The Hidden Science of Imagination & What AI Still Doesn't Understand

Humans are remarkably bad at predicting how long something will take. Whether it's a home project, a work assignment, or even a quick trip across town, we almost always underestimate the time involved. Psychologists have studied this tendency for decades—and understanding why it happens can help you make better plans, meet more deadlines, and avoid unnecessary frustration. https://thedecisionlab.com/biases/planning-fallacy We tend to think of imagination as something artists, inventors, and children rely on. But imagination is actually one of the most powerful—and overlooked—abilities we all possess. It shapes the decisions you make, the problems you solve, the goals you pursue, and even the life you create. There is fascinating science behind how imagination works, and once you understand it, you can use it far more effectively. My guest is Dr. Cassandra Vieten, clinical psychologist, Director of Research at the Arthur C. Clarke Center for Human Imagination at the University of California, San Diego, and author of the book Imagine That: Unlocking the Science of Imagination to Transform the Way You Think, Feel, and Live (https://amzn.to/4bqbZl7) Here is the link to discover your imagination profile: https://imagination.ucsd.edu/profile/ Artificial intelligence can write essays, diagnose diseases, generate stunning images, and defeat the world's best chess players. Yet ask it to perform certain tasks that are effortless for a human child, and it can stumble badly. Why? What exactly makes AI so powerful in some areas—and surprisingly weak in others? Beneath the headlines are some fascinating realities about how AI actually works and where it still falls short. Here to explain it is Toby Walsh, one of the world's leading AI researchers. He is Professor of Artificial Intelligence at the University of New South Wales, Chief Scientist at its AI Institute, and author of The Shortest History of AI: The Six Essential Ideas That Animate It (https://amzn.to/4hCjycc) When a highway lane is closed ahead, should you merge immediately or wait until the very end? Most drivers have strong opinions, but traffic engineers have spent years studying what actually keeps traffic moving. The answer is NOT what many people expect—and understanding it may change the way you drive through every construction zone. https://highways.dot.gov/sites/fhwa.dot.gov/files/FHWA-HRT-23-070.pdf= PLEASE SUPPORT OUR SPONSORS AIR DOCTOR: Head to ⁠https://AirDoctorPro.com⁠ and use promo code SYSK to get $250 off select AirDoctor air purifiers, including the 3500, 4000, and 5500 models. Plus, you’ll receive a free 3year warranty, an $84 value, and AirDoctor’s 30-day money back guarantee WAYFAIR: Ready to upgrade your home for way less? Head to ⁠⁠⁠⁠⁠⁠https://Wayfair.com⁠⁠⁠⁠⁠⁠ right now to shop all things home and get your space ready for less. QUINCE: Elevate your everyday. Go to ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://Quince.com/sysk⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ for free shipping on your order and 365-day returns. Now available in Canada, too! SHOPIFY: It's time to turn those "what ifs" into CHA CHING with Shopify Today! Sign up for your $1 per month trail and start selling today at ⁠https://Shopify.com/realm⁠ INDEED: Get a $75 Sponsored Job credit to help get your job the premium status it deserves at ⁠⁠⁠⁠⁠https://Indeed.com/PODCAST⁠⁠⁠⁠⁠ Learn more about your ad choices. Visit megaphone.fm/adchoices

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Mike Carruthers?

  I thought I had a pretty good idea of where my money was going. And then I started using Monarch. And it turns out, there's a big difference between thinking, you know, and actually seeing it all laid out in front of you. I've spent some time with Monarch now. I've connected everything and really tested it out. And I like it a lot. Your accounts, investments, savings goals, spending, they're all there in one place. But what really got my attention is Monarch doesn't just show you what's happened. It helps you spot what's happening right now. They have an AI weekly recap, and it can flag a spending spike or a change in your net worth or upcoming expenses,

• 0:44Mike Carruthers?

  things that are surprisingly easy to miss on your own. And I found that, once you actually see the numbers, you look at your money differently. If you've ever wondered, where exactly is all my money going? I think you'll find Monarch pretty eye-opening. Write your own money story with Monarch. Use code SYSK at monarch.com to get your first year of Monarch Core half off at just $50. That's 50% off your first year at monarch.com with code SYSK.

• 1:23Mike Carruthers?

  Today on Something You Should Know, why human beings are pretty bad at predicting how long something will take. Then the science of imagination that can take your thinking to new heights.

• 1:36Speaker 01

  We kind of have a fuzzy idea of imagination creativity as people who do art and people who make Disneyland and make movies. But really, your imagination is woven through every part of your day and almost every part of your cognition.

• 1:49Mike Carruthers?

  Also, the right way to merge cars on a highway. And one of the world's leading AI researchers explains what many of us don't understand about artificial intelligence.

• 2:01Speaker 02

  The thing I suspect many people misunderstand about artificial intelligence is they forget there's that word "artificial" at the front, that it's actually fundamentally different to human intelligence. I think maybe we should have made it even more explicit and called it "alien intelligence."

• 2:16Mike Carruthers?

  All this today on Something You Should Know.

• 2:21Mike Carruthers?

  There's probably no better endorsement for a company than this. I keep buying their stuff. That's exactly what's happening with me and Quince. It started with their clothes. I've bought sweaters, polos, pants, jeans, shoes, a belt. And Quince has become one of the first places I look when I need something new to wear. And my wife and son buy from Quince too, all the time. What keeps bringing us back is the quality and the price. Quince uses premium materials like 100% Mongolian cashmere, organic cotton, merino wool. Their cashmere sweaters start at just $60. And with colder weather coming, they have down jackets

• 3:06Mike Carruthers?

  and classic wool outerwear as well. Quince prices everything 50% to 80% less than similar brands by working directly with ethical factories and cutting out the middlemen. And it's not just clothes. They have bedding, bath, kitchen essentials, and furniture. We even have Quince pillows. So this isn't one of those endorsements where I try something once because they sent it to me. We shop at Quince, and I absolutely recommend you try it. Find your next fall favorites at Quince. Download the Quince app for app-exclusive offers or go to quince.com. Get free shipping on your order and 365-day returns. Now available in Canada and the UK too. And when Quince asks where you heard about them,

• 3:51Mike Carruthers?

  please let them know it was Something You Should Know. It's the best way to support our show.

• 3:58Mike Carruthers?

  Something You Should Know. Fascinating intel, the world's top experts, and practical advice you can use in your life. Today, Something You Should Know with Mike Carruthers. You know, human beings are pretty lousy at predicting how long almost anything is going to take. And the reason why that is, is what we're going to start with today on this episode of Something You Should Know. Hi, I'm Mike Carruthers. Welcome. So you've probably noticed that a project that you think is going to take an hour often ends up eating half the day. And we all know that every home improvement project seems to drag on for weeks longer than expected. It turns out, this isn't just bad luck.

• 4:44Mike Carruthers?

  It's a well-documented quirk of human psychology called the planning fallacy. Researchers have found that people consistently underestimate how long tasks will take, even tasks that we've done many times before. In planning how long something takes, we imagine that everything is going to run smoothly while overlooking things like interruptions and unexpected problems and little delays that almost always happen. That's why deadlines get missed and "I'll be there in 20 minutes" often turns into 40 minutes. One of the best ways to make a more accurate prediction isn't to think harder about the future. It's to look to the past.

• 5:25Mike Carruthers?

  Ask yourself, how long did a similar task actually take last time? Your own experience is usually a much better predictor than your optimism. And that simple habit can save you a lot of frustration, missed deadlines, and unnecessary stress. And that is something you should know.

• 5:47Mike Carruthers?

  Imagination. Most of us think of it as something children use to make up stories or pretend to be a superhero or create imaginary friends. But what if imagination isn't something you outgrow? What if it's actually one of the most powerful tools your brain has? Every big invention, every solution to a difficult problem, every plan for the future begins with someone imagining something that doesn't yet exist. So what is imagination? Where does it come from? Why are some people naturally more imaginative than others? And can you actually strengthen your imagination to become more creative, make better decisions,

• 6:31Mike Carruthers?

  improve relationships, and even be happier? Here to explain the fascinating science behind imagination is Dr. Cassandra Vieten. She is a clinical psychologist, director of research at the Arthur C. Clarke Center for Human Imagination at the University of California, San Diego. And she's author of the book Imagine That: Unlocking the Science of Imagination to Transform the Way You Think, Feel, and Live. Hi, Cassandra. Welcome to Something You Should Know.

• 7:03Speaker 01

  Great to be here.

• 7:05Mike Carruthers?

  So what is my imagination? When I use my imagination or someone says, "Use your imagination to come up with something." What is that process? How do you describe that?

• 7:17Speaker 01

  So what you're doing when you use your imagination is that you're using sort of a Swiss Army knife or something that can be used to help you with memory, to help you with planning, to help you with how you see your body. And it actually impacts what happens in your body when you use it. You use it for all kinds of things, even as you just go through your everyday life. A lot of what you are perceiving is in fact coming from your imagination. Because imagination, it's a gap filler. It fills in the gaps with the information that you don't have as you're moving through your world. And it just sort of assumes and makes its best guess about what should be there or what could be there.

• 8:01Speaker 01

  But it turns out, when you start to realize that, you can begin to put things there or decide how you want to see the world. And that kind of changes your whole perception of reality.

• 8:11Mike Carruthers?

  When I think of imagination, and I think when most people think of imagination, they think of creativity, of coming up with new things, of new ideas, of... those two words are closely linked, imagination and creativity. And it's something for people who have a lot of time on their hands to do. That's what it is.

• 8:35Speaker 01

  Yeah, I mean, what we say is that imagination is a way of both simulating, reconfiguring, and transcending reality. And what that means is that you can use your imagination to simulate reality. You can say, "Hey, I'm going to go in for a job interview later this morning, and I'm going to try to imagine exactly what it's going to be like so I can rehearse." But you can also say, "I want to reconfigure reality. I want to see the way my bedroom is set up right now. I don't really love it. I don't like the way the sun shines in my eyes when I wake up in the morning, so I'm going to reconfigure it in my mind before I make changes." And the third is to totally transcend reality and say, "I need to think outside the box right now.

• 9:18Speaker 01

  What if a dragon flew overhead right now and saw all of the situations in my life and told me what it thought from its perspective?" Now, that sounds a little bit crazy, right? But what people find is when they do those sort of fantastical imagination exercises, often that dragon will tell them something that they couldn't learn any other way.

• 9:41Mike Carruthers?

  Yeah, well, that brings it home. I understand that, that yes, you can imagine something, how it would be instead of how it is, or how it could be instead of how it is. And that, I guess, that's the imagination at work.

• 9:56Speaker 01

  Right, yeah. I mean, that's the function. You think through evolution. Evolution probably selected the little creatures that preceded us to select for creatures that were pretty good at trying to figure out what might happen, instead of just being stuck in what is happening. So if you're, you know, a little creature that's coming up to a bunch of rocks, and you say, wait a minute, be cautious, because there could be a tiger on the other side of those rocks, you're going to survive more than the creature that just says, oh, that's a bunch of rocks, you know. So the imagination gets a lot of survival advantage for us. But then it goes much further than that. Just imagining what

• 10:40Speaker 01

  could be moves into imagining brand-new solutions to problems. So making tools, or creating agriculture, or in our world, trying to figure out how the heck you're going to deal with a problem that's been vexing you for months or years. It's often because you're stuck in, well, I could either do this, or I could do this, but I can't do this because of this, and I can't do that because of that. You've got to break outside that box and say, what are 50 things that I could do, real or not, feasible or not, practical or not? And just by doing that exercise of writing down 50 things, even if they're kind of wild, like I can go join the circus, that exercise allows your mind to start to consider new possibilities.

• 11:25Speaker 01

  And sometimes there's a nugget of truth in those fantastical solutions.

• 11:31Mike Carruthers?

  So what I, I think, when people think of imagination, they think that it's very kind of touchy-feely, it's wonderful, but it's hard to really grasp. But you talk about imagination science. What is imagination science?

• 11:50Speaker 01

  Well, I'm really happy to be the science director at the Arthur C. Clarke Center for Human Imagination, and also the co-chair of the World Imagination Network, which is bringing together scientists all over the US and the world to talk about the science of imagination. So if you think, okay, every new invention, every medical advancement, every social movement, every major solution has come from someone imagining it first. The invention of the telephone, the invention of the cellphone. People thought about these things long before they happened. They were imaginary when people thought about them. We're facing so many complex problems in our individual lives and in our society, that

• 12:33Speaker 01

  we must study, how do people come to those realizations? How do people have those insights? How do you explore this territory of the mind that's not very familiar to us? Like you said, it's mostly, you know, Disney and creativity and paintings and art, and those things are very important. They really are helpful to people. But imagination is a working tool that, in a way, has become a little bit atrophied, because we haven't focused a lot on the science of it. So I'll give you one scientific finding. Your brain, when you imagine something, differentiates between that imagined thing and a real thing, a lot less than you might think. So if you imagine, right now, something that you are terrified

• 13:18Speaker 01

  of, that you really are scared of, that you, it is not your jam at all, and you really embody yourself into that imagination, you are going to have some of the same biological processes that happen if you were to encounter the real thing. Cortisol being released in your body, the fight-flight system being activated. And if you imagine those negative things every day, you're actually causing chronic stress in the same way that chronic stress can be caused by a real threatening situation. And so the science tells us this, because we can put people in scanners, look at their brain, have them encounter something for real, have them imagine encountering the same thing, and the same areas of the

• 14:02Speaker 01

  brain light up, and the same kinds of chemicals are released in the body.

• 14:06Mike Carruthers?

  So you often hear people say, oh, Joey has such a great imagination. He's so imaginative. Do people's imaginations differ? And is it just a gift? Or is it something you can develop? Or maybe it's both? Or what?

• 14:25Speaker 01

  Well, we call this the neurodiversity of imagination, that everybody comes into this world likely with some genetics and predisposition to certain types of imagining or styles of imagining. And so some people have a very, very vivid and detailed imagination, where they can put themselves into the situation and tell you all about it. Whereas other people's visual imagination is quite blurry or blocky, like, it's not very vivid. Some people don't see anything at all. They call it aphantasia, where someone, you know, you ask them, imagine your house at home, you know, imagine your home. And they'll say, well, I know I have a home, and I know what rooms are in it, but I can't picture it at all. And so there's

• 15:08Speaker 01

  this wide diversity of imagination. But what most people don't know is that there is, there are other forms of imagination other than visual. There's storytelling imagination, where somebody might or might not be having a lot of visuals, but they can tell a story that is very detailed and very powerful. There are people who have very good social, emotional imaginations, where they are very good at empathizing with other people, because they can imagine very clearly what they might be feeling. And in fact, our lab has an imagination profile survey that your listeners can go take. It's at imagination.ucsd.edu/profile.

• 15:51Speaker 01

  And you'll answer a lot of questions about imagination. And what you'll get back is a diagram of your tendencies, your styles, your profile. I hesitate to call them strengths, because I think every profile can be a strength, and we need all kinds of imaginers.

• 16:09Mike Carruthers?

  And we have all kinds of imaginers. But I think, you tell me, but it seems to me that a lot of, like, the things that you said, the telephone and all these things that have been invented, that much of it is accidental, that it's a couple of ideas that collide, and somebody figures it out, more than it's, let's sit down and imagine the telephone.

• 16:33Speaker 01

  Right, right, yeah. And, you know, most of our major imagination eureka moments are those sort of lightning-strike moments. Now, most of the people who had those euricas had been considering and studying and thinking about this for a long time, but not always. Sometimes people just get that lightning strike. But we can't really afford to just wait for those random lightning strikes and hope they happen. So what we know is that you can practice your imagination. You can learn it, just like a language, or just like an instrument, or even building a muscle. You can imagine imagination as a muscle. And if you don't use it, it gets kind of atrophied. And one of the things that we're concerned

• 17:14Speaker 01

  about is with so much digital technology, so many screens, so much receptive information, and now artificial intelligence, imagination is getting sort of flattened. And so I really think that we need to get more imagination training out into early education, later education, and even in our own lives. It really helps to move people into an imaginative space. Let's say you have a one-hour board meeting where it's all business. If you can take the first 10 minutes of that board meeting and do something imaginative with your group, the remainder of the board meeting is going to be much more productive, even in the real sense.

• 17:58Mike Carruthers?

  I want to get a better sense of what you mean by do something imaginative. So we'll dive into that in just a moment. A big thank you to today's sponsor, Vinted. I think most of us have what I call someday clothes. I'll wear that someday. That'll fit better someday. Someday, I'll have a place to wear that. Meanwhile, it's been hanging in the closet for three years. Vinted offers another possibility: sell it. Every day, people can sell clothes they're no longer wearing and make some extra money. And they've made the process pretty painless. Take what you don't wear, list it on Vinted. Listing is quick and simple. And when an item sells,

• 18:42Mike Carruthers?

  Vinted creates the prepaid shipping label for you. And importantly, there are no seller fees on Vinted. So when you sell something, you keep everything you earn. I like this whole idea, because instead of clothes just taking up space, somebody else gets to wear them, and you get some money out of the deal. Your closet looks better, too. I mean, everybody wins, even the closet. So maybe it's time to stop waiting for someday. Download the Vinted app for free and start listing with no seller fees, so you keep everything you earn.

• 19:24Speaker 03

  The action's on the court. And as the official beer sponsor of the NBA, Michelob Ultra is getting fans closer to the game with a chance to win courtside seats, merch, and more. Visit michelobultra.com/courtside to enter. Michelob Ultra Courtside, 26 to 27. No purchase necessary. Open to U.S. residents 21-plus. Begins on October 1st, 2026 and ends on June 30th, 2027. Multiple entry periods. See official rules at michelobultra.com/courtside for free entry, entry deadlines, prizes, and details.

• 19:52Mike Carruthers?

  We're talking about your imagination. And my guest is Dr. Cassandra Veaton. She is Director of Research at the Arthur C. Clarke Center for Human Imagination at UCSD. And she's author of the book Imagine That: Unlocking the Science of Imagination to Transform the Way You Think, Feel, and Live. So Cassandra, you said a few moments ago about doing something imaginative. And I'm not sure I know exactly what that means. So could you describe some of the techniques that work to build your imagination?

• 20:25Speaker 01

  Yeah, so one is that we know from the emerging science that you can use your imagination to influence your future, to become a co-author of your future as opposed to kind of letting it happen to you. So one exercise might be that you imagine your future self, let's say 10 years from now. And this future version of yourself has accomplished or embodied a lot of the qualities that you hope to embody in the future, that you're working on in yourself. That future self finds themselves in a realistic, but aspirational place where you want to go. And sitting down and having a conversation with that future self. You can do it visually. But if you're not a big visual imaginer,

• 21:10Speaker 01

  you can also just do it by verbally talking to this future self or even writing it down. Hey, future self, what did you do to get to where you are? What's the most important thing you did when you were back 10 years ago, when you were where you are now, where I am now? What are some of the practices that you engaged in? What are some of the decisions that you made? What did you focus on? And when people come back and say, oh, I had this conversation with my future self. I have a better sense of direction. And I even have a decision-making tool. Is the decision I'm about to make going to move me closer to that future self or further away from that future self? And so this has been done in jails and prisons

• 21:52Speaker 01

  with people who don't have a lot of hope for the future, taking them to visualize their future self and resulting in not only more hope and optimism, but the people who do it notice their behavior changes because they have that directionality. It's been done with veterans who are just coming out of service. This is a study at Arizona State University, where they actually sit down with the veterans. They ask them to describe their future state. And they pair them with an artist who then sketches out that future state as they're talking and gives them that image as something that they can use as sort of a map of their future.

• 22:32Mike Carruthers?

  Oh, I like that. I like the picture because that's so visual. And then you have this visual reminder that every time you see it, it reinforces the message.

• 22:43Speaker 01

  That's right. That's right. Yeah. And one of my colleagues, Stephen Asma, told me a cool one the other day where he does a workshop where people will sit down and write out a statement about how they want to come into the future. So I will be loving and confident and full of gratitude, let's say. They take that statement and they take only the consonants. And they create what they call a sigil by putting all those consonants together into like a little symbol. And then they put that symbol up by their bedside table, you know? And when they wake up, they see it and they know that that symbol means I am working towards that mission statement that I wrote. So it's just a different way of,

• 23:27Speaker 01

  you know, we all know about vision boards and affirmations. And I think those are great. But, you know, taking a different approach and saying, I really want to reinforce this in my mind, something that I've imagined, I want to reinforce it.

• 23:40Mike Carruthers?

  See, I would think that you would have said to build your imagination, do imaginative things, you know, paint or write music or, you know, play an instrument or something that, in a way that nobody else does. It's something more doing than discussing.

• 24:01Speaker 01

  Well, here's what's interesting is that imagination actually differs from creativity in a couple of important ways. So imagination, you can think of as the whole field of possibilities. It can be real and practical possibilities. And it can be unreal and impractical possibilities. But it's the entire field of possibility. And you can imagine that you're standing by a well and you drop a bucket down into the well. And when you pull it up, up comes something that's like, wow, this is a new idea. And for some reason, it's important to what I'm dealing with right now. Let me see if I can use it and create something out of it. Then you move into creativity.

• 24:45Speaker 01

  And creativity actually imposes some constraints. The medium imposes a constraint. So if you decide, I'm going to paint this, then you start to paint it. I'm going to write a novel about it. I'm going to write a short story. I'm going to write a poem. I'm going to create an email. I'm going to create a website. I'm going to make my new business card. That's creativity is when you bring it into this world of form. So what I'm encouraging is that people spend more time in that earlier phase of imagination before rushing into creativity. Because if they can spend more time there, they're going to find more possibilities. And if you get used to that territory and you learn the rules,

• 25:28Speaker 01

  you learn to speak the language of imagination, then you start to bring out these truly new ideas, generative imagination, which is different than that simulation imagination.

• 25:40Mike Carruthers?

  The way you look at imagination as someone who works in that world is very different, I think, than the way most people view imagination.

• 25:51Speaker 01

  Yeah, well, Mike, how in your, let's say, your K through 12 education or your K through 16 education, did you ever have a class on how to access your imagination, navigate it, visualize, ideate, sense, transform, apply? We never got trained in any of this stuff. And so we kind of have a fuzzy idea of imagination creativity as people who do art and people who make Disneyland and make movies. But really, your imagination is woven through every part of your day and almost every part of your cognition. So learning more about it and finding out, this is a strong influence in my life. And if I become aware of it

• 26:32Speaker 01

  and I learn how to notice it, recognize it, and apply it, then I'm going to have kind of a superpower.

• 26:40Mike Carruthers?

  Well, you've certainly changed my image of imagination. I mean, my sense of imagination is that it's something for artists and musicians and creative people who use their imagination to create. But when you talk about the science of imagination, boy, it really, really gets interesting. I've been talking to Cassandra Veaton. She's a clinical psychologist and Director of Research at the Arthur C. Clarke Center for Human Imagination at the University of California, San Diego. And her book is Imagine That: Unlocking the Science of Imagination to Transform the Way You Think, Feel, and Live. And there's a link to that book in the show notes. Cassandra, this was great. You really explained it well.

• 27:25Mike Carruthers?

  I really enjoyed the conversation.

• 27:27Speaker 01

  Oh, you're so kind, Mike. That means a lot coming from you. And you're a great interviewer. I love your podcast.

• 27:36Speaker 03

  The action's on the court. And as the official beer sponsor of the NBA, Michelob Ultra is getting fans closer to the game with a chance to win courtside seats, merch, and more. Visit michelobultra.com/courtside to enter. Michelob Ultra Courtside, 26 to 27. No purchase necessary. Open to U.S. residents 21-plus. Begins on October 1st, 2026 and ends on June 30th, 2027. Multiple entry periods. See official rules at michelobultra.com/courtside for free entry, entry deadlines, prizes, and details.

• 28:05Speaker 04

  Pink Pantheress here. Coach and Spotify are teaming up to bring music and style together. And they want you to be part of it. Grab your friends and head to the Tabby Tour, hitting cities around the country with live performances from special artists, exclusive experiences, and personalized products. And discover custom playlists on Spotify, booked just for you based on who you are and how you listen. Live your story with Coach. Learn more at coach.com.

• 28:35Mike Carruthers?

  Artificial intelligence. It can beat grandmasters at chess, write complicated computer code. It can pass difficult exams and hold surprisingly good conversations. Yet AI also makes ridiculous mistakes. Confidently invents facts and sometimes fails at problems a young child could easily solve. So what's really going on? Why is AI brilliant one moment and baffling the next? The answers reveal a lot about how this technology actually works and where it's headed from here. Joining me to explain is Toby Walsh. He is one of the world's leading AI researchers.

• 29:18Mike Carruthers?

  He's professor of artificial intelligence at the University of New South Wales, and he's chief scientist at its AI institute. He's author of four previous books on AI, and his newest book is The Shortest History of AI: The Six Essential Ideas That Animated. Hi Toby, welcome to Something You Should Know.

• 29:41Speaker 02

  A pleasure to join you today.

• 29:43Mike Carruthers?

  So what is it that you think people may not understand about AI? Because as a consumer who uses it, I mean, it seems pretty straightforward. You ask it a question, and depending on how you ask the question, you'll get an answer, and that's kind of the beginning and end of it.

• 30:01Speaker 02

  The thing I suspect many people misunderstand about artificial intelligence is they forget there's that word "artificial" at the front, that it's that is actually fundamentally different to human intelligence. And it's, I mean, it's a natural human conceit that the only intelligence that we've experienced before was our own. We had a very visceral experience of that. And so it's natural to project the sorts of, our sort of intelligence onto the machines, not realizing that it's quite an artificial, quite a different type of intelligence. Indeed, I think maybe we should have made it even more explicit and called it "alien intelligence", just to make, just to make very crystal clear to people that intelligence does come in different forms.

• 30:43Speaker 02

  And the intelligence that we build in machines, I mean, although it comes across when we're talking to ChatGPT or one of these chatbots, it's maybe somewhat similar. It does have some somewhat different characteristics. I mean, to blow our own trumpet, artificial intelligence is much less efficient than human intelligence. You can show a young child a teacup and say, "That's a teacup", and the young child will be able to recognize teacups from then on. Whereas to get a computer to understand what a teacup is, you have to show them thousands of teacups before it gets the idea of the concept of a teacup. So we're still, I'm still really impressed by human intelligence and the efficiency of human intelligence. We don't need vast data centers to run human intelligence.

• 31:28Speaker 02

  Human intelligence runs on just 20 watts of power. That's a dim light bulb.

• 31:34Mike Carruthers?

  Right, but the fact is that I think what I find helpful about artificial intelligence is ChatGPT, one of the other ones, has access to a lot of information that would take me a long time to go find myself. They can just pull it up, use it, translate it, and boom, there's my answer.

• 31:55Speaker 02

  Yeah, well, when I said it was different, that's one way it's different, is that it has got, it's read more books than a human could read in a lifetime of books. And so some of how it's impressing us is just because it's so well-read, and it's better read than any human could ever be. So in a number of areas, AI has got better than humans. Not because it's in some sense more intelligent, it's just because it's better read. It's got access to more information than any human could possibly have.

• 32:24Mike Carruthers?

  Well, the one thing that has become really clear to me is it's all in how you ask the question and what information you provide when you ask the question will determine in large part the quality of your answer.

• 32:42Speaker 02

  Indeed, and that's an example of how limited the AI intelligence that we have today is because you and I, it doesn't matter really how we ask the question, as long as I get the meaning, I'll be able to answer the question. But with AI, you're right, it's very critical that you ask the question in the right way, that you provide the right information for it to work. And human intelligence is much more robust. You can ask questions in many different ways and still be able to get a decent answer. So it's a good example, actually, of how the tools still are at a very early age. I'm sure they'll get much better. I'm sure in time, we'll be able to talk to them in any way that we choose

• 33:26Speaker 02

  and not have to worry that we've written the prompt in the right way to get a good answer.

• 33:32Mike Carruthers?

  But the thing I've noticed is that the more I use it, the more it does seem to know me, that it remembers what I've asked it before and the conversations I've had before. And it must use that, and that enters in to what I'm asking it now to determine the answer.

• 33:52Speaker 02

  Oh, yeah, certainly. The history of your conversation is being remembered and a summary. Not all of it is being used to help it write more appropriate responses in the future. So your listeners should be very mindful of that. People are sharing very intimate details with the chatbots, and they are recording and memorizing that. That makes it good for having a conversation. But equally, you know, they are starting to use that information for other purposes, sell your stuff and things like that. So people should be mindful of not oversharing with the chatbots.

• 34:29Mike Carruthers?

  Well, wait a second. I mean, in fact, I think I asked ChatGPT at one point. I said, "How reliably can I assume that what I talk to you about is confidential?" And it said, "Pretty reliable." It's not, it's, yeah, it made it sound as if this will go no further.

• 34:53Speaker 02

  Well, you don't know what deals that OpenAI, the company behind ChatGPT, is going to sign in the future. They have said that they're going to start selling stuff through their services, start using advertising, so they're certainly going to use that information. And we're sharing a lot of personal information. 60% of the prompts, the queries that people type into ChatGPT are not professional. They're not about work. They're about personal matters, about our relationships, about our friends, about things in our personal lives.

• 35:28Mike Carruthers?

  Well, something I've witnessed and noticed about AI is no matter what you ask it, you will get an answer. I've never once had it come back and say, "I don't know, I have no idea." So even if you ask a very personal question, you'll get an answer. But how would it know that if it doesn't know you personally?

• 35:53Speaker 02

  Yeah, that's one of the things that's annoying about ChatGPT, that underneath it does have probabilities. It does, there are certain times where something's a coincidence, it doesn't know the answer. But it's been programmed in a way to always confidently answer the question and never to say, "You know what, Toby, I don't really know the answer to this question. You better go and look this up yourself." Or, you know, "Opinion varies, experts disagree as to the correct answer to your question here." It will always, and this is one of the problems, I think, with it at the moment, is it always confidently says an answer, and it's easy to be taken in.

• 36:33Mike Carruthers?

  Well, I've heard that there are things you can say, like key words you can give it, that will make it more truthful and less, because it always tells me how wonderful I am and how smart my question was. But that you can turn that off somehow, and I've never done that because I like to hear how wonderful I am. So why would I turn that off? But is that true? If you type in a certain word, it'll stop telling you how great you are and tell you the truth?

• 37:05Speaker 02

  Yes, you could do that. In fact, I did that. You can change what's called the system prompt. That's the prompt that's added to everything you ever ask at the beginning. And you can say, "Just be much more critical, push back when you think I'm wrong. Don't always compliment me." And so I did that for a couple of weeks. And after, I must admit, after a while, I got really annoyed. But it kept on pushing back, and sometimes it said, you know, "Read the room. Sometimes I'm right, and I don't always have to disagree." So again, I think that's a sign that the tools are rather zero and one, black and white. They don't have that subtlety yet of understanding, you know, with humans, that, you know, you can disagree some of the time. But if you disagree all the time, if you push back all the time, people are going to get really annoyed.

• 37:50Speaker 02

  So eventually I had to turn that off because it was just as annoying as the way it works today, where it's a bit sycophantic, it's a bit, it is a bit complimentary. It doesn't push back perhaps enough as it should. Equally, when it pushes back all the time, it gets a bit annoying as well.

• 38:07Mike Carruthers?

  Do you know how many people use chatbots or ChatGPT in particular? Or any of the others? How many people are using it?

• 38:16Speaker 02

  I mean, that's the remarkable thing about ChatGPT is how many people are using it. It's been used by over, on a weekly basis, by over a billion people. That's one in eight of the world typing their deepest, darkest thoughts into ChatGPT at the moment.

• 38:35Mike Carruthers?

  One in eight, what was that number again?

• 38:38Speaker 02

  One in eight, a billion, it's a billion weekly users.

• 38:42Mike Carruthers?

  But they're a billion different weekly users or a billion queries?

• 38:47Speaker 02

  A billion different users. A lot of people, and of course, it's available not just through OpenAI, through the ChatGPT interface, it's available in Bing Search, it's available in Snapchat. There's an AI there called MyAI. See, you've got your own AI now. So yes, it's available through a number of other partnerships as well. But its reach is quite remarkable because, you know, less than half of the world, I think, is on the internet. So it's, in terms of the internet users, it's a real fraction of us have got to use it. And that's, I think that's one of the challenges. We've never had technologies that we've so easily, so quickly, so cheaply put in the hands of so many people.

• 39:27Mike Carruthers?

  One of the things that we all heard from the very beginning about AI is that ChatGPT-type AI is that what it's doing is predicting the next word. And it sounds very odd. I don't know what that means, that it's predicting the next word, because... It seems so complete and it seems so thoughtful. It seems like it's doing much more than just predicting the next word.

• 39:59Speaker 02

  I mean, I think the easiest way to understand it is to look back. When you first started using a mobile phone, when you were typing a text message, an SMS into the mobile phone, it predicts how to finish the word. And it's based upon a dictionary of words and the frequency with which the words get used. And if you type APP, well, then it knows probably, you know, the best way to finish APP is apple. So it finishes the word apple for you. Well, what ChatGPT does is that on steroids. It's not using a dictionary. It's using the internet or about half the internet. And because it's trained on so much more data,

• 40:44Speaker 02

  it predicts not just the next word, but the whole sentence, maybe even the whole paragraph, even the next page. And if you're given enough data, that seems to be enough to be able to give quite meaningful answers. But equally, if we reflect on how we speak, do you know how, anything more than about the next word that you're going to say? Have you thought that much further ahead?

• 41:08Mike Carruthers?

  Sometimes maybe, but you know, I don't know. I don't know. But you would think that it, the way it comes out, that it's kind of been all thought through before the first word comes out. But, and maybe it has, but it just seems very counterintuitive that what it's really doing is predicting the next word as opposed to conceptually coming up with an answer.

• 41:35Speaker 02

  Well, in fact, that's how the recent advances, the recent progress has happened with these agentic AI systems, which are built on top of language models like ChatGPT, is that they go through a number of phases. First of all, they plan how they're going to answer the question. And then they take the steps and the plan and they break those steps down into smaller steps. And eventually then they answer each of the individual steps. So there is a process. And if you're actually asked to see the thinking in one of the chatbots today, well, you can see it says, you know, I'm, you know, searching the web, assembling an answer.

• 42:16Speaker 02

  It says out loud the steps it's going through. So there's a much more planning process now. And that's how it's got to be more capable and answer more complex queries.

• 42:30Mike Carruthers?

  One of the things that, I don't know if this applies to other people, but one of the things that ChatGPT gets wrong for me is I'll ask it for sources. Like, can you give me a link to a place that I can find this out? And it'll give me a link. And I, but I will say in my prompt, give me working links that I could go to. And very often it gives me broken links. And I write back and I say, these links don't work. And they go, oh, good catch. Yeah. Well, but I asked you not to give me bad links and it didn't. But if it's a bad link, how does it even know it's there?

• 43:11Speaker 02

  Well, in some cases, it's literally just making it up. I mean, that's again, along with always confidently answering your questions, it's been trained to always, always try and be helpful. And unfortunately, that includes making some of the answers up. I mean, of course, you want it to make things up. You want it to be able to, you know, write a funny poem or write a witty speech for a wedding. So you do want it to say things that have never been said. But unfortunately, it's hard to tell the AI where you want it to be able to make things up and where making things up, like giving you proper links, is going to be actually not the answer.

• 43:53Mike Carruthers?

  Where do you think this is going? Do you have any sense of that? I mean, is it going to get, it seems like it's getting incrementally better, but it's not, we're not going to wake up one morning and everything's going to be different.

• 44:06Speaker 02

  Well, I mean, the ambition of the field is to get what's called artificial general intelligence, an intelligence that matches human intelligence in all capabilities. And then, and then why would we stop there? Well, maybe we'll get to what's called artificial superintelligence, intelligence that's significantly better than human intelligence in all aspects. I mean, that's, that's the ambition of the field. That's the ambition of the companies like OpenAI, to build intelligence, genius level intelligence that we have on machines. And I'm not sure it's going to be as easy as some of the predictions, or at least the stock market is thinking it's going to be that it will only take a year or two.

• 44:49Speaker 02

  I think it might take a decade or two. But I doubt many of my colleagues would say it's going to take a century or two. It's not that sort of scale of a problem left, I suspect. So that means that in our lifetimes, well, certainly in the lifetimes, certainly in the lifetimes of our children, we're going to have intelligence that's at least matches, probably, most likely exceeds human intelligence. And that, I think, is going to be pretty transformational because the world that we live in is the world that our intelligence has made. The, you know, we've doubled life expectancy because of all the great things that medicine and technology have brought to our lives. But what will that greater intelligence bring to our lives?

• 45:34Mike Carruthers?

  Is it fairly well agreed upon when artificial intelligence first began?

• 45:41Speaker 02

  Funny story. It started on June the 18th, 1956. So it just celebrated its 70th birthday. June the 18th, 1956 was the first day of the first meeting for artificial intelligence. And the words artificial intelligence were invented to describe the topic of that meeting. They'd never been used before. Although we say it's 70 years old, of course, you can trace the genesis, the idea that we might build machines that were intelligent back thousands of years to, you know, to the ancient Greeks, to ancient Chinese civilizations, to ancient Indian culture. There's lots of, lots of ideas that go back thousands of years.

• 46:25Speaker 02

  But it wasn't until the 1950s that we had the computer, the computer became commonplace, that we could sort of think about turning those dreams into reality. And that's when the field started and people got together and said, well, how far can we take this? How intelligent could we make these machines? Could we, could they match human intelligence? And that, that was the first day of the first meeting.

• 46:46Mike Carruthers?

  When you look at the timeline, the history timeline of artificial intelligence, do you include things like the calculator? Isn't that artificial intelligence?

• 47:00Speaker 02

  Well, yeah. I mean, people will argue that that's doing something that's entirely mechanical. But, but the people who invented calculators were trying to mechanize mathematical thought. It turns out you can do that quite easily. You can do it actually mechanically. You can, you can build a, you know, IBM started out building mechanical calculating machines before they moved into electronic calculating machines. So that's, that's quite an easy task, it turns out to be. And now we're getting computers to do mathematics, to do international mathematic Olympiads, to do, prove theorems that mathematicians have never been able to prove themselves. So we've gone from just calculation to doing things that are quite creative.

• 47:45Mike Carruthers?

  But is calculation a building block of artificial intelligence? Or calculation is just a mechanical calculation that's off on its own?

• 47:55Speaker 02

  Well, if you look under the hood, what the, what the computer is doing, it is in its CPU and its GPU, it is doing just simple, simple calculations. It's just to get mathematical. It's just linear algebra underneath. So it is all reduced at the end of the day to adding numbers together, subtracting numbers together, comparing numbers. That's all it's doing.

• 48:18Mike Carruthers?

  Well, it's hard not to be fascinated and mesmerized by how it does what it does. It's fun to watch, that's for sure. And I appreciate your insight into it.

• 48:29Speaker 02

  A pleasure. I hope your listeners realize that it's not maybe as magical as it maybe it seems. Because, you know, when a computer speaks back to you, it does seem like magic. And to realize actually there's some quite simple ideas underneath. And there's also some quite significant missing pieces that just because it can confidently answer that question doesn't mean it's necessarily as intelligent as us yet.

• 48:56Mike Carruthers?

  I've been speaking with Toby Walsh. He's one of the world's leading AI researchers. And his latest book is called The Shortest History of AI, The Six Essential Ideas That Animate It. There's a link to that book in the show notes. Thanks, Toby.

• 49:11Speaker 02

  It's been a great pleasure talking to you today.

• 49:17Mike Carruthers?

  If you're driving your car near and through a construction zone and you see a lane ending up ahead, your instinct is probably to merge over as soon as possible. And if someone stays in the closing lane until the very last second, you probably think they're being rude. But in heavy stop and go traffic, traffic engineers say the opposite is often true. Using both lanes all the way to the merge point and then taking turns, called a zipper merge, keeps more pavement in use, which shortens traffic backups and can reduce congestion dramatically. The problem is that most drivers don't know when it's appropriate. In light traffic, merging early is still the safest approach

• 50:02Mike Carruthers?

  because everyone is moving at highway speed. But once traffic slows to a crawl, filling both lanes and then altering one car at a time is actually the most efficient. So the next time you see someone staying in the closing lane until the end, they may not be trying to cut in line. They may be following the strategy that traffic experts have been recommending for years. It only works, though, if everyone understands the rule. And that is something you should know. You can help us by talking up this podcast, telling your friends about it, asking them to listen, telling them how much you enjoy it. It really helps. So share this podcast, use the share function on the player you're using and, and help us grow our audience, please.

• 50:47Mike Carruthers?

  I'm Mike Carruthers. Thanks for listening today to Something You Should Know.

• 50:53Speaker 04

  Pink Pantheress here. Coach and Spotify are teaming up to bring music and style together, and they want you to be part of it. Grab your friends and head to the Tabi Tour, hitting cities around the country with live performances from special artists, exclusive experiences and personalized products. And discover custom playlists on Spotify, booked just for you based on who you are and how you listen. Live your story with Coach. Learn more at coach.com.

• 51:21Speaker 05

  This episode is brought to you by State Farm. Football has a play for just about everything. You can go deep, keep it on the ground, or draw up something nobody sees coming. And whether you want to run a flea flicker or the tush push, you need a plan to fit your unique situation. That's why State Farm takes the time to get to know you so they can help personalize coverage for your life and budget. Now that's a solid game plan. Like a good neighbor, State Farm is there.