tv Firing Line With Margaret Hoover PBS May 20, 2022 11:30pm-12:01am PDT
high stakes in the age of artificial intelligence. this week on "firing line." >> i know that you and frank were planning to disconnect me. >> from 2001, a space odyssey and "star wars." >> behave yourself r 2. >> to "the matrix" and "the terminator," hollywood has predicted the rise of the robot, both as friend and foe for decades. but with artificial intellence, ai, already telling us what movies to watch, listening to our commands -- >> alexa, order trash liners. >> -- and soon to be driving our cars, will the movies soon be reality? former google ceo eric schmid says ai is changing everything. >> we could really be altering human beings' experiences.
>> the computer scientist is now taking a hard look at whether the united states and the world are ready. >> democracy needs to have a digital update where the digital systems are consistent with the democratic values. plus new ways to tackle disinformation and climate change. what does eric schmidt say now? "firing line" with margaret hoover is made possible in part by robert grenier ri, charles r. swab, the mark rhett and daniel lobefoundation, the family foundation, and by craig -- damon button and the simmons family foundation. corporate funding is provided by stevens inc. and pfizer inc. eric schmidt, welcome to "firing line." >> thank you for having me. >> a 100-year-old who's alive today has seen many firsts.
those first include a nuclear bomb, the polio vaccine, a man on the moon, tt tube baby,he personal computer, the smartphone, the creation of many new words, including verbs like, to google. you are the former ceo of google, the head of schmidt futures, and coauthor of a new book about artificial intelligence, or ai. how different, eric, will the world look 100 years from now? >> well, we say it's going to be incredibly different because we're going through an epical change. this isn't just a technology change with artificial intelligence. it's really a change similar from the age of faith to the age of reason, which created the enlightenment. the reason we say that is we've never had a situation where humans had a human-like intelligence to help them, partner with them, and travel through the day with them.
the presence of this new kind of intelligence is going to change society in enormous ways. we'll be richer, poorer, faster, slower, happier, sadder, more anxious, and more complacent at all the same time. >> the book, you have written alongside dr. henry kissinger as well as a computer scientist. it's called "the age of ai and our human future." and you say, quote, slowly, almost passively, we have come to rely on the technology without registering either the fact of our dependence or the implications of it. in the book, you give an example involving search engines. can you share that example with me and the audience? >> well n our case, with google, we didn't realize we needed google until we had it. and google uses very sophisticated ranking and ad targeting. a lot of this technology was invented over the history with google. i think the important point here
is that it's very difficult today to imagine modern life without the internet and the social media, the information sources. google and the others we think make us smarter. but because they're so powerful, they also change the way we think in ways that are subtle. when ai comes in with its ability to target precisely what you want, all of a sudden we're going to be in each of our own little filter bubbles, where we see exactly what we most want to see and will most get us excited. this only leads n my view, to bad outcomes but nevertheless it's coming. >> bad outcomes, why? >> because society is always about new ideas and about building consensus. and the current social media players are optimizing around revenue. the best way to optimize revenue is for engagement. and the best way t optimize engagement is outrage. these technologies because of their targeting drive us to one side or the other. >> so, do i hear you, as t former ceo of google, taking a
step back as a critique, evaluating it as not necessarily a positive influence culturally? >> well, 15 years ago, we thought that t correct answer for bad speech was more speech. and what we've learned is that the weaponization of information through social media is really harmful to society. i'll give you my position. it's very simple. i'm completely in favor of free speech for humans. and by the way, donald trump is a human being, and therefore he should have free speech. what i'm not -- >> and he does. >> well, he shouldn't be eliminated from platforms. but what we shouldn't do is we shouldn't take his aggressive speech and automatically amplify it through bots and other sort of ways of having a single individual have a huge voice. what's happening is that individuals who are particularly charismatic but they can also be wrong and lying and outrageous, seem to drive out rational
conversation, fact-based conversation. that's not because the people are saying the wrong thing. it's because the algorithms are finding that speech and promoting it. >> i would like to go through some examples of ai and how it might be used in the future and get your reaction. every year, as you know, more than 1 million people are killed in car accidents. now, research has shown that ai-driven cars will become safer than human drivers. do you welcome the day when cars are truly driving themselves? >> i do. and the reason is that artificial intelligence, one of the first big wins was artificial intelligence vision is better than human vision. and furthermore, it doesn't get tired at night. it doesn't get drunk. it doesn't make the kinds of mistakes that lead to terrible outcomes. we should be driving in self-driving cars. >> i'm going to tick through a few more. artificial intelligence has already detected breast cancer
earlier than human doctors. do you welcome ai as a medical revolutionary force? >> we do. the biggest area i think we're going to see wins in ai will be in biology and medicine. >> should ai machines replace teachers? >> well, they won't replace teache. what typically happens when you talk about artificial intelligence is everyone assumes that people will not have jobs. all of the evidence is that there will be more jobs for humans. after all, we have this huge shortage of people who want to work right now. but when they do show up at work, they're going to be smarter because they're going to have this digital assistance. and let me give you an example of what's possible in the next, say, five years. it will be possible to take a digital assistant and personalize it to you. it becomes a digital second self. now, this is under your control and trained by you. now, eventually, that second self will be watching you and learning. and it will watch you toward the rest of your life.
and when you ultimately and tragically die, as w all do, it can survive as a digital replica of who you were. and who knows, maybe it can learn some new things even after your death. >> in 1997 -- and you write about this in your book -- the computer "deep blue" defeated the chess mast e, also a "firing line" guest. but in that era, programmers had taught the machine how to play chess. now, five years ago, as you know, google's deep mind created alpha zero. and this time the program taught itself how to play chess. humans put in the rules and within four hours, it had become the most powerful chess player on the planet. explain to the audience how computers teach themselves. >> in this case the team worked on an algorithm called reinforcement learn. if you gave the computer the set of game rules, it could figure out optimal play. it did this in four hours for
chess, and it did in roughly a day for the world's most complicated game, which is called go. that technology ultimately beat the smartest korean and chinese go players in the world, who were brilliant young men. and i know because i was at both of them and saw it happen. >> you saw history made here tonight. three straight wins. >> what's interesting about go was tha in the go game, which has been played for 2,500 years, it appears that the computer invented a new move. it also looks like when the top player, the top human players played chess and g with the computer helping them, their own skills got better. so, one way to think about this is that we're in a good period right now where the computer is getting smarter. it's not capable of orthrowing us yet and hopefully never will be. but it can augment us. it can make whatever you're good at even better. and i defy you to argue that that's bad. making humans smarter and more
capable, more productive, has got to be good. >> so, what we know is that these computer programs have become very powerful at specific tasks like playing chess or playing go. but human common sense is a more difficult skill to impart upon a machine. and what we know is that the next step is something called artificial general intelligence, agi, as opposed to ai. can you, for the audience, tell us what that means? >> well, today the computers and the things we're talking about seem pretty simple to me as a computer scientist, although they're powerful. the human decides what the human should do based on what the human thought was interesting. with general intelligence, the idea is that the computer can begin to fix its own objective fun function. against its own thinking, it n decide what it wants to pursue. and in the most extreme view of agi, not only will it be able to pick where it's going to go, but
it'll also be able to write code to do so. we don't know what real agi looks like. but one way to think about it, it's unlikely to be human-like intelligence because the human intelligence is, to some degree a burden, right? it's biologically determined. but a computer doesn't have that restriction. if i think of evil opponent -- let's think about putin in russia. there are some things we know about him. he still has to sleep. he still has to eat. he still has a physical life span. the computers that we're talking about won't have any of those kinds of constraints. what happens if they go off on left field in some way that's completely non-human and not appropriate and we don't even know how to constrain that? so, today we know that in the next five to ten years we're going to have incredibly powerful conversational systems. you and i will be talking to the computer. it will help us. it will be super smart. it will generate pictures.
it will generate movies. it's going to be a lot of fun and so forth. what we don't know is once it can start doing its own objectives, where does the president to go? >> even in your tone, as you talk about agi, your tone goes towards sort of the what can go wrong. you know? u have said, quote, we'll need to be strictly guarded to prevent misuse. >> i think it's fair to say that these computers -- and there won't be that many in the world because this is very expensive, very difficult. they'll e up being very similar to plutonium plants, which are heavily guarded. they'll have limits on who can use them that are very carefully examined because of the potential for misuse. imagine if i came up to one of these things and i said, i want you to come up with a drug that will kill 1 million people that are different from me and here's the parameters. that's obviously not okay. those kinds of questions will have to be banned. so, the computer properly done, there will be the computer which has general intelligence.
but then there will be a computer ahead of it which is trng to make sure it is only asked appropriate questions and only gives appropriate answers. >> okay. i want to have a little bit of fun here. i'd like you to take a look at a very famous fictional example of ai. this is the character data from "star trek" after he gets the emotion chip. >> i believe this beverage has produced an emotional response. >> it looks like he hates it. >> yes. that is it. i hate this. >> the jet sons got it wrong. cars still don't fly. but how did "star ek" do? in the sense, like, are we going to be walking around with datas in our midst? >> well, they're highly unlikely to be humanoid in form. it's much, much more likely that this digital stelligence, this digital second twin, this digital partner will be something you access through your phone or through your computer or through other kinds
of devices. what's interesting about your clip is that emotion is something that can be learned too. and you could imagine, for example, that ai systems could learn how to be the world's best salespeople. so, salespeople, for example, learn to never say the word "no" or "negative," and whatever you say, they say something which is confirming and positive. but you can imagine that's relatively easy for a computer to learn. so, is the couter being emotional, or is it just -- has it learned how to sound emotional? one of the core problems with ai is we don't understand consciousness. and we may never know, certainly not in our lifetimes, if these things have any form of real consciousness. but we'll certainly have things which look an awful lot like human behavior, especially if it has a goal-seeking behavior like selling something or telling a story or talking about love or so forth. but does it really understand the importance of love? probably not. >> open the pod bay doors, hal.
>> i'm sorry, dave. i'm afraid i can't do that. >> there's "2001: a space odyssey" to "terminator" to "the matrix" movies and tv shows that have been warning us of the hypothetical dangers of artificial intelligence for some time. and the late stephen hawking sounded the alarm about ai and what it could bring to humanity. likewise, elon musk has warned it could lead to superintelligence or civilization ending. you've called musk, quote, exactly wrong. why aren't you worried about singularity and when machine intelligence becomes unstoppable? you know, killer robots and the like? >> well, the first place, nobody to my knowledge is building killer robots right now. but if they were, we would be watching them very carefully. so, these scenarios that are sciee fiction where we end up
in a singularity and the computer outrace us and so forth, that's a wonderful movie plot. but if we ever get to that point, there will be so many people watching and worrying and so many detective systems and so forth, i'm not too worried about that. what i'm really worried about is the change in information space. imagine today we have books and we have authorities and so forth. a world which is ai-stoked will have great dynamism, all sort of new content. it'll be very difficult to tell what is made and photographed versus what is false and doctored. so, this issue around truth becomes more important. and since we don't have a uniform definition of truth, it's hard to build a system to enforce it. >> there are some companies who are using ai technology to screen candidates when they're hiring, and they use ai to monitor productivity. but the u.s. government says using ai in this way can
discriminate against people with disabilities. and similarly there have been estions about how ai algorithms use and determine who gets bail and whether those algorithms per petsch rate inequalities, specifically racial inequalities. why are you so optimistic? >> the reason i'm so optimistic is that know so many peopl are working on solvi these problems. we're talking about hundreds and hundreds of people who recognize the problem that you rightly described. we should be able to solve them with varus techniques. what i'm actually worried about are the unintended affects of things. when i look at social media as the current bad example of this, i'm very concerned that with ai, when we are -- because ai is so powerful, when it begins to affect the way we think, the way we learn, the way our friends influence us,e don't have any idea what happens to human beings. we have no precedent for this. and we need to get ahead of it.
there also issues around what are jobs like in the future? how does national security look like in the future? but the most important one is, what does it mean to be human when there's another kind of intelligence that's similar to ours butot the same? do we wait for it? do we defer to it? do we criticize it? do we view it as a lesser intelligence even if it's smarter? are we prejudice for it or against it? we don't know. >> i'd like to take a piece of that. let's talk about labor and jobs in ai in the future and how that impacts the economy. you know, you said something earlier, which struck me. you said all the evidence suggests that there will not be fewer jobs in the future but there will be more jobs. can you dive into that more for me? >> so, there's been a lot of economic research on jobs. and the consensus right now is that at least for the next 30 years, three decades, there's going to be not enough people to
fill the jobs that are going to exist. the reasons have to do largely with demographics. many of the most advanced countries have a replacement ratio below two. the solution, by the way, is immiation, which for various complicated reasons, people don't seem to want to do. so, if you're not willing to have more young people either by birth or by immigration, which we're not doing, you're going to have not enough people to do the jobs. the ai technology today is not powerful enough and not right enough to put it in a life critical decision. you don't want an ai system running -- flying the airplane. you want the human flying the airplane with an ai system giving it advice. with doctors, you don't want the system making the health decision. you want the doctor to have the ai system scan everything, tell me what's going on, give me your assessment, and i think think about it. >> you say in the next couple of decades, but what about the next decade? am i hearing that you're
actually not concerned about ai disrupting jobs in the next decade? because there is a lot of research from mit and oxford and mckenzie that suggests the next decade will actually displace quite a number of jobs. >> we're talking about humans -- so, the problem is the people will need to be retrained. so, the jobs will be there, but the right people won't be there. >> you wrote in the book, quote, societies need to be ready to supply the displaced. not only with alternative sources of income but also with alternative sources of fulfillment. what do you mean by that? >> well, we know that humans need meaning and that meaning often comes from work. and so we're going to have to find way to give people meaning in this new, more digital world. my own view is a lot of that will come from the tools themselves. the computers are getting good enough now that they can really serve to make you smarter and more capable.
and i think probably more relevant to the society around you. >> your coauthor, dr. henry kissing jer appeared on "firing line" many times and in 1975, he spoke about preventing nuclear war with the soviet union. take a look at this clip. >> in the world of nuclear superpowers, in the world in which american per is no longer as predominant as it was in the late 1940s, it is necessary for us to conduct a more complicated foreign policy. without these simple categories of a more fortunate historical past. >> i've heard you say that the reason we're alive today is because the doctrines that dr. kissinger and his colleagued pursued during the cold war. and you've also said that when it comes to the threat of large scale ai warfare, the time to
act is now, before we have real tragedy. what kanind of policies can prevent this? >> so, i was part of a commission that was created by congress called the national security commission for ai. and we looked at all this very carefully. what we concluded today is that the united states ahead of china but not by much and that the united states needs to get ready for ai-enabled conflict and security. we don't have a doctrine for how to deal with ai-enabled warfare because it will happen so quickly. our military spes its time in what is called the oodal loop where it's observe, orient, decide, and attack. that loop is around human decision making time. mutually assured destruction, containment, all of those doctrines are under enormous threat because of the speed problem. we just don't have time. and the algorithms, at least today, are not precise enough to
know exactly what they're going to do. they need human oversight. >> so, you know, i'm glad you mentioned china. i wonder, as the former ceo of google, the company that famously quit china years ago, what responsibility did u.s. companies have that have invested in ai, in china, and in the expansion of china's surveillance tape? >> what china is doing with surveillance is really a violation of the way we think of human rights. it's really a surveillance threat. and it's not okay for american threats to help there. in practice, the collaboration between china and the u.s. over the higher end of computer science and of information is going to be stopping. and the reason is not because we don't like what they're doing. it's because china does not want our liberal, democratic information into their information space.
china will literally prevent all of that infmation from getting in. that's why youtube was banned. that's why google was banned. that's why twitter is banned. that's why facebook is banned. and you should expect more of that. >> let me ask you about climate change. is there a role for ai? >> there is. >> in climate change? >> first place, climate change is not climate change. it's climate destruction. and ai is useful in a number of ways. we have funded a series of research on climate modeling, plants using less fertilizer, changing the ecosystem in farming so that there's more essentially co2 absorption. i can go on. over and over again, ai is the understanding or the way in which we will adapt these systems. we have to deal with climate change. the reason we have to do it now is that every year, the compound effect of the damage gets harder to reverse. >> the critics, as i know you know, say that tech and ai are
contributing to the climate crisis and that the tech sector's estimated 2020 global carbon footprint compared to that of the aviation industry, larger than that of the country of japan, which is the fifth largest polluter in the world. so, how do you square that? >> so, i think the tech industry -- you're correct in criticizing the tech industry. but what is not to correct is to say the th industry is not doing something about it. the leading solutions to most of these problems are probably going to come out as the by-product of the tech industry. most of the climate change things for businesses are good for business because the cost of energy is not going down. >> eric schmidt, the book is "the age of ai and our human future." thanks very much for joining me on "firing line." >> thank you margaret. thank you so much for "firing line." >> firing line with margaret hoover is made possible by --
and welcome to "amanpour and company" from kabul. here's what's coming up. he tells me his country is doomed if women are not part of its future. then, deborah lions has had more meetings with the taliban than any other western official, and we discuss what's to come. also ahead, chairman of the democratic national committee, jaime harson tells michelle martin that today's republican party is tougher on mickey mouse than on putin. finally, new orders. talk about why it's so important for young men to talk about their