was talking to a colleague of mine it's with the university school of dentistry and and he just you know kind of during dinner conversation dropped the fact that in China they're going to have this huge problem with dentistry because they essentially do not have the profession of dentistry in mainland China and it wasn't necessary because they were eating a traditional you know Asian continent diet which was not significant risk for for dental disease but they're starting to see this latest affluent generation is starting to get significant they're they're starting to develop pediatric dental disease and they have nothing to deal with it so that that's going to be a very you know you're going to have a very you know an emerging affluent population that's very comfortable with technology that is going to have a public health crisis and no really good way to scale we're not going to be able to you know train up enough dentists and dental technicians and hygienists to go out everywhere so there will be some changes like that but they'll largely be driven I think in developing and emerging economies so if we don't do it someone else will someone else will end up the robot dentist yes exactly do we are how are we doing on time should we switch to questions do okay great who has a question this question is for Michael Schmidt and I'm very curious about two aspects of what you said I'm sorry I'm James Barrett I'm a documentary filmmaker two aspects of it one is the inscrutable nature of evolutionary algorithms what is it about their input and their how they work and what they output that's that's difficult to understand and you could probably use John Koza in that answer and then the other question is are you alluding to when you talk about intelligence or intelligent machines getting out of control you're referring in any way to the intelligence explosion referring to ij good yes so in a way yeah so let me talk about what evolution algorithms are and the idea is traditionally AI has been focused on trying to simulate what a what a human brain would do it has like a memory center and it has this computational unit and it tries to connect these to sensors what an evolutionary algorithm does is it just gives a randomness like a random computer program random ones and zeros and it competes them in an evolutionary process so that they start with the population and you figure out which ones do slightly better than the other ones and you throw away half and then you recombine those and eventually a solution emerges because you can do this billions and billions and billions of times similar to how in natural evolution there's been millions and millions of generations you can do this in seconds or minutes or hours on a computer and get very very good results that's what a genetic program or or an evolutionary algorithm does is it simulates this process such that an intelligent complicated solution emerges we don't have to design it from the top down it's something that comes from the bottom up so it's not fundamentally limited by our own understanding that's the idea so it can produce things that are beyond our capability potentially to understand things that are too complicated or problems that we just haven't been able to tackle ourselves efficiently and there's been a lot of different applications like antenna design things like this so that's kind of what an evolutionary algorithm does and why it's difficult for us to understand them because they're stochastic they they kind of come up with results automatically that emerge we don't design them from the top down we provide a simulation such that they come out they arise and you mentioned also the when my alluding to I think what's sometimes called the singularity or the the technological explosion things like this so I think these are I don't like to use the term singularity because I think it's a very heavy-loaded term but closer related to that I think something very very similar where we'll we'll when I talk about breaking away this is very closely related to the singularity some people might consider that to be a technological explosion and I think it's a good thing and something we should we should push for I'm very much more optimistic than than others so we should let another question one more question I think Andrew Askelund College of Law Arizona State University I wonder in a way what we do when we use computers is we we perhaps lapse into kind of a probability fetish that you know that the programs are good at sort of identifying we use the salient features so that's what we focus on and they're good for that but it might be the case that sometimes what's really interesting is not amongst the probabilities or it's easy to imagine a program that for example writes a story about quarterback Q completing so many passes so many yards despite so many interceptions that could sort of generate over and over again but the extent something odd takes place in a game that we think is really important the computer would not would not notice that and it seems like data in generally could play that way something really important doesn't fit into the probability story and it's not just as it missed we may sort of have a tendency to to focus on probabilities and missed the outliers outlying facts outlying data do you want to answer that yeah so one of the things we focus on are the outliers because again for us to compete with the ESPN's and the yahoo sports of the world we have to create content that's different and compelling and so doing the same thing is writing about the same things they are won't sort of meet that bar and so the things actually we do focus on are trying to make sure that we look for the outliers because again that's what computers are good at i mean you say they're focused on you know kind of the probabilities that are most likely to hit but they can also be good at hitting you know looking at things that aren't likely to hit just as easily if you're willing to program them to look for that and in fact that's exactly what we do because that's one of the benefits that you get with software is you know you for the same amount of cycles that you burn trying to look for the things that are highly probable you can also look for the things that are unlikely to occur and again there's a lot of value in that that the typical human can't do because they can't process all that information um okay i think we're out of time um if you have other questions for these people they'll be probably around here for a little bit but thanks so much and we'll have another panel soon i'm farhad manju for slate v for the last few months i've been looking into the rise of robotics and artificial intelligence software in american workplaces as i was analyzing real life robots in science medicine journalism and the law one thing that struck me was how different they are from robots that were used to seeing in movies and on tv for starters on screen usually find robots doing humans dirty work look at walley he's a waste collection robot or remember those butler robots in woody allen's sleeper are you thirsty mr. Monroe me no no thank you today's robots aren't anywhere close to being able to do the kind of work that robot maids do take rosy from the jetsons she can recognize objects she speaks and understands english come and get it she has fine motor skills she can navigate through a whole apartment all these skills are very difficult for our real-life robots creating machines that can manipulate the physical environment in the way that a human can is one of the most challenging tasks in robotics and here's the irony while today's robots aren't very good at doing our dirty work they're great at doing our mental work part of the reason is that computers are getting much better at understanding language the movies have long suggested that when computers acquire speech we'll control them just by talking to them you know like the computer in star trek computers send a subspace message star fleet command security channel authorization alpha four seven authorization required to activate security channel modern speech recognition isn't as good as that although it's quickly improving but the real advantage in computers learning language is that they can start to understand documents today legal robots can search through a whole stack of evidence in a complicated trial and discover and criminating emails for example and robots can also be creative there are now machines that can write sports stories software can also be funny researchers recently created a computer that's sure to amuse fans of the office it can recognize when it would be hilarious to say that's what she said when computers are shown to be funny the joke is usually that they fail to understand human sensibilities like vicky from small wonder i'm not programmed to smile well then i'll program you too this is a smile and computers are rarely shown to be creative they're almost always portrayed as rigid single-minded bent on their pre-programmed mission there are exceptions of course in star wars r2d2 seems to understand funny when he sees it help i think i'm out this is all your fault and then there's howl from 2001 he certainly reacts creatively when he sees a threat to his mission i know that you and frank were planning to disconnect me and i'm afraid that's something i cannot allow to happen but the most interesting thing about howl is the work he does he's identified as a crew member on board the ship and he takes care of most of the ship's functions he renders the humans almost irrelevant in a few years time after the robots have taken all our jobs we may come to see that as a pretty good call this conversation can serve no purpose anymore but fine okay so our next panelists can come up so in this in this panel i i kind of want to talk about what we just ended with which is whether um artificial intelligence is a fundamentally new kind of technology and will affect the economy in a way that um older that technological improvements that we've had in the past you know didn't whether it'll sort of actually put us out of jobs or just do what other technology has done which is spur economic growth and make room for people to get new kinds of jobs so the the person who has argued sort of the the more pessimistic view is is martin ford he is a technologist and he's the author of a book called a great book called the lights in the tunnel which is you know our use i thought very convincingly that things are going to be very bad um and uh so martin is from california he's he lives in silicon valley he he writes software is that right right software and so and you know i know a thing that i found in this series is that people who are on the technology side of it are much more um believe much more that robots will come for us that than than other people and maybe it's because they understand the technology or maybe they just have a lot more faith in the computers um and then we have uh michael lind who is from here the new america foundation uh he's the policy director for the new america foundation's economic growth program from what i can tell michael believes a lot of what martin says but disagrees with some of the timing so we can talk about that um and then we have tyler crown who's a economist that many people have heard of who know well and i think that i was hoping that tyler would tell us that both of these people were completely wrong but i think he's not going to say that so we'll see um because we we chatted yesterday and he seems pessimistic as well but maybe i'm wrong um okay but let's start with with you martin tell tell me why um you know we've had a hundred years ago most of the population worked in agriculture then we had lots of uh automation and um you know great technology come about now nobody works very few people work in agriculture but the economy wasn't ruined why won't that same process work out with uh ai okay first of all let me say that um i don't really argue that everything is going to be terrible but i do think that we're going to have to adapt to our economy i do think that if if we're unable to make the necessary adjustments then things could potentially be quite negative um historically you know what we have found is that technology has advanced and the economy has always adapted to that uh uh as for had said the the example of the mechanization of agriculture is is one really good example um you know you used to have most of people working in agriculture and now you've got two percent and you know what we found is that in the short run that did cause significant disruption and unemployment but over the long run people moved to other sectors you know they moved to manufacturing first and later they moved to services but what we found historically is that these technologies going back perhaps 200 years or so have been primarily mechanical or perhaps electromechanical and they've tended to focus in single industries or single employment sectors and that's certainly what happened with agriculture but if you look at what's going on today it's really quite different we now have this special new field called information technology um it's really ubiquitous IT has its tentacles everywhere I mean it's going to impact every single industry in existence and and perhaps more importantly it's going to impact any industry or or new sector that arises in the future and and so that really becomes quite different now it's much more broad based um and uh in essence what you're seeing is that things are becoming much less labor intensive you know as jobs are destroyed in traditional areas now it's much less likely that a whole new sector is going to rise up that can absorb all those workers in the way for example that manufacturing arose when agriculture was mechanized and the reason is the information technology is going to invade everywhere at once and it's going to you know increasingly form the basis of new industries and new employment sectors that are created in the future so I think what's happening overall is that you know the nature of machines is changing they're becoming more autonomous they're moving from becoming or they're moving from being compliments to become substitutes that's happening sort of overall on the average and the whole economy is basically becoming less labor intensive Tyler what do you think of that I mean is it fundamentally a new kind of thing I do think it's a new kind of thing well let me outline a more optimistic case and we can all decide what to make of it since this is what you want when smart machines really take off there will be much much more output things will be more efficient there'll be more stuff it will be higher quality medicine will be much better so there'll be a lot more wealth that wealth means it's possible to support many people so even if wages are low in a lot of sectors if you own a pretty small amount of capital you'll still be quite well off an alternative scenario is that governments own capital to some extent and have more redistribution there's a kind of guaranteed annual income in this scenario because it's easy enough to afford it and maybe a lot of people don't have jobs in the contemporary sense but again they still do fine so as long as total output is going up which which clearly it is in these scenarios there's always optimistic corners to these pictures the other point I would make is I think smart machines will always be compliments and not substitutes but it will change who they're complimenting so I was very struck by this woman who was the doctor sitting here a moment ago and I fully believe that her role will not be replaced by machines but her role didn't sound to me like doctor it sounded to me like therapist friend persuader motivational coach placebo effect all of which are great things so the more you have these wealthy patients out there the patients in essence are the people who work with the smart machines and augment their power those people will be extremely wealthy those people will employ in different ways what you might call personal servants and because those people are so wealthy those personal servants will also earn a fair amount so the gains from trade are always there there's still a law of comparative advantage I think people who are very good at working with the machines will earn much much more and the others of us will need to find different kinds of jobs but again if total output goes up there's always an optimistic scenario doesn't your optimistic scenario require a long transition period so if if this technology was going to come about in over a period of a hundred years maybe we could sort of adjust society to that model but if it's going to come about in the next 20 years let's say wouldn't it be much harder to to get there we're going to have a long transition scenario you look at something like chess which is highly manageable highly regular it took really quite a long time to get chess playing machines to be able to beat the best humans go the best humans are still better shogi it's close you look at a lot of different areas there's medicine there's law there's economists and they're going to proceed at different paces there'll be a kind of slow gradual turnover of the economy where different sectors get turned into smart machines and people shift sectors and I don't see why it's the singularity scenario that we wake up one morning and the terminator arrives and it's like oh my god I think that's pretty unlikely Michael what do you think well I recommend martin Ford's book I read it when it came out it's the best discussion of the subject I've read and and one of its benefits is that he has anticipated all possible objections and so I was looking at it this morning and he basically said what I'm going to say it's you know one of the many cases two in particular first that he's premature and second that related to that in the short run I think that you will create new labor intensive jobs at least in the short run that will absorb a lot of the people shunted out of what kind of labor intensive jobs well here's the thing historically rich people have tended to be the pioneers for the next wave of employment I came across a great quote from Woodrow Wilson around 1900 when he was president of Princeton saying that nothing excites socialistic feelings more than the site of an automobile right and of course then you have Henry Ford everybody has as a car so if you have this process of fortism what is happening is result of productivity growth a lot of mechanical tasks are being mechanized or automated an increasingly number of intellectual tasks are automated so you would see a growth of personal service jobs that cannot be automated well we have those jobs today they're jobs of the servants of the rich right so if you take a working class person you know wins the lottery with the ticket becomes a billionaire that person is going to be surrounded by human beings right it's going to be surrounded by people doing things that working class people do for themselves or do without so there are these two categories things that you would pay other people to do you know chores and personal shopping and and you name it or things that you don't even know you have as a desire yeah you know like like you know scented aroma therapy in a three-day spa until you actually have the resources for it so the question is if the prices are falling you know from within these these amenity services and at the same time the pool of labor is expanding historically that's how we've gotten over this this sounds like what tyler saying right i mean it's it's the it's the optimistic view and so the kinds of jobs that would be there are kinds of jobs like like personal shopper where it you want i would want someone to be my personal shopper i would want a human to be my personal shopper because it makes me feel good to pay a human to do that rather than a computer i want a human's taste these are tell me why these are jobs that i that machines wouldn't do well you know part of it is demand it's simply uh my grandfather was an engineer who worked in new york in the 1930s and he thought that the automat was was the wave of the future this was a restaurant which had the first vending machines so you should go go in put in i guess is a penny and then the vending machine dropped and so his vision was by the year 1950 that's how everyone would eat right well you go to a restaurant now uh if it's a moderate income restaurant you have one person seats you one person brings you know one person uh explains the specials and so on so clearly there's you know we want to be around other people you could come up with software to design your own house but think about it right now only very affluent people can afford architect designed homes i think what is more likely that people are going to spend and also technology is labor saving are you going to spend the time to master the architecture software or would you have let's say the working class person can afford a working class architect with you know very good CAD you know uh tools uh in the long run i think martin Ford is absolutely right that is eventually if you you get incredibly sophisticated how 9000 type you know Star Trek computers you do have to rethink wage labor which is about 150 years old in the north atlantic countries and much younger before that so so that could be an interesting question too so so martin what do you think of that that suggestion well i i think that you know it's a valid argument but i i have concerns primarily on it's about the concentration of income and the concentration of consumption that we see and i think uh the trends that the you know the technology trends are going to probably amplify that i mean you're talking about jobs like personal shoppers and how many people really are there that are going to hire a personal shopper i mean you're talking about a very small number but i guess they're saying that you know if personal shoppers cost a tenth of what they do now then you everyone would get one i mean is that right is that a good summary if i earned 10 million dollars a year i would hire a person just to take around my dry cleaning but i think basically we'll end up hiring other people to cheer us up the restaurant example is a very good one to go to horn and harder is depressing you go to a restaurant someone smiles hello mr kawin they bring you to the seat the waiter waitress comes by they cheer you up a lot of jobs will be about motivation just like the doctor here is motivating her patients motivation will be one of the biggest employment sectors in this future i don't know if you guys can tell this but today for the first time so i i traveled and i didn't have a i forgot my razor at home and i went to a barbershop and i got somebody to shave me which cost me $40 and i could have bought bought a razor for three dollars but by the way on this on this point right i'm paying this person to do a task that i could have done right we always hear that the the kids should be studying stem and mathematics and whether the children are going to be competing with Darwinian algorithms you know why do that teach your children to be obsequious and friendly obsequity is going to be a great but i have to say this optimistic scenario doesn't sound very i mean it doesn't i don't want to i don't want my child to be a personal shopper like can we escape that i mean could people do intellectual work that's part of it i i'm still doubtful about the numbers me keep in mind we've got a workforce in the united states of 150 million people at least that many people available to work um i don't believe we're going to take all those working class jobs that are ultimately going to be eliminated by technology and have all those people working for a tiny number of very wealthy people no no no that's not what i'm saying what i'm saying is this only works if you have service sector fortism to coin the term in other words if you take what are now luxury services and you mass produce them like the spas they offer in the mall that's a different scenario but then that implies my fortism just let's talk what explain what you well the fortism in the 1920s was the idea that you have mass production but also mass consumption because the wages rise and the prices fall as a result of technology and and until the automobile owner a worker can afford an automobile so so service sector fortism and in different societies this may not happen but it would be a situation where the home health aid can afford the the you know modestly paid personal shopper or the modestly paid caterer but the question is going to be can they afford housing and can they afford health care and can they afford energy that's you know the bulk of people's budget so you this technology may drive down the price of manufactured goods it may drive down drive down the price of these kind of personal services but how are the people who are performing these services for a wage going to earn enough to survive in in the real world that's a very different argument because if you're really an optimist about smart machines they'll make health care education in real estate much much cheaper if your son is a personal shopper keep in mind your son's basically an owner of capital he only needs to work enough to buy some capital and earn this income stream because there's so much output so to be a personal shopper for 10 hours a week it's not as bad as it sounds i think for a barber a lot of people today would trade in for that and then you have all the people who work with the smart machines and augment their value and surely not everyone can do that but it's more than just some tiny elite so wait you're so go back to this argument about tell me why it would be 10 hours a week because because you you just have to work for a little bit because everything would be so cheap absolutely you could work more if you wanted but oh but i think i i i disagree with that you know it was the assumption of canes and various others that as incomes went up people would sort of max out and they would translate that into leisure so by the essentially we would have decided in 1955 we would be satisfied with magnivox tv's and then we would stop working 40 hours a week my guess is that some of what happens is when you know the both luxury goods and luxury services become cheap and mass produced at that point people keep working longer hours to get you know like the new bio genetically designed pet or something that's the big status symbol i do think some will keep in mind the last 20 years the trend is that only the very wealthy work more and the rest of us work less so people will decide but if people want to work less they'll be able to but you're saying you'll be able you'll because the because the machines will be doing all this work and making everything cheaper will have everything will be cheap enough that we won't have to work all the time again i'm putting on the optimal yes which was the request so okay well i mean do you actually believe it then put on the other hat you're going to get a lot of different results and it will depend a great deal on the country and most of all how politics responds one issue i also worry about distinct from this issue i think in some ways very good drugs will be quite a threat to jobs explain that if drugs are really fun and totally safe which may not be possible we don't know but you could imagine that as a technological advance it's really going to cut into a lot of this the spending on the personal servants and getting your hair done and going to the spa just stay at home and take your drugs right well rock stars may need to do all of these things simultaneously sure so that's a competing force against smart machines and in part the future will be determined by you know which which set of forces are winning that race okay maybe we got a little far so so martin i got this question a lot in comments to on my series which is why are we assuming that the machines will get really good and the humans will stay behind you know a lot of people who think about the future you know the singularity ray Kurzweil think about machines making us better and so what why won't that happen well i think it may but but you know sort of the starting point for me is is the recognition that most of the jobs most of the work required by the economy has always been on some level fundamentally routine i think i think if you go back in history you know you had people working um working the land that's a very routine job later they moved to routine jobs and manufacturing today they're more in the service sector but most people um and you know at all skill levels they come to work and they face the same kinds of challenges day in and day out with not exactly the same but within some narrow range of variation and those types of tasks and jobs are increasingly going to be exactly what is appropriate for technologies like machine learning for instance now as we fast forward into the future you know if you believe what people like ray Kurzweil say we may have you know brain implants and such that will make us much much smarter but that doesn't mean that the best way to take on a routine task is going to be to you know take a working class person and give him a brain implant to make him smarter i mean it's still gonna be the case that a specialized machine is going to be best for performing a specialized task so in order for that that kind of vision to result in employment for everyone you have to imagine an economy that creates you know 150 million or more jobs for people to do non-routine creative type tasks and that simply hasn't been the case historically i mean you know the people who really get paid to do creative things are relatively small in number and most other people have always done routine things um Tyler how do so you suggested there were two you know there were multiple different scenarios sure how do we what are some of the things we need to get to your optimistic scenario i mean sort of regular you know in government in society education things like that i mean you need health care and education to become sensible and accountability based sectors and to some extent deregulated and they're very far from that i can imagine a world where basically every child has a tutor that won't employ everyone but it will be a large number of jobs every old person has two people helping him or her out that will be a large number of jobs but right now health care and education are total messes in ways which i think are pretty obvious to us all and that risks the crack-up scenario where those things stay expensive and everything goes wrong right so so so your scenario wouldn't work if we needed to if those sectors didn't got they got more expensive as they got automated but i think the smart machines will do an end run around the regulations especially in education health care will be tougher and it will happen in other countries and people here will demand it and they'll basically just get it through the internet so on that i'm reasonably optimistic explain what you mean by the machines doing an end run around like in education like education the schools might stay terrible and require a lot of credentialism and certification but if you can just buy an education box for your kid you know at target homeschool your kid sit it by the education box and have a private tutor come in sometime in the week and pool with other parents that's going to work even if there's no real reform in education so i'm somewhat optimistic there well i i'm a bit more pessimistic on the on the political side in as much as until the last generation really a couple of decades ago you had a third of the human race living under marxist linanist regimes where the way you had access to the goods of society was to warm your way up in the communist party hierarchy you had much of the rest of the world latin america and much of east asia was military dictatorships where the people at the top were not capitalists they were not technologists i mean they were people they were soldiers and as long as you have a world of territorial states which we are going to have you're going to have different social systems different social orders and and that was you know my point about why you need to take alternatives to the wage system seriously for most of human history people did not earn a living by selling their labor in a market this is about a century less than two centuries old even in britain and united states but you had a class of farmers most of them unfree farmers slaves or serfs and then you had a class of parasites the landlords who happened to be the warlords to keep the farmers in line and to extract rinse from them so this whole experiment in everyone going out and being able to sell enough of your labor to afford the necessities of your life it was tried briefly in the late 19th and early 20th centuries and it was considered such a failure in every industrial country that it was replaced by welfare state systems even it whether it didn't matter whether the country was a democracy or a junta or uh there is no free labor market no free retirement system no free market in in a number of basic labor related subsistence goods in any modern technological society i mean if you're a libertarian you can fantasize about what it would be like but there's not a single territorial state on the on the planet that does this so it's already rigged right it's been rigged since bismar right and and franklin roosevelt and and you know perone and and you know you name it so i i think that's the question uh and just one more point briefly on this why did you have these welfare states and immigration restrictions and minimum wage laws and all this stuff it was because the political elites who were not terribly enlightened necessarily uh they feared their people right they either feared the people marching on them overthrowing them or they needed the people to fight wars right and one thing i worry about abusing you know martin's thoughts on this as you move towards increasingly robotic military where it's essentially capital intensive you don't need the peasants anymore you know uh they're not there if anything they're a drag they're not they don't add to your national strength and at the same time uh it's really difficult for highly dispersed populations you know simply to overthrow the government the way it was back when you had one capital city surrounded by a lot of poor people uh i just think you have to factor in the political side you know you can't just assume this is all going to be solved by the market that's the pessimistic yeah well i i think it's true that it's hard to be optimistic if you look at the politics of it ultimately i think the solution is going to be something that tyler mentioned earlier which is a basic guaranteed income you know we're going to have to get away from the idea that people have to have a wage-paying job in order to survive and that's what i advocated in my book but you know politically it's very very hard to imagine that in terms of what's what's happening today but it's not just politically hard to imagine that i think it's socially difficult to imagine that i mean a lot of people find meaning of of their life through work a lot of people kind of think about their work as being that's right and that's the idea that i've advocated is that we should take a basic income and then we should modify it with certain incentives most important which would be education just just as an example suppose we paid everyone some minimal basic income but if you graduate from high school or pass your equivalency test then you actually get a somewhat higher income imagine what that would do for the high school graduation rate and so we could we could build some basic incentives some of which might occupy people's time and give them some purpose into into that sort of a scheme but you know a basic income i know it sounds very radical probably a lot of people see it as a very leftist welfare state run amok type thing but actually it's it's very much a free market idea uh Friedrich Hayek supported the basic income Milton Friedman supported a negative income tax which Franklin Roosevelt and Lyndon Johnson opposed it because the whole basis of the welfare state liberalism in the united states is that a job is a right you know for whatever reason in that historical period work was central to your identity as a citizen so Johnson and Roosevelt hated paying people who weren't working you know they wanted public works right but they did pay farmers for not planting right and it's not that much was stretched to me you know they're somewhat similar but i think in the future we may have to you know involve in that direction but there simply aren't going to be enough paying jobs out there for a lot of people and and even to the extent that the new if your theory is correct that there are going to be all these personal service jobs you can see right away that there's going to be a structural mismatch and we already have a problem that we can't employ working-class men and and women are doing better and you know if you're talking about personal shoppers and this type of thing you're going to skew things even more in that direction i think Tyler you mentioned the the guaranteed wage what do you think of that idea it can't be afforded now but parts of western europe already have something not so far from it and it would enable us to get rid of a lot of other bad government policies so i think if we can find a way to get there it's a better outcome and just piling more in different parts of the welfare state on top of each other so i i don't think it's so unrealistic it is for 2011 and i think it would be a mistake to have it now but if you're thinking about 70 years from now i more or less expect it well we have it now in the sense that tax expenditures like the child tax credit or income tax credit these are really wage subsidies you know going to working people so already you have part of your income is your actual market wage and then part of it is this government subsidy so you could actually get to something like martin describes by gradually shifting the ratio what about this other problem that martin talks about which is if we do if we do move to an economy where the job that humans do is this personal service work there may be a lot of people maybe men maybe you know other people maybe people with bad taste people who just don't have the aptitude for that kind of work what what did those people do i think this new drugs drugs the new order you can already see this will favor conscientiousness as a personality trait and on average that will favor female labor i think i think there will be some decent percentage of people who not quite for economic reasons but for reasons of temperament are simply going to do very badly in this order and i think some of those people will end up living in a kind of shanty town getting a very low income with other people who with other people like that yeah and today we call those prisons or you have tent cities or in developing countries a place like brazil you just have people living in low rent areas and not getting good public services and i think that will be part of the new equilibrium for some people people basically who are not conscientious enough to simply take a job there's a rich person willing to pay them what they're supposed to do isn't that hard the wage isn't bad but they just can't do it and those people in my view will be the big losers do you think michael's right when he says that the things we should teach our kids is the obsequious the diligent and obsequious an alternative option for these equilibria is people that people have a lot more kids so it'll be a pretty cheap supply of very high quality labor people who can look after your kids and you can monitor their quality and reliability much better than you can now so i think in a world of smart machines it could be that the norm is to have five to seven children again not necessarily through traditional birth people will get a lot more pleasure from children uh not everyone but i think you'll see a lot more people doing this you already see this among a lot of billionaires or very wealthy millionaires that they have a lot more kids it's a new trend you see it in the numbers it's now a small number of people but when you get bored with a lot of other stuff you reach a certain age technology gives you greater flexibility it's like why not have six kids and and overpopulation won't be a problem because of the machines will take care of things well i didn't say it won't be a problem right there'd be an environmental issue that probably will be a problem but it's a way of thinking about where the jobs will come from if everyone has six kids and they all need nannies smart nannies reliable conscientious nannies it's an awful lot of jobs well the point i'd like to make is and we were sort of joking about obsequiousness and diligence and all of that but with every shift in the majority of the workforce from one sector to the next the workers in the previous sector despised the emerging sector as beneath the dignity of of an ordinary american at least speaking the anglo-american tradition so when you first had factories in massachusetts and in new england they could only get farm girls to do it because it was unmanly to work in a factory job now you think of the factory job as being like the he-man sort of thing but that was for women real men were farmers well and so then it took this whole cultural thing where you could actually work in a factory and take orders from a foreman without knifing him to death you know and all of this and so then you had office work right but that was wimpy right i mean the blue collar worker you know was much tougher than word cleaver going to the office in the 1950s so i i do think that there is you call it feminization or you know whatever but but there is this trend to get away from the self-reliant male yeoman who's completely self-sufficient and it's just it's a it's a cultural shock for men in particular i think and breadwinners and so on but that that's nothing new i mean that's been going on through the industrial revolution i mean it's a long term trend well one thing i'd add though is that i feel the discussion right now is taking on kind of a working class favor flavor you know what are we going to do with these working class people but we need to remember that part of what we've been talking about here is that the all this stuff is going to impact what i would call the upper middle class the educated middle class i mean we're going to see these technologies unfold and begin to really undermine a lot of skilled professions i mean we've talked about doctors and lawyers but i think the people that are likely to be hit the hardest are going to be the corporate drone type people the people that sit in a cubicle inside an organization and they don't really have any of the protections that doctors and lawyers have so as that kind of unfolds to me it it's gonna ultimately raise a question of what do we do with these more skilled people that on are should be on the path to success what do we do with new college graduates what do we do with people who are in their 40s and 50s and are perhaps impacted by these trends you know in the middle of their careers i mean in the first panel you heard you heard them talking about sports writing and how i could automate this and all that to me what what jumps out is that a lot of entry level positions are really going to be heavily impacted because those that are more routine jobs and so how are we going to get the next generation sort of on the path to success well should we weep though for the mid-career lawyers are going to be replaced by legal software i mean this is a fairly elite group you know if you go again back to the 19th century you had these very highly paid welsh iron mongers had to be imported from wales you know for the early iron factories and then you had the Bessemer steel process and then unskilled people could do it i think you're right you know and and you may have written about this too you know a lot of this the professions a lot of human capital it's going to be replaced by digital capital and software the number of people who are affected is fairly small and they're overpaid frankly right now and society gains so much from this it just doesn't seem to me to be a big well if you just looking at lawyers and doctors and professor if you don't just focus on labor income think about capital income anyone who inherits anything from their parents which admittedly is not everyone but they can just live off the capital their whole life labor is 68 percent of GDP if you imagine all that labor all those jobs going away every single one and we have more output you just have these staggeringly high returns to capital unless you have eight kids even then divide by eight people will still be doing pretty well so having access to capital income i think is the main question not what your job will be i'm martin i so i'm starting to be won over by their argument about kind of the future of personal shoppers like and nannies and nannies i mean especially if this is sort of the the you know the admittedly the most optimistic view but i mean i guess i wonder do you think that it's impossible do you think that kind of thing could happen are you wavering i'm not wavering i don't you know i think one thing we need to consider is the impact of all this on on consumption as as jobs begin to erode as income becomes more and more concentrated consumption also becomes more concentrated right now we see the top five percent of the population in terms of income doing over 35 percent of the consumption now the trends that we're talking about here are going to impact even that five percent i mean a lot of these highly skilled jobs that pay higher wages are going to be impacted all of the people that make up the the bottom 95 percent are going to even more heavily be impacted you know it's hard to see how you maintain consumption going forward as as i just want it seems to me there's a possible contradiction in what you're saying because on the one hand your argument is based on productivity growth on the other hand you're sort of when you're talking about consumption assuming we're having this rapid productivity growth right now we had rapid productivity growth in the depression and i think you're quite right about that but that means even with stagnant wages your purchasing power increases as as some of these other goods become cheaper and cheaper right in terms of you know some manufactured goods that's true but again in terms of if you look at the budget that the average person faces productivity increases don't necessarily make land cheaper right they don't they don't make housing cost cheaper at least not in the shorter run well you know elizabeth land you know yeah it depends on where you live uh elizabeth warren and her book on the middle class is something absolutely fascinating uh basically people pay much more now uh for real estate than their parents did in order to be in good neighborhoods with good public schools right and the reason they're in the good public schools is because it's pre-college right and the reason you want to go to college is you want to get the skill intensive non yet automated professional degree i wonder if you don't have have all this unravel if you no longer want your kid to grow up to be a professional then there's this kind of falling dominoes then you don't need to buy your education box by the education box yeah that's right you don't have to be in that suburban neighborhood and so on uh well i i i ask you because you mentioned education right now people most people having taught at various universities they want jobs that's their goal they're not doing this out of pleasure in learning if you don't need that in terms of the job then why would people spend more money on education well i think that that you know as this unfolds though you've got a lot of people right now that are living in homes they've got mortgages as unemployment increases you know we're going to see more of the faults i mean it's going to impact the entire financial i mean i think there's a you guys are fundamentally disagreeing about kind of the time horizon here i mean tyler said earlier that he thinks it's going to take you know not going to happen tomorrow it's going to take more time under your scenario you're talking about people right now um it seems it sounds like you know you think that this is going to happen a lot faster than they do well i think it's i think it's already having an impact i mean i i don't i mean we if you look at the last 20 30 years what we've seen is the wages have been completely stagnant um productivity has obviously increased but the fruits of that productivity increase has has really gone to a tiny number of people at the top i mean it hasn't been broadly distributed at all and yet at the same time we continue to see housing costs health care costs are rising in the shorter one what i see is unemployment increasing perhaps gradually um across the board at both skilled and and lesser skilled professions whereas the longer one impacts in terms of really bringing down health care costs or really bringing down housing costs are i think further out there um so i think we're we're going to get into a situation where unemployment and stagnant wages and and increasing concentration of income and consumption are going to occur over the next 10 years but i'm kind of doubtful that you're going to see a dramatic reduction in health care costs for instance over the next 10 years okay um we are at time at time for questions um do we have some questions hi my name is i'm one of shragi um i just wanted to ask what you guys thought that all of this i didn't hear anything about social mobility but i think that that plays a really important role in at least america's rise in america's story and you know for someone who might be aspiring to be a medical student would you say now that maybe a two hundred thousand dollar investment in their education is is a bad expenditure that they should just save that and and you know try to use their inheritance to become part of i don't know the capitalist class yeah what i mean we hear more and more about people who spend a lot on on on law school not getting good jobs is that is it a good idea to spend money on that kind of education uh or returns for lawyers are great i think they need to be much lower yet in your case with medical school i suspect the cartel will protect you long enough so i would say go for it yeah i think if you look at where we have cancerous cost inflation in this country it's in two areas where you still have a guild craft-like organization of production it it's law and medicine and maybe even the professoriate with this kind of the tenure professors and the you know the the lecturers someone is going to figure out whether it's done by government regulation or by the private sector you're going to have the henry ford of higher education in medicine and law and the prices will drop and actually wages may rise at the bottom i mean the nurses may get raises and the lecturers may be paid more even as the professor's pay goes down so so you know in that sense to invest in the medieval education for the skills of the craft that it's about to be annihilated i would think twice about that okay one more question yeah charging persistent man i would like to know to know what would be the economy system what would be the wage for the robber and who would possess those wage what would be the economic system you said yeah i mean i guess we don't talk to them but well what i've advocated is they were still going to have a free market economy but then we'll we'll adapt that with some sort of a basic income to provide support for people um if we don't do that i think there's a risk that we could at some point in the future you know decide that socialism is what is the only thing that works i mean if capitalism really fails well i i think if you as i was saying earlier about politics uh the roman emperor septima severus gave his dying advice was keep the soldiers happy at the end of the day as long as you have territorial states and the they depend on the police and the military who tend to come from you know undistinguished backgrounds you're going to have some rigged system by which at least some of the gains from from growth even in a robotic economy are going to ordinary people we'll have a meritocracy of i you and conscientiousness for better or worse okay all right thank you thanks everyone