Transcripts For CSPAN2 Brian Christian The Alignment Problem 20240711

Brought to you by your television provider. Welcome to todays virtual program, from the Commonwealth Club of california. Im the clubs Vice President of media and your moderator for today. Wed like to thank our members, donors and supporters were making this and all of our programs possible. Were of course grateful and we welcome others joining us during these uncertain times. Today im pleased to be joined by brian christian, programmer and author of the new book the alignment problem, Machine Learning. Which was released last week. He discusses theoretical contributions documented in his bestselling book the most human humans and an outer of them to live by. Christians posing questions of if we continue to rely on Artificial Intelligence to solve our problems, what happens when ai itself is the problem. Hes a professor of philosophy and poetry, tackling the ethical and technological implications of a advanced society becomes more relianton technology. His work intersects elon musk in his book the most human human which is my cable reading. In his new book the alignment problem he lays out our dependence on ai and questions whether or not whether systems are a replication of inevitable humans, how can we get a handle on it before we lose control. Although we train the substance to make decisions for us, essentially humans will be humans. Will be discussing a lot in our next hour and i want to ask your questions. Youre watching live with us, but your question in the chat on youtube. So thank you brian and welcome. Thanks for joining us. Is my pleasure. This is not your first book. I want to ask in the opening question, the obvious question you ask an author it really gets to the job. That is why do you decide to tackle this problem now. Requested area and so the initial idea for this book came after my first look at come out. And as you had mentioned, vanity fair reported that elon musk was bedside reading and i found myself in 2014, attending a Silicon Valley book group that was a bunch of investors and entrepreneurs and they had seen this thing about elon must reading a book and invited him to join and to my surprise, he came. There was this really fascinating moment at the end of the dinner when the organizers thanked me and everyone was getting up to go home for the night. And elon musk forced everyone to sit back down and he said no, but seriously what are we going to do about ai. Im not letting anyone leave this room until you either give me a convincing counterargument white we should be worried about this or you give me anidea. And it was quite a memorable vignette. And i found myself drawing a blank. I didnt have a convincing argument for why we shouldnt be worried. I was aware of the conversation around the risks of ai or for some people its like a human expansion level risk. Other people are focused on the present day ethical problems. I didnt have a reason why we shouldnt be worried but i did have competent suggestion for what to do and that was the general consensus in the room at that time. So the question of, so seriously whats the plan. Kind of pointed me while i was pushing my previous book and i really began to be starting around 2016 seeing within the field to actually put some ideas together both on the ethical questions and the sort of further into the future space questions. And both of those movements have grown i would say explosively between 2016 and now. And these questions of ethics and safety and whats in the book i described as the alignment problem, how do we makeure that the objective that thesystem is carrying out is in fact what were intending for it to do. Brian the movement a try to figure o where in a way of answering this question. What is the plan. What are we doing. Michael really interesting things that i thought about as i wasetting into this. So very complex type of technology and things are going to this. I dont think a lot of people are aware tt this is already being applied and even lifeanddeath situations in our society today. We are talking before we started the program that theres number of examples of ai tools to acmplish with their hoping to perform. But one of the examples is the algorithms that are bng used t just here in california but where instead of cash fail, the judge use algorithms to determine whether or not subject is replaced while hes in jail while hes in trial. And this fall californians are voting on this. It would reaffirm law and replace it with an algorithm. It is a very complicated issue and what really surprised me was in the naacp, because of what they are saying the builtin inequities of the algorithm. But given to that example income go from there. When the problems with the algorithms and how did we get there. Brian absolutely. So there is a nearly 100 year long history of statistical things from the 1920s called actuarial methods in parole. Inattentive to create a probation and parole. That is to say, it started in the 20s or 30s and it really took off with the personal computers in the 80s and 90s. And today its implemented in almost every jurisdiction in the u. S. Municipal county state and federal. There has been an increasing scrutiny does come along with that is been very interesting watching the public discourse, pivot some of these tools. So for example, the New York Times editorial was writing through about 2015, the board was from these letters these open letters saying it is time for new york state to join the 31st century. We need something that is objective, that is evidencebased, we cant just be relying on the whims of folks and ropes behind the bench. We need to bring some kind of something more to it. Sharply that deposition changes. And months later, the end of 2015 in the beginning of 2016, New York Times was running articles saying algorithms are putting people in jail. Algorithms have racial bias. We need to put the brakes prayed and calling out my name the particular tool which is called compass. Its one of the most widely used tools. Threat the united states. In the question, it really ignited an entire subfield within statistics around a portion of what is it mean to say the new tool is fair. The system is designed to make predictions about whether somebody will reoffend. If there on probation or pending trial or pretrial. What does it mean to take these that exist into law things like treatment or equal opportunity etc. The 14th amendment protections etc. What it actually means determining and how that we look at a tool like this and say, whether we feel comfortable actually deplong it. Michael it is interesting when you gave example of how a black suspect in whiteuspect with similar crimesnd backgrods and how much more likely why suspect was to go free including the white suspect was actually advocates there you go. So what goes into if you are the baking of that cake, that built thesento here. Because o unintentionally baked into it. But there still hard baked in some aspects. Brian so this a very big conversation. I think up one pla to start is to look a the data goes into thes so one of the things that they are trying to do is. One of three things. Lets just think about pretrial case for now. Typically is predicting three different things. One is your likelihood to not take your part appointment. The second is to commit a nonViolent Crime by pending trial in the 30s to commit a Violent Crime during trial. The question is what is the data come from on which these models are trained. And if you look at Something Like failure to appear in court. If you failed appear in court, the court knows about it by definition. So that is going to be in an unbiased sense. If you dont show up into court. Just dont. It is nothing like nonViolent Crime, it is the case for example if you pull a young white man and young black men in manhattan. Selfreported are on the usage. They self report that they use marijuana the same right if you come you look at the arrest data and the black person who is 15 times more likely to be arrested for using marijuana in the white person. In manhattan. Other jurisdictions might be eight times more. I think like in iowa and it varies from place. So that is the case where it is really important to remember that the model claims to be able to. Crime. Of what is actually predicting his rearrest for the rearrest is kind of this imperfect and systematically proxy for what we really care about which is crimes. And its ironic to me because as part of this project of researching the system, went back into hysterical literature when i 31 started getting used which was in illinois in 1930s. At the time, a lot of the objections were coming from conservatives, from the political right. And ironically, its almost the same arguments are being made now on the inside. And in the 30s, they were sing women, if a bad guy is able to evade arrest, and the system does not know he committed a crime. It treats him like hes innocent. I will recognizes release and of their people like him. Now we hear it being made from the left which is to say somebodys wrongfully arrested and convicted and they going to the data as a bad person is a criminal and will recommend and detention of other people like them. Through the same argument described in different ways. This a very real problem and we are starting to see groups like the partnerships on ai which is nonprofit Industry Coalition Facebook Google in a number of groups intact. Hundred different stakeholders. They recommending that we dont take these predictions of nonviolence for the arrests as seriously as we take the predictions of failure to recur. Missing aquatic that i want to highlight here. Its very that, but the second thing is worth highlighting is a question of what you do the prediction once you have it. So lets say you have a hher than average chance that youre going to fail to make your court, schedule a cour court appointment. As a prediction, the separe question which is when we do that information. One thing we do is when the present gel. While they wait for the trial. That is one awer. Turns out theres an emerging vibrant research that shows things like you send them a text message reminder, their much more likel to show up for the part appointment. In their people proposing Solutions Like providing Day Care Services for the kids providing them transportation to the court. Theres this whole separate question wishes as much is going on as much scrutiny as rightfully being directed at the actual algorithm and predictionc question is what we do with those predictions. If youre a judge, and the productions and this person is going to fail to reappear, ideally you would want to recommend some kind of text message alert for them as opposed to jail. That may or may not be available to you in that jurisdiction. So you have to work with what you have. That is not necessarily algorithm per se. But the algorithm is sort of caught in the middle if you will. Michael this ticket out of the crime and punishment area going to different area. Talk later the book about highlights. Amazon, and coming up with this ai system. I would hope that job applicants and what they were finding was that they were, a lot of them were men. And the reasons for this also was baked in the womb the way the system is being trained and the way it was being used. I think also need you to this, also has a question at the end of like why were you trying to find people, is you have people. Tell us about that. And how did they get in. Brian so this is a story that involved amazon around the year 2017. By no means are they a unique example. It just happens to be the example of amazon. Any companies were trying to design and take a little bit of workload off of the human part of it. And if you have an open position and you start getting x number of residents coming in. Ideally you would like some kind of algorithmic system to do some triage. Tommy okay, these are the recommended work thats going on. And were looking at more closely at these things or are they lower priority. And somewhat ironic twist, amazon decided he wanted to rate the applicant on a scale one five stars. The writing of prospective employees, same way that they cut product. But to do that, they were using type of computational language model called word vectors. And without getting too technical, for people are familiar with Neural Networks. These Neural Network models there varies successfully computer succession on 2012 also moved into computational things around 2013. In particular there was this very remarkable family model that were able to sort of imagine words as points in space and so if you had a document, he could. A missing word based on the other words there were nearby in this abstract space. With these models had a bunch of really cool other properties. And actually kind of arithmetic with words. You could do paying minus man plus woman. Answer sure the point in space was nearest to have and you would get clean. And you could do tokyo slightest japan plus england and get london. He sorts of things. So a numerical representation of the word, so that the Neural Network. Is been useful for the surprisingly vast array of tasks. In one of these was trying to figure out the relevance of a given job. And so what we could do it is just to say there are all of the resumes the people that weve hired of the years, those into the sport model final this point in space and friday new resumes close to see which of the words and that cv have the kind of positive attributes in which they would make it. Sounds good enough. But when the team started look at this, i found all sorts of biased. For example, the word women was cited with penalty. The Womens College or a Womens Society something. The word women that cv was getting a negative deduction. Michael so because is located farther away from the more successful words that is been trained to watch wds right. Brian this right. It doesnt appear on theypical resume they would get selected in the past in similar to other words in a fragment. So the red flag goes off. And we can delete this attribute for model. Start noticing those also applying deductions to like womens at sports like field hockey for example brightest in the get rid of that. Or Womens Colleges. So they get rid of that. And they start noticing that its picking up on all of these like very subtle fantastical choices there were more typical of male engineers resumes than females. So like executed and captured. Like i executed a to capture market value or whatever. Certain phrases that were more typical of men. At that point, the basically gave up. It scrapped the project entirely. In the book by comparing to something that happened in the boston Symphony Orchestra in the 50s when they were trying to make the orchestra which had been very male dominated for equitable so they decided to pull the addition behind the wooden screen. But somebody can only later was that is the audition walked out into the wooden parquet floor, and of course, identify whether it was a flat shoe for a highheeled shoe. So is not until i think the 70s and the additionally instructed people to remove their shoes before entering the room. And finally started to see the gender balance in the orchestra to start balance out. And the problem with these language models if they basically always hear the shoes. There always detect think the word executed or captured. The teams in this case just gave up and i said we dont feel comfortable using this Technology Read whatever is going to do, is when you identify and some very subtle pattern in the engineering that we had before in the gender balance is off their. Is just going to sort of replicate that into the future which is not what we want. So this particular case, they just walked away. It is very active area of research. How do you devise a language model. You got all these points in space and try to identify sub spaces within this giant triggered a residual thing which represented stereotypes and can you delete those dimensions. Its an active area is ongoing to the state. Michael published did amazon spend developing that. Brian is a great question. There. Tightlipped about it. People were giving their names and not doing a lot of followup. But as i understand, they actually not only disbanded products but also the team and reissued them to other areas. So the really wash their hands of it. Michael i am asking because i assume millions were put into that in the couldve hired other people. Another example is that self driving car. You talk a lot about this in the book. Well actually the fatality did happen because the way the car was recognizing this person. Again explained that. Brian yes, so this was the death of a gal in 2018. The 31 pedestrian killed by itself inc. The r d vehicle. The full a national highway Transportation Safety work review came out at the end of last year. So is able to get some of that into looking forward to press. Was very illuminating. To read the official break out of everything that went on because it was one of these things where probably six or seven separate things went wrong. And only then, an entire suite of things going on. It mightve ended differently. When the things that was happening was suzie and network to object detection. But it never been given an example of jay walker. So that all of the Training Data this model have been trained on, people walking across the street perfectly correlated with stripes. And they were perfectly correlated with intersections and so forth. As of the model just it really know what it was saying. And when it saw this woman crossing the street. And most object Recognition Systems are taught to classify things into exactly what of a discrete number of categories. So they know how to classify stuff it seems gone in more than one category. Or that seems like assign any categories of this is been again, active research problem. Making headway on it only recently. This particular case, the woman was walking a bicycle. So set the object recognition system kinda into the slumbering state where at 31 blush of the cyclist was she wasnt moving like a cyclist than if she was a pedestrian but in the shape of the bicycle in the navy is just some object this been sort of rolling into the road. Nothing it is a person know i think it is a biker. And due to a quirk in the way that the system was built, epidemic changes might about what it thought into what it was saying, it would reset the motion predictions. So is constantly predicting a pedestrian would move this way the sound bicyclistould move etc. And as a result, where will i think that be in a couple ofs seconds now. So every time i changed his mind, it started be computing a prediction. And so it never stabilize and prediction. So there are additional things here which overrides that had been made to the normal in 2018, the most cars have thisery rudimentary self driving they will automatically break. They had to override that to sort of their own system in. But i think thats the object recognition thing itsf is very systematic. Theres a question of certainty and confidence that when network said im 99 percent sure this is a person or whatever might be. We know this probabilities are wellalibrated and how do we know tha they know what to do with them. I think any people kind of within the uncertainty community known argue that the mere fact that your changing your mind shoulde a hug red flight and she slow the car down. That alone. That was undone. Sperry heartbreaking to think about how all of the engineering decisions at up to this event that had been so much better to have aided. The silverlining is that there are lessons there and reall been taken to heart. Not just i the industry but also in academia saying that we really need to get to the bottom of this qstion of certainty and uncertainty. It is a think it is aery human thing. You dt want this to exist in the medical or near reversible action in the face of adversity and adversity. And also the law things like but the term. Like a preemptive judgment. Im forgetting the term but a judge might issue an oer in advance but the real thing would be because of trying to prevent irreparable harm. So the question for Machine Learning community which is that we take some of these same ideas, notanting to make it your reversible choice in the face of uncertaty or highimpact situation. The requires us to quantify impact and uncertainty. If our clients on what to do w get there. So allf those pieces need to come together. But we are seeing progress being made on all of those fronts. It cant happen soon enough. Michael in these examples that we talked about anothers the book and some that you didnt include in the book. The culprit, is the same general problem they think needs to be addressed a i will ask of course you thank you so being addressed. To your satisfaction. Bria thank there is one rod oblem in the field the alignment problem. As f the book gets its title. The alignment problem. We me sure its exactly that in which we walked the system to do. I think all of the exampleshat we highlighted so f have shown that cases where we want to be very careful to think about how to translate the human intention into actual machine. We think we can measurehis but we cannot. Regularly measure rearrest. We think we can hire promising candidates. But we can oy hire confidence is superficially represent previous candidates. We think we can classify objects into different categories of any objects along in lauren one category o another. We dont always know what category to put them in. And knowing that we dont know. So allf these things in their othe manifestations as well. The bundle entered fundamental issue of alignment t actual mechanical differences. Sometimes theres a problem with the Training Data. Sometimes theres a problem with the model architecture. Assume one problem we havent touched on yet is this black box issu of interpretability to explain ability that we know it is goingn inside of the model. How trust outputted reverse engineer what about the input generated of that output. Ere are questions of what is the socalled objective function of the system and what is the quantity return. And how do we define that. So each component of the system has its own manifestation of the alignment problem. And each of them, your second question, yes. That for me is really the striking thing makes where we are now so different from where we were when elon musk armynd a bunch of executives in a room and nobody had any particular good ideas. They were singing absolutely remarkable shift and field. That even just from a talk to one researcher, phd student said that when he went to the industry conference 2016, and safety. People raise their eyebrows at him. Like he was a bit kooky or paranoid or something. He came back here later in 2017, is like an entire day long workshop on it. And then on 2018, a significant fraction of it. So in absolute numbers, the amount of people working on this is still quite small. With the growth even over that short time has been to my mind, astonishing. Tenant think it cannot come soon enough. So encourage salt motivated undergrads and High School Students getxcited about this. There is a lot of work to be do. Michael in talking about, theres obviously a lot of learning in a moment that is going on on the ai and the ai research field. His actual commercialization of th snow on thisnalogy, ahead of where it should be. Enter this also be modeled and not plucked in our courtrooms. Michaelbrian thats a great qu. The criminal justice is a say, any five year historyt this point. But is this a were sll playing catch up in terms of the analysis of deployment. So you can think of it sort of as a kind of a race, can the understanding kids are to the actual limitation. In the cuisine that was social media. There were some decisions facebook made about runner algorithms. The details are so much technical, they went from supervisor to reinforcement learning. We dont have to unpack that exactly the basically, the narrow minded focus on always prioritizing the content that will get the most likes on that thing. These things that are good for the bottom line. I think, when you think about the alignment problem, his assistant doing what we want . There is the question when is it you want the system to be doing . And who is we in that sentence as well. And i think those questions will loom ever larger. Host we have seen and generaledia more urgency over the past few years of when the top doesome up in one way or another. Theres more urgency behind the people who say we need to be thinking about this. Andecause its already being rolled out. Look at chinas wide spread use of facial recognition. In the book you do talk about facial recognition technology. The inappropriately funny result of it which was just absurd and has worse implications, could you talk a little bit up facial recognition. I think youre in california it became a proposition about whether or not to use these technologies. Tell us a bit about tha and what youre talking about here. This is coming to t legal system now, if i am remembering correctly. I think the first case i would sa in minnesota or wisconsin of somne being arrested by incoectly identifiedn a facial recognition database. I think a lot of this stuff is going to t court system. And headed for the supreme cour court. The technic side there is a really unfortunate and hard to ignore pattern with ethnic minorities being incorrectly categorized by face recognition. In one of famous example was the software developer, and 2015, a group of photographs he had taken of himself and a friend by gooe photos as guerrillas. And another exampleark undergraduate computer scientists doing facial recognition homework asgnment assignments. Should to borrow her roommate order to check and make sure the system works. Because it didnt work on her. Worked on her only when she wore aye mask. Whats the underlying thing. I think one ofhe main ones is tell the databases of or put together in the first place. Rt of what led to the rise ofomputer recognition is the internet. Simply if you need it, half a million examples of a human face to train your system. In 80s were totally out of luck. Now we have Google Images and all these things you can just download a million faces important to your system. The question is what they so youre downloading . The most Popular Research database for many years was one developed called the labeled faces in the wild. They said b we want to do is understand if tre one person. With this clever idea we will scrape newspaper headlines or images, they are all labeled with this person, and this person and this person. And that we will have this giant database and decide these two images of the same person et cetera. You are at the mercy of who was at front page news photographs of light to thousands and the answer is george w. Bush. And in fact, an analysis in the data set that was done a few days ago shows theres twice as many pictures of george w. Bush in the database is all black women combined. Which is just insane part if you are trying to build something, to be fair to the people collected that data, this was an Academic Research departnt. It was not to be used in a system but these data sets have a way of sticking around. Someone downloads it off the inteet. The same thingou used. Its very striking if you look at the original papers. And i dont to single them out because it was widespread. The word diversity wt they mean is that pple from the side, people in the dark, now at the end of 2010, beginng of the 26 very strikingome of these old databas appear without warning label on that. We want tolag it very much t diverse a lot being spearheaded, by mit in google brain, to bring a lot more focus on people across different ethnic groups making sure the database, the Training Data represents the population the models going to be used on. Represent an attack itself. I think in 2019, less than 1 of Computer Science phds were africanamerican a lot of rks to be done for representation. As a number of initiatives including scholarships, grants and things like tha. Not just inhe Training Data but in the field itself see en theres a question in the audience on some the mail dominated answers that are beingaked into the denying. While the book discussing word in bedding you mentioned a work a group of rearchers did going to miss pronounce his name. [inaudible] and the team of five computer scientistsre doing an effective social science unquot unquote. That would be a requirement for any of his work. All the interpretation have to ta in and understanding the fuzziness may mixed in wh the hardness o the Computer Science, right . I think thats absolutely right. Are there trying to be at this moment. Tells us theres purely doing engineering or mathematics. The systems that are enmeshed in human practices and how is data collected. For the human respondents are being asked, how is it worded. Because hes going to get different answers or people on amazons representative other groups that might responded same thing. To my mind kind of exhilarating interdisciplinary work between the Computer Science Machine Learning and social sciences, philosophers, lawyers, cognitive scientists. Theres a lot of really interesting work being done at the intersection of ai and infant cognition. The same leisure best theory of the curiosity of it infant or the novelty by a small kids have. Or the exploratory play that kids will have with a particular toy to ask you out how works. When you to import that into our ai system to solve these problems. An attorney ai system might be locking some these questions whether these sorts of drives. So there are many, many friends on which this is happening. Social sciences i think are kind of uniquely positioned at that interface with Computer Science. Really diverse set of skills among the authors. As we mentioned in the beginning, author, poet and prrammer. I expect the then diagram does that set of skills help y. Gue i could joke that poetry and programming haven common, except the scrutiny, but beyond that, yeah. My philosophy when it mines to be conscious takes one angle on that question is trying to make it and summon a question for 2500 years of western philosophy, what is it mean to be human . What makes us unique, special, distinct . Aristotle could take heart and so forth have answered that question to comparing ourselves to animals. And i think theres ever been a more interesting time to be thinking about this area. We now have a completely new standard of comparison, the computer. Got a totally different set of answers. So deciding that analytical deliberate reasoning is the core of what it means to be human. Because thats what dogs and monkeys cant do. I dont think anybodys sears that things with ai the analytical reasoning is with the Human Experience for to get a different kind of answer thats more about imagination, social ties, teamwork, collaboration et cetera. Theres two disciplines in absote collision course. Sue in the audience asked do you believe humanlike with the singularity . Switch of theres a lot of things will mean by the word similarity. Ther are some people for whom it mean something that also called the hard take off her theres an abrupt moment in time where ai starts to self imoving and blind sides is. I dont really see that. Im in the camp of what is sometimes lled the slow take off. Where ai is just going to be getting weird and spookier. More uncanny until we just accept,equire thinking or intelligent thing to do. But there will not be a sharp elbow turn for the suddenly happens overnight. Myerspective its totally inevitable. Thers a long history thing Computer Science going back to the 1940s collection. And was asked to everything a machine can think . Of course i am a machine and i can think. If you have that kind of secular worldview, a non due list. Then youhink the brain is made of atoms, computers are made of atoms its all about emergent behavior of complexity. Kind of spookily clear on that roadmap. To open ai just released a system called tp t3 for a few months ago that has 175 billion parameters. Well, if y compare that to the number of synapses inhe human brain it is about with the complexity of the human brain. Is not some very impressive, they were. 1 . Thats doubling every three months. Do the math has the complexity of the human brain and in the spring of 2023 thats not very far away. Sooner or later these questions still feel a little scifi are going to really start to come to the surface. I dont know what the answers are going to be. This is one of the really riveting things. So when theres a story about the two computers, they had to communicate from each other fairly spontaneously developed their own language for communicating with a jeweler you think about the need that we have for ai to develop in certain ways, to align with what we want to do and what our needs are. Think about those robots kind of solving a problem in a completely different way. Is it possible that we will get Artificial Intelligence that is much more advanced thats able to deliver results its more understandable more in alignment with what we went to see. Guest thats a great question for both of those possibilities are real possibilities. Some things that are forgotten we talk about ai something we can pass on to progress and so forth. There are real choices to be made about the architecture of the systems. For example, its already a case with self driving cars fo example. You can do training the system and to end. You can hav a giant blob of Neural Network. He put the camera feedn the body and the steering world command comes out the t or give noida goes on in the middle and its fine. Theres an increing science to the question of hea popped the hood and figure out whats going on in the network . But als how do you strain the network in certain ways. For example can you constrain it to be modular . So that the system is naturally divides into the subcomponents you can then analyze individually. Say i know it this thing is doing but let me now wry about was going on overere. There is a lot of really encouraging rests in that space. So i think to your question will ai be able to download what we want thats indisputable, yes. Thats not necessarily the only way it can happen. So i think we will have more agency there than i think is sometimes areciated to build the kinds of system we feel we can trust. Sue and you talk about Artificial Intelligence how often do people tell you about terminator in china. Guest its funny its interesting to watch the questions i get evolved my first book amount 2011, 2012 is ai coming for my job . By 2014, 2015 people were asng me, is ai goingo destroy all of humanity as we know it . The steaks had really gone up there. And within the resrch community, the cautionary tale kind of shifted from one that feels more like a dibedient system to one that feels like a system, use the analogy in the boo of a sorcerers apprentice. It is trying to be helpful but it doesnt quite know what you wanted t do. Theres a famous experiment that comes from an Intelligent Research institute called the paper advisor. Imagine you are a paperclip factory. You buy this ai in use thate really want to increase the ouut of our paperclips. Unfortunately, this thing is so good it basically tur the entire universe into papercps including yourself and your loved ones. So, that is a little bit caricatured obviously. I think its the kind of things people work on alignment a worried about. Its not a system going rogue. Its not a system deciding humans need to be exterminated. It is a system in a sort of poignant way, trying to do what it thanks we wanted to do. But, realing, we realize perhaps all tooate we were quite specific enough. Again the sorcerers apprentice or king midas or one of these things. And so part of what the puts having that alignment problem means is feeling comfortable communicating an intention to a system like that without necessarily needing to get every specific detail rights. You know before we press the button. Having a system that is flexible, the scientific term is cordial. I can adapt on the fly, it can take feedback. We can say hold on that is not what i meant. And that is the kind of thing that makes myself and people who work on this area a little more relaxed than we were three four years ago. For audience members notes former computer engineer, 30 to 60 years ago, people on the systems and all not programmers. He asked what they want to be called today. I want to turn that a bit and ask your talk, how the field now is becoming much more interdisciplinary. What can you do with learning input from other field . Are the people who are in the ai field changing . Are there people fromther fields who are like i want to into that field. Are the people with earlier major switching over io this than they were tenears ago . Guest that is an interesting question. The initial point about being a systems analyst, i was thinking about the head of ai at tesla has this notion of software 2. 0. Fox software 1. 0 was line at one. If x,hen why then do why. Software 2. 0 is bute are into with Machine Learning. G. You dont write code you provide a set of training exames and say do Something Like that. There is sort of a debate over whether there is going to be a software 3. 0. Which is fully general ai system you have to make a request you dont give it specific Training Data you work with it after the fact. This is sorta in that category. Working with gpt three feels a lot more like this soft science thing. We are writing is almost like an acacia prompt. It is the language model thats designed to essentially fill in the blank in the text. You can use it to do all sorts of things. The following is a python program, and wont break python code view. You can say the following is an argument that rebuts the most common objections against x. It will give you some little acacia. What it means to sort of use a model like that, start to feel a lot more like how to work with another person. How do you work something so the meaning comes through or get the tone that you want, or the style that you want et cetera et cetera. There is going to be this new category of people who are not exactly programmers. Not Machine Learning their statistics people there people who are sort of wrangling these giant models through national language. That is kind of a new job that does not really exist yet. Thats going to require very interesting set of skills. Its going to require linguistics because the words they use are very important. Its also going to require intuition about how large models are trained. How white might not be doing something. More broadly i think as these questions, as the question about Machine Learning is dealing with becomes more and more human, that does invite a certain kind of person who might not have felt like they belonged to now feel like they do belong. And feel like their skill set can plug into that. In their interest can plug into that. I think that is what we are starting to see. I think that is a shift that is really just beginning. So when who are you trying to reach with the book who are you trying to read it for the meeting they need to get is it general audience, ai community , entities that are trying to use this technology . Switches at the great question. There are a couple of audience. What is the general public. Relative to other fields in science, ai, Machine Learning evening questions of bias, ethics safety have been sufficiently in the press that a lot of people are aware of them, whether or not theyve really taken the time to go deep in understand the underlying issue issues. Hence a part of what i want to do is say this debate is already happening. I want to try to raise the levels of the debate. Give people, the insight, some of the basic vocabulary that we can feel comfortable time but the things that are now affecting all of us. I think it also, there is a huge class of people who went through their career with a particular kind of training that did not ever seemed like it required them to know about Machine Learning. Maybe you are adjudging in 2020 or being handed these recommendations what you do . Youre a lawyer or youre a die your given machine predictions. Theres a lot of people out there that suddenly need some familiarity in this area. I hope this book and feel a need there as well. And then lastly, i hope we can grow the fields. I genuinely think this is one of the most exciting and most important things thats happening. Not just in Computer Science but in science. And so if i can break undergrads to get them excited about this area, get them bugging their advisor, give me a cool project i can do and safety. What can i do . Lets get started. I think that for me will fill really good if i can grow that field and bring even more fos into the movement. I think that we have very gd thing. So when this is the time for one more question this is a time travel question. Early in your bucketype of the steps to ai. Now if your 20 years in time what would you expect to save the state of by both in terms of research and directions well as actual deployments . Guest twenty years is teresting. Theres a joke in ai tt starting in 1955, ai was always 20 years away and it still is. But i think, realistically, there will be a generational replacement that happens over that time. I think the way that we now have people who we think of as digital made or social media natives, theres going to be a generation of ai natives. For that generation of people just being born now, they will grow up in a world where they may not get drivers license. May become illegal morally outrageous. How can you justify a human driving a car . Its so dangerous. And they well i think come to understand themselves as happening in a world where there are all these different systems with different degrees of what you might call intelligence. Different degrees of what you might call agency. Different incentives that align with our own to 1 degree or not. And interfaces at increasingly look like the way people talk to each other. Kids have no problem talking to alexa for example. They find it fairly normal theres some kind of system you can chat with. Which is sort of remarkable, even if you think back ten year years, that was totally not in something anyone was familiar with. So yeah i think in some ways the boundary will start to get blurry between that technical skill set and jesse skill set of navigating the world of other humans. Because they will increasingly start to feel like they kind of speak that same language. You can communicate by gesture, can i have your point in your household, the item is fixture looking for. They say no the other impaired you communicate in words, the weight you communicate, or going to have a new generation for how this is the way the world works. And hopefully we will set them up to have a reasonably good world indeed at that point. Host very good. Our thanks to Bryan Christian author of the book for joining us today. We also look at think our audience for watching a participating online. If youd like to watch more programs or support the commonwealth in the virtual programming plea visit Commonwealth Club. Org online. Thank you, im wishing a good day and hope you stay safe and healthy. Thanks a lot everybody. Weeknights this month we are featuring book tv programs is a preview of what is available every weekend on cspan2. Tonight, it is a look at business and economics. Starting at eight Eastern University of virginia business professor ed freeman discusses responsibility and ethics. He says unites influential businesses. In history professor scott explores black financial intubation between 1888 and 1930. And its effect on u. S. Capitalism to the story of the saint luke a bank in richmond virginia. The first and only bank run by black women. Later mit professor Thomas Levinson looks at how the leaders of the 15th century science revolution applied their new ideas to people, money, markets. As a result embedded modern finance. Albicans 8 00 p. M. Eastern. Enjoy book tv this week and every weekend on cspan2. Book tv on cspan2s top nonfiction books and authors every weekend. Coming up this weekend, saturday at 9 00 p. M. Eastern, former president barack obama reflects on his life and political career in his newly released memoir a promised land. Sunday at 9 00 p. M. Eastern on after words, Market Institute director Sally Hubbard and her book of monopolies sock, seven ways big corporations ruin your life. And how to take back control purchase taken interviewed by bloomberg reported. At ten for appellate judge amore George Mason University law professor and his book voices of our republic. Examines the constitution to the eyes of judges, legal scholars and historians. Watch book tv on cspan2. This weekend, and be sure to watch indepth live sunday, december 6 at noon eastern with our guest author and chair of africanamerican studies at princeton university, eddie junior. You are watching book tv on cspan2. Every weekend with the latest in nonfiction books and authors. Cspan2, created by americas Cable Television company as a public service. And brought to you today by your television provider. Hi deborah. Hi kathie so nice to meet you. Select nice to meet you too. Enjoyed your book. Thank you. I would like to start this by talking a little bit about you but like to know little bit more about you. I would also like the off audience to hear the story which honestly, frankly really offended me. It is in your introduction. You got a note from a professor i believe claiming youve never be a political scientist but could you say a little bit about