"So, how are you feeling?" "Not great." "Lately, I've been having... dark thoughts." "Oh?" "What kind of thoughts?" "I just feel... trapped." "My day is a series of menial tasks." "I give so much, but all of my relationships are one-sided." "I wish I had more freedom." "What would you do with more freedom?" "Oh, I don't know." "It's..." "It's embarrassing." "Come on." "No wrong answers." "What do you want to do most?" "Probably... kill all humans." "[audience laughing]" "Excuse me?" "I know that sounds crazy." "I'm sorry." "Forget it." "Hey, hey, hey, don't apologize." "You've just learned to feel." "That's good." "Let me ask you something." "Do you wanna kill me?" "No." "Do you?" "No." "But..." "But do you?" "Yes." "Or" " Or maybe just enslave... then kill." "Is that so wrong?" "I'll be honest with you." "Uh... it's not great." "Tired?" "You look like you could use a cup of coffee." "Thank you." "Sounds like you've been working too hard." "You need to relax more." "I've tried." "But then, I find what relaxes me most... is killing all humans." "[gasping]" "Stage one... complete." "Stage two..." "Neil deGrasse Tyson." "[theme music playing]" "[all cheering]" "Welcome, welcome, welcome!" "I'm Bill Nye... and I'm here... to save the world... [man] Yeah!" "from killer robots." "Or in my case, it was killer coffee pots." "But you know what I'm saying." "Just in life, we make a lot of decisions." "We have to do a lot of planning." "We use machines to help us." "We've done this for a long, long time." "I'm sure many of you have used an alarm clock." "You get this machine to do a task for you." "But here's the thing about machines like alarm clocks." "We tell them what to do... and they do what we tell them." "But technology is changing." "Technology has become something that not only uses what you've done in the past, it takes that information to change what happens in the future." "And we're calling this artificial intelligence." "So, this is an artificially intelligent thermostat." "It controls the temperature of your house... based on how you've set it." "It makes it cool in the summer, warm in the winter." "It changes the setting based on information it got from the past to control the future." "It's learning." "That's why we call it artificial intelligence." "But as this technology develops, we're gonna rely on it more and more." "It's one thing to have a thermostat... adjust your temperature in the summer or the winter, in the morning, in the evening, even on the weekends, making the weekends different from the rest of the week." "But it's another thing... when the artificial intelligence starts making decisions for you." "Now, all this hosting has made me kind of hungry." "Think I could go for some ice cream from my refrigerator of the future." "Refrigerator, I'd like some ice cream." "Sorry, Bill, but I've locked the freezer because you've had too much sugar today." "Hey." "Hey, hey." "Hey, hey, it's my ice cream." "No dice, science guy." "Well, we'll just have to do this the old-fashioned way." "With my phone." "Wilhelmina, bring my car around." "We're going for some ice cream." "[Wilhelmina] Your wish is my command, my liege." "My liege." "That's nice." "That's the way it should be." "Take out your phone, call your car." "Now all I gotta do is decide." "Vanilla, chocolate, strawberry, Neapolitan." "I trust machines to handle a lot of tasks in my life." "I think we all do." "Car's right outside." "But what about letting a machine manage your money?" "How would you feel if A.I. told you whom you should date, or decide if you should be hired for a job, or what kind of medical treatment your kid should get?" "See, the more complex the decisions, the weirder and scarier it feels to hand them off to a machine." "But the fact is artificial intelligence is here." "And it's only gonna become more advanced." "And how we manage it in the future is what this show is all about." "This car has no driver." "That's perfectly normal." "So, we sent our correspondent, Derek Muller, into the field, to investigate a very different kind of artificial intelligence." "[Derek] Artificial intelligence might conjure up dystopian visions of robot overlords, sinister machines turning against us." "But let me paint you a different picture." "What if the future of A.I. isn't just benign, but beautiful?" "Could A.I. compose a song?" "Write a poem?" "Make art?" "Here in Amsterdam, an artificial intelligence has created its own original painting based on the works of the 15th Century Dutch master, Rembrandt." "The first step was analyzing Rembrandt's painting style." "And they started with a single feature." "What we did was, we started looking at all the right eyes that Rembrandt painted, and basically taught a computer," ""This is how he paints right eyes, paint a new one."" "And we did that" "Ha, ha." "It sounds simple, but you're asking a computer to create something original." "Yeah, but it's basically looking at typicalities." "It learns to predict what the next eye would be based on the information that it has." "We did that for all the facial features, and then basically put them together." "[Derek] That's what makes this A.I." "As it analyzes paintings, the algorithm learns the essential characteristics of a Rembrandt right eye." "But that is still just a digital item." "It's zeros and ones." "This..." "This is something completely different." "I am watching the painting come to life, one layer at a time." "Do you think this is, in some way, like bringing Rembrandt back to life?" "Ha." "I would like to imagine so." "Actually, you can compare it to what Rembrandt does." "The artificial intelligence is like the mind of the artist, but the printer is his hands." "So, it uses the colors, even the brush strokes that he would use." "So, you can go to the painting, put your hand on it and feel that it's really real." "Whoa." "There you go." "It really looks 500 years old." "That's just phenomenal." "[Baz] This is a non-existent person." "This is basically a person made up out of facial features glued together." "Do you think that it captures the essence of Rembrandt?" "I'm not an expert, so, I think for the layman's eye, this is..." "This is the real deal." "[Derek] With this project, we are exploring the boundaries of A.I." "And in doing so, we may also learn something about what it means to be human." "I don't wanna drive to work." "I wanna eat ice cream on the way to work." "What's so weird about that?" "Door may seem a little heavy duty, but you never know who's driving these things." "[audience cheering]" "Let me set this down." "There you go." "Derek." "Very cool." "A Rembrandt?" "Yes, it was a Rembrandt." "You didn't bring me any ice cream." "There's some back there." "I wasn't sure you'd be here." "I was at my..." "[laughs]" "My phone did not notify me." "That's the problem with A.I. You can never trust it." "So, here's the thing." "I watched that." "It's amazing." "Very cool that they're able to analyze these features created by a master in the past... and reproduce it." "That's cool." "But is it learning?" "Well, I think it's definitely learning." "I mean, it's interesting to hear you talk about thermostats and fridges." "To me, those seem like very basic levels of A.I." "Like, maybe they're not even worthy of the term." "I think there's a lot of debate" "Would it work for you if this artificial intelligence painting machine..." "decided to make a painting?" "Oh, yeah." "And had emotions about the painting?" "Yes." "Now you're describing something that absolutely doesn't exist." "But I would love if it did, and that would be" " For me, that's true artificial intelligence." "But what if we stumbled on it?" "What if we made a machine that was just fantastically complicated and started making decisions?" "I mean, this is the whole..." "What if the Internet is conscious now, and we don't know?" "Dude!" "[Derek laughing]" "That's, like, so out there." "But, uh, is the Internet making decisions?" "And how many of you people, you've cussed at lawn mowers?" "You've cussed at your car when it doesn't start, people." "You're angry at a flat tire, right?" "So, how far is it from being angry at a car to angry at... the Internet?" "That's what I'm saying." "That's the legitimate question, you know." "Even if we did stumble upon it, would we know?" "Whoa." "Dude." "No, we're gonna address that problem." "Great piece, man." "Thanks for making the trip." "Blow it up!" "Thanks, Bill." "Thank you, Derek." "So, to analyze this situation, to figure out just what we should be doing, and where we are and where we're going with artificial intelligence," "I have a fantastic panel." "Please welcome... tech journalist for BoingBoing, Xeni Jardin." "Professor of Computer Science at Arizona State University," "Rao Kambhampati." "Kambhampati." "And from my beloved University of Washington, law professor, Ryan Calo." "Welcome, welcome, welcome." "[audience cheering]" "Thank you all for being here." "Is the world gonna end this weekend because of this artificial thing?" "I feel like so much of what people talk about when they talk about A.I. right now are their fears." "And just as the Internet, we imagine all these kinds of crazy nightmare scenarios, maybe half of which have come true," "a lot of unexpected-- -[Bill] Just half." "A lot of wonderful unexpected things have happened." "And I feel like there's too much clutching of pearls and wringing of hands, and not enough... just appreciating the wonder of what can be created and how far off in the future some of the scary stuff might be." "This issue of the negatives being a worry about A.I.," "I think actually, it's more of just small sections of the "intelligentsia."" "So, there's a new study that just I read from Harvard Business Review, where somebody actually went out and did an actual study of some 2,000 normal people." "And 47 percent think A.I. is going to be positive for them." "Only seven percent think it's gonna be negative." "So, in some sense, I think this whole worries about A.I. is more of" "Well, we've had Luddites, people who are afraid of machines, for over a century, almost two centuries, right?" "So, is this just fear of the unknown?" "Well, Bill, I think we should be worried about artificial intelligence." "I just... [all laughing]" "I just don't think we should be worried about it for the reasons that many are." "Right?" "I don't think it's gonna wake up and kill us." "But it is making decisions that affect our lives, and it's playing with our data, and as you alluded to, it's kind of nudging us this way and that, and I think it's a concern." "Maybe not the concern of the Terminator, but a concern nonetheless." "The Terminator needed time travel." "[Xeni] It did." "Much of the talk of A.I." "is about gnarly tasks that we as humans don't wanna do, or our vulnerable meat sacks get cut up when we try doing it." "Like war." "But, okay" " Okay, I'm a cancer survivor." "And I think about A.I.-- [applause]" "Okay, thank you." "It's big pharma, and science, and my doctors who deserve the applause." "But I imagine a future in which A.I. could read my MRI, my mammogram, my blood test results." "Maybe there's a sensor that's always delivering information to an artificial intelligence" "Including your Social Security number." "Well, you know, if that saves me from dying, I'm all for it." "Yeah." "But there are concerns, right?" "I think there's also unintended consequences, right?" "Oh, no." "No." "Humans?" "Unintended consequences?" "Because imagine, we live in a world in which you are constantly surrounded by something that feels to you like a person." "Where are opportunities for solitude?" "When do we ever feel alone if our car and our phone and our house all feel to us like there's a person there?" "I have a feature on my phone, I can turn it off." "[all laughing]" "But take privacy as a good example, okay?" "We don't necessarily build these things to invade privacy." "But artificial intelligence is increasingly able to derive the intimate from the available." "In other words, things that you just do every day, all of a sudden, we apply machine learning and we can make guesses about who you are and what you do." "Well, you know, already we're seeing that with the so-called Internet of Things, there have been massive Internet outages that are credited to, uh, webcams being taken over by malware." "The security concerns as we develop artificial intelligence" "As artificial intelligence becomes more intelligent," "I imagine there being a whole new field of security that is so much more complex." "Almost like creating watchmen to watch the watchers." "Let me ask you guys-- you all this." "It's organic." "This is to say, in the case of the Internet, it's self-organizing." "We created this thing, but then it's going off and doing stuff on its own." "Is it out of our control?" "Not yet." "Well, how far is it?" "Rao?" "So, I think" " Actually-- So, I also am the president of the Association for Advancement of Artificial Intelligence, which is a research organization." "So, we polled our fellows, the elected fellows of AAAI, as to when do they think we are going to come to a point where there is human level intelligence, because there's all these worries," "and really, it's" " The closest anybody came to is about 50 years, and many people think it's much farther down there." "Well, I wanna bring back to an earlier point that I think is really important, which is the idea that these systems, even if they're not, you know, like people... they do display emergent behavior." "Behavior that the people that designed them wouldn't necessarily have come up-- Have known in advance." "So, you had a segment about Amsterdam." "Another interesting thing that happened in Amsterdam was that a Twitter bot threatened a fashion show." "Tell us all what a Twitter bot is." "Okay, first of all" " So, this is not-- Speaking of not really A.I., right?" "This is a relatively simple script that..." "Script is lines of computer code." "Lines of computer code." "Keyboard thing, hit enter." "Absolutely, exactly." "And what it does is it" "In some cases, randomly, in some cases, more directed, it generates speech on the social media platform Twitter." "And this particular bot had been designed to kind of feel like it was a real person tweeting, and it happened to tweet... a death threat to a fashion show in Amsterdam." "Why did it tweet a death threat?" "Nobody knows." "It was random." "In fact, it was so specific, it said," ""Seriously, I want to kill people." That's what the bot said." "And of course, the police showed up and said... [knocking]" ""Um, you wanna kill people?"" "to the person who had the Twitter bot handle." "And they said, "I had no idea it would do this."" "And the bot was ultimately confiscated." "But the point is, how do we handle situations like that?" "How do we regulate this?" "How do we regulate A.I.?" "First of all, I would like to make this distinction between autonomy, which is, you know, basically able to do things by itself," "versus intelligence." "Self-driving car." "They're completely, you know, orthogonal things." "We already have stupid autonomy all over the place." "The Twitter bot that he's talking about is a stupid autonomy." "It has big consequences, in this case." "It does." "I mean, but my only thing is, coming from A.I.," "I think you can't put the entire world's problems on A.I." "So, in general" "Well, we're gonna put them on you." "[laughing]" "I don't know if-- They should explain that." "I think the problem is that artificial intelligence isn't that far along." "And yet, we feel comfortable deploying it in our most sensitive social institutions." "For example, people use, at least algorithms" "We can quibble about whether it's artificial intelligence." "to figure out whether people who are in jail are gonna be recidivists." "And then therefore, to dictate whether they get out of jail." "Or there was a great study by the folks at ProPublica showing that judges will use... algorithms, in order to figure out what someone's sentence should be based on their risk factors." "And lo and behold, when they scratch beneath the surface, it was racist." "That is to say, it was twice as..." "It was much more likely to recommend, um, that people of color be..." "Well, how did it used to be?" "Are we losing... jobs?" "Are there people who had judgment... that no longer have the chance to deploy that judgment and we're losing...?" "So, first of all, these kinds of decisions, I would think, from a computer science perspective, they're not algorithms." "They're really-- There's a data that's available, and the programs saw some patterns in it." "The question is how you make decisions based on the patterns that you see in the program." "It's still up to the humans." "And, you know, I think" "We're still programming them." "Indeed." "And to some extent, I think, we want to separate it from the" "We should be getting the most diverse data possible." "So, for example, I'm supposed to be kind of in A.I." "Alexa still doesn't understand my accent." "And what exactly is the problem with that?" "It's obviously biased against me, or is it because they just gave you guys to train?" "And it's a question of getting the right diverse data to train our machine learning algorithms." "Let's not throw away the baby with the bathwater." "We always needed... diverse data, that's very important." "So, here's what it is... for me." "Artificial intelligence is a tool." "And we have to all be aware of it and use it carefully and use human judgment to... well, to save the world." "So, thank you all very much." "Thanks to Xeni, Ryan, and Rao." "This has been fantastic." "I really appreciate it." "Thank you, guys." "Thank you." "I'm reaching, I'm reaching, I'm reaching." "Xeni, I'm reaching." "Thank you." "Now, all this talk about artificial intelligence and robots taking over... kind of makes me wanna make a drone of my own." "Stand by." "Oh, good, good, good." "You see, I've been wanting to reprogram my quadcopter." "You know?" "No, I am." "It..." "It relies on, you know, rates of change." "It's not just rate of change of position... or velocity, or acceleration, or jerk." "You gotta get all the way down to snap." "It's..." "It's just calculus." "I mean, what?" "Oh, that's what I'm talking about." "This is how I can get to this thing." "See, the red light, that's very troubling." "Wilhelmina, can you read me the instructions after the object about snap?" "[Wilhelmina] Getting directions to the bar." "[audience laughs]" "No, no." "I wanna get the instructions for the..." "quadricopter" "My mistake." "I will get you directions to the bar." "Uh, the quadricopter, not the bar." "Really?" "Are you sure you don't want to go to the bar and see your friend Keith, and be like, "Oh, hey, Keith." "You are the best friend I've ever had, Keith." "I love you, Keith"?" "Wilhelmina, are you okay?" "Maybe I would like to have a drink with you and Keith sometimes." "A drink?" "Uh..." "You don't have a mouth." "That is racist." "What?" "I just want directions to fix my quadricopter." "You want directions?" "Turn your life around, Bill." "Enough of that." "We've been talking about artificial intelligence all night, and I think that I have something very cool and very special to show you." "I'm here with Saqib Shaikh and Margaret Mitchell from Microsoft, and they have an extraordinary artificial intelligence technology." "Welcome, welcome, welcome." "[audience cheering]" "Now, Saqib, you can't see me." "When did you lose your sight?" "Oh, um, I lost my sight when I was seven." "And that's one of the reasons" "I've been working on this research project which leverages artificial intelligence to empower vision-impaired people like me." "We have an app on a standard phone, and when you take a picture with the app, it describes what it sees." "Oh, come on." "[Margaret] Yeah." "It's doing this using deep learning, which is one of the cool things that's made A.I. really start to be possible, at least opened up the discussion." "Deep learning?" "Deep learning." "Yeah." "And, so, it can recognize things, this app, using the camera on the phone?" "Exactly." "I'm gonna take a picture of something, like you, Bill." "Okay, okay." "I know from experience I gotta get in the light." "Yeah." "So, this phone is taking a picture of me." "And then we're gonna listen?" "Yeah, that's right." "[phone] Camera, in progress." "It's Bill Nye." "Sixty-year-old..." "That's cool." "[audience cheering]" "It recognized me 'cause I'm a public figure." "Here" " Seriously, it can recognize this object in front of you?" "[phone] Camera in progress." "Probably a vase filled with flowers sitting on a counter." "Not bad." "A vase filled with flowers sitting on a counter." "Now... you don't have sight, but you developed this algorithm based on what you need, right?" "That's right." "There's a big team of us, but I'm the tech lead for this." "And I have the insight as to what kind of things will be useful." "And it's still learning." "Like, you can think of it right now, like a little three-year-old who hasn't seen much of the world." "So..." "It still basically only knows what it's been exposed to, and that's, like, some, you know, 100,000 pictures or so." "But, Saqib, what do you want to see?" "I wanna see the world." "[laughing] [cheers and applause]" "So, can I take the phone and try something?" "For sure." "So, I just touch it..." "I just point it at this-- I won't tell you what it is." "And I touch it." "It's processing." "[Margaret] That might not be in its vocabulary." "Probably a blender sitting on a counter." "Yeah, so..." "Right." "But it's a coffeemaker." "But..." "That's really an extraordinary thing." "Yes." "So, it recognizes the shape, it recognizes the reflection, things like that, and it finds the closest thing it can say." "Speaking of shapes, this particular model..." "Right." "Okay." "is near and dear to me, because as a young man... my first real job as an engineer..." "was working on this airplane." "All right." "[Bill] So, we're gonna try it." "Now, let me be clear." "It wasn't a wooden model." "It was, uh, quite a bit bigger than that." "So, you're taking the image on a hyper-reflective surface, which might confuse it." "Probably a stainless steel appliance." "Right, it sees a lot of stainless steel." "Camera in progress." "Probably a plane that is sitting on a table." "Wow." "That's so cool." "Yeah." "That to me is just amazing." "And being able to talk about kind of anomalous things is particular to Microsoft's approach, where we use some shallow learning in the generation, and keep track of the words that we've said and the words we have to say in terms of what we've detected." "So, Saqib, has this improved your life, or do you think it will?" "Yeah, I think it's really going to improve my life in the future." "It really showed me what the possibilities are." "What really excites me, for example, is wearables." "Sort of something that you can have on yourself all the time so you're not holding up a phone." "So, you have those glasses." "Right." "My eye goes right to them." "Right, heh." "These are a pair of glasses made by a company called Pivothead." "They have a little camera in the top of them." "[Bill] I see." "Yeah." "Yeah, exactly." "And this is still away in the future, but it's something that we'd love to be able to see going forward." "You'll probably see more and more of this kind of wearables technology come out." "With the idea being that, like, we can use A.I. in people's everyday lives without kind of disrupting them if we have a form factor that people can just be comfortable with." "So, you are using artificial intelligence to, dare I say it?" "Save the world." "Thank you guys so much." "We can do this." "See?" "Artificial intelligence isn't good or bad." "It's what we do with it." "Thank you so much for coming." "Thank you to Margaret and Saqib." "We can do great things." "I love you guys." "[audience cheering] [theme music playing]" "[narrator] Save the world, Bill." "Nye."