podcasts / The Global Story / episode
The Global Story cover art
podcast
📅 2026-11-17 ⏱ 27 min episode 📝 Transcript · 4,890 words

AI can imitate a human. But can it think like one?

The Global Story · · Transcript & summary by WhipScribe

🔔 Get every new The Global Story transcript by emailOne email per new episode. Unsubscribe in every message.
One email per new episode, unsubscribe in every message. Not shared with the publisher.

Summary, chapters and key takeaways for this episode are on their way; the transcript below is complete.

▸Transcript

Our own transcript of the public episode, with a timestamp about every 90 seconds. Tap a timestamp or the link button on a paragraph to copy a link that opens this page there.

0:00This BBC podcast is supported by ads outside the UK. to score big. From seasonal finds to favorites you'll use all year long. Plus, with Wayfair, enjoy fast and easy shipping with delivery in time for the holidays. Save up to 70% off early Black Friday deals. Shop now at Wayfair.com. Wayfair, every style, every home. Booking a Vrbo vacation rental means you get VrboCare and 24-7 life support, verified reviews from real guests, and top-rated homes with the Love by Guest filter. I just booked my Vrbo because there was a sweet wine fridge. We all have our reasons. If you know, you're Vrbo. Terms apply. From this point forward, all of United States documents, and hopefully the world's, will be changed to use the much more accurate term,

1:02super, as opposed to artificial. So it's super intelligence. That is Donald Trump addressing the United Nations General Assembly back in September when he decided to rebrand artificial intelligence as super intelligence. Trump's been obsessed of late with the concept of super intelligence. It's artificial. This is extreme intelligence. It's superior intelligence. It's supreme, also, intelligence. But the word super is the best word of all, and it's the simplest. In a post on Truth Social, he declared that the White House would consider anyone who used the term artificial intelligence as, quote, the enemy. But what does super intelligence actually mean? And for that matter, what does intelligence mean when we're

1:52referring to inanimate machines? For over 75 years, computer scientists have been trying to figure out whether computers can think. But with the advent of highly sophisticated chatbots like Claude and ChatGPT, might we finally have our answer? From the BBC, I'm Tristan Redman. And I'm Asma Khalid. And today on The Global Story, how smart is AI anyway? And will it ever be smarter than us? This week, all week, we're bringing you a special news strand of our show called The Global Backstory, where we're exploring the history behind the news today. You hear stories about AI all day, every day, but most of them are about where AI is going. But on our show, we're looking at where all of this came from. My name is Thomas Germain. I'm the BBC's technology

2:52columnist and host of the BBC's flagship tech podcast, The Interface. Great. And a fellow American BBC host. Yes, indeed. Yeah. Fantastic to have you on the show. We're taking over. Reverse colonization. Not yet. Not yet. Yeah, exactly. Well, it's great to have you with us, Tom. We're going to be spending a lot of time on the topic of superintelligence, not because Donald Trump is telling us we have to call it that, but because it applies to our conversation. What does it actually mean, though, Tom? Yeah, well, it's hard to say, and not because I'm uninformed on this topic, but because no one really knows. A while ago, we started discussing the idea that someday, maybe soon, perhaps never, we'll build an AI that is better than human beings at all or most tasks, right? And people started referring to that

3:46as artificial general intelligence or AGI. Then the conversation became incredibly fraught. Some people started calling it super intelligence. There's no meaningful, widely accepted definition of what that is. So it just makes a complicated issue even more complicated. Well, all of this relies on some kind of intelligence test. Like, how do you evaluate intelligence? Where does the story of intelligence testing full stop start for you, Thomas? Well, I think the best way to start the conversation is with a French psychologist named Alfred Binet. In around 1905, Binet wanted to figure out a way to measure which French school children needed additional help in the learning process, and he came up with the IQ test, the

4:31intelligence quotient test. Apparently, it was very useful for that particular task. But then it became super popular throughout the 20th century in a lot of ways that I think are problematic. In particular, it became a favorite of the eugenicist movement, like people who believed that certain people and races were inherently more or less intelligent and more or less valuable than others. And there was this test that gave you a number. It sounded scientific, and it was used for all kinds of hateful things. And I think the real problem with it that even Binet acknowledged in the early days is this test is limited. It's not measuring your innate ability. It's measuring your ability to perform on a certain test. And we know that

5:15people who aren't used to taking tests or this kind of test perform worse on the IQ test. Does that make them less intelligent? We also know that it only measures certain kinds of intelligence, right? Certain sorts of logic and spatial reasoning problems. It doesn't get at creativity or most kinds of critical thinking or social skills and things like that. All things that I think we would wrap up into intelligence. Well, you know, all this week we're talking about artificial intelligence. And so I'd really love to understand how we move from the world of intelligence testing on humans to the idea of trying to test the intelligence of machines. Well, this idea of a machine that is intelligent

5:58or conscious maybe goes way back before the invention of computers. But if we fast forward a little bit, I think we can really pin this, or at least the most important next chapter here, in the 1950s with the computer scientist Alan Turing, one of the most famous computer scientists of all time. He was a big part of the British war effort. He did a lot of really important work on code breaking, and he did some of the foundational development on modern computers, both in building the machines, but also developing the theory that would underpin the advancement of computers decades into the future. So not just a foundational, like practical builder of these tools, but also an important philosopher in what

6:42computing would ultimately become. Turing had this idea of an AI or computer, rather, to be able to fool a human being into thinking that they were talking to another human being, right? They would, like, go back and forth in a conversation. If the person was sufficiently fooled, we would say that that computer had achieved intelligence. Did any machines in those early years, as he had this idea, actually come close to matching Turing's definition of intelligence? It wasn't long before we had computer programs that were, you know, questionably passing the test. The most famous examples in the 1960s. Now to Horizon asks, who will be the master and who will be the slaves in a future dominated by super intelligent machines?

7:30A researcher built this computer program called ELISA. ELISA is a computer program that anyone can converse with via the keyboard and it'll reply on the screen. It was a chatbot, sort of like ChatGPT and Cloud, except it was a lot less sophisticated. Basically, the way that it worked is you would say something to it and it would kind of parrot it back in a different way. So you'd say, I'm having a bad day. And it would go, why are you having a bad day? Men are all alike. In what way? They're always bugging us about something or other. Can you think of a specific example? Well, my boyfriend made me come here. Your boyfriend made you come here? He says I'm depressed much of the time. I'm sorry to hear that you're depressed. It's true. I am unhappy.

8:14Do you think coming here will help you not to be unhappy? People were totally fooled by this thing. In fact, some people who were told that it was a computer program didn't believe the researcher. Weisenbaum's secretary fell under the spell of the machine and I asked her to my office and sat her down at the keyboard and of course I looked over her shoulder. After two or three interchanges with the machine, she turned to me and she said, would you mind leaving the room, please? People quickly developed this sort of attachment with this tool. This has come to be known as the Eliza effect, where people develop a sort of psychological closeness with these tools. As technology developed, we kept, you know, setting up a new system where like, well,

9:03surely a computer will never be able to do this. This is the domain of human beings. And I think the most famous example was back in 1996 when IBM built a computer called Deep Blue, and it went up against the chess grandmaster of the world. Some people say the best of all time, Garry Kasparov, in a highly publicized, televised series of tests. Are we missing something on the chessboard now that Kasparov sees? He does not look, he looks disgusted, in fact. He looks just, whoa! Deep Blue Kasparov has resigned. They went head to head. Kasparov won a match, but then ultimately Deep Blue came out ahead. And Kasparov was in such disbelief that he very publicly accused IBM of cheating, that there must have been a human on the other side of Deep Blue that was beating him.

9:57I interviewed some of the researchers once who built Deep Blue, and they were upset about this. They were proud of their work. They took it very personally. More recently, Kasparov, as we're candid, he accepts that no, Deep Blue, and they were upset about this. They were proud of their work. They took it very personally. More recently, Kasparov has recanted. He accepts that, no, Deep Blue did what they said it does. It really was a machine. Yeah, but there was a story about a chatbot that called itself Eugene Guzman that was developed in 2001. Eugene Guzman isn't your average seventh grader. In fact, he doesn't go to school, but he already seems to know everything. That's because Eugene is a computer loaded with artificial intelligence and the wisecracking

10:30awkwardness of a 13-year-old. It was supposed to be a teenage Ukrainian boy, and as he was talking to people, he would even bring up he had a pet guinea pig. His father is a gynecologist, his mother a famous TV personality. There was like a whole backstory. How about an artificial love life? He says he has, quote, a load of female acquaintances. Why was Eugene Guzman a 13-year-old Ukrainian boy? I think it was partially to excuse some of the awkwardness that came with chatbots that were at that level of sophistication in the early 2000s, right? They, like, weren't that good. Are you saying that Ukrainian 13-year-old boys are unsophisticated? I'm saying they're not particularly, I would assume, sophisticated speaking English.

11:14When asked why the AI is modeled after a teenager, creator Vladimir Veselov says, 13 years old is too old to know everything and too young to know nothing. And then in 2014, there was this battery of tests that was set up with a panel of judges. Now, the University of Reading says the computer program is called Eugene Guzman has passed the iconic Turing test. And Guzman, you know, he didn't fool most of the judges, but he did like about a third of them believed they were really talking to a Ukrainian child. Later, people clamored because they just really don't want computers to pass the Turing test. They said, oh, well, there were like the way the test was set up is problematic. Like it doesn't really count.

12:02Another story with IBM, which for a long time was the leader in artificial intelligence. They developed a computer called Watson. And Watson started to look more like the kinds of things I think that we're thinking more of intelligence now. You could ask it complicated questions and it would look them up in this massive database. And they pitted Watson against Jeopardy champion Ken Jennings in 2011. From the T.J. Watson Research Center in Yorktown Heights, New York, this is Jeopardy! The IBM Challenge. And now, here is the host of Jeopardy! Alex Trebek. It's a trivia quiz, essentially. Thank you, Johnny Gilbert. Thank you, ladies and gentlemen. You are about to witness what may prove to be an historic competition.

12:51And you have to answer in the form of a question. The clue will be like... Alternative meaning is 400. Four-letter word for the iron fitting on the hoof of a horse or a card-dealing box in a casino. Watson. What is it? Chew? You are right. You get to pick. Literary character APB for 800. And Watson beat it. Now we come to Watson. We're looking for Bram Stoker. And we find who is Bram Stoker and the wager. Hello, 17,973, 41,413. And often this is seen like you're a genius if you can do well on Jeopardy. So the fact that IBM's Watson was able to do that was a truly impressive feat and a really monumental moment in the history of computing. Well, we've just seen history made here on KXAN this afternoon.

13:43Watson, the supercomputer, wiping the floor with the two greatest human champions ever to play Jeopardy. Watson absolutely wasted his human competitors. And as has happened many times, including more recently with modern AI tools, there was a lot of existential dread in the media that computers were developing and eating into more and more of what we used to think you need for the season ahead. Save on kitchen and dining, bedroom furniture and mattresses, holiday decor, and way more. Now is the time to score big, from seasonal finds to favorites you'll use all year long. Plus, with Wayfair, enjoy fast and easy shipping, with delivery in time for the holidays. Save up to 70% off early Black Friday deals.

14:43Shop now at Wayfair.com. Wayfair, every style, every home. Tires matter. They're the only part of your vehicle that touches the road. Tread confidently with new tires from Tire Rack. Whether you're looking for expert recommendations or know exactly what you want, Tire Rack makes it easy. Fast, free shipping, free road hazard protection, convenient installation options, and the best selection of Redestein tires. Go to TireRack.com to see their Redestein test results, tire ratings, and reviews. And be sure to check out all the special offers. TireRack.com, the way tire buying should be. But here we are in 2026, and you have these large language models like ChatGPT, Claude, that have become ever so pervasive. And I think it's fair to say that in this moment,

15:32it feels like the Turing test has passed. I don't know that we've even had a conversation really about that in public, but it feels like we've all just assumed that we've moved beyond that point. I mean, is it accurate to say that these large language models broke the Turing test? It's out the window, the Turing test in the post-ChatGPT era. In fact, what was really notable to me covering at the time I thought about this is people didn't even really bring it up. Like at that point, computers had become so sophisticated that people weren't even really thinking about the Turing test as something that would be remarkable. One of the interesting things is that just like how people were fooled with ELISA in the 1960s, they didn't have the context,

16:19right? And now I think the more you use modern AI tools, the harder it would be for one of them to pass the Turing test with you, right? There's a particular flavor. There's a style that large language models have. I can even tell like, oh, like that one kind of sounds more like Claude than ChatGPT. Really? Yeah, absolutely. There's like certain turns of phrases or like the sentence construction. They all kind of have their own feeling to them. And I think Turing actually gives us a really interesting framework. He called it the imitation game. And on some level, that's what these computer programs are doing, right? People call large language models spicy autocorrect, right? Which is kind of disparaging.

17:01It predicts what the next word should be for an answer that will statistically, according to all the math that it's done, be a satisfying answer. But the thing is, math is just what it's doing, right? You plug in some words, it turns them into numbers, and then it does these really complicated math problems that are so big, a normal human being couldn't do it in their lifetime probably. But in terms of ability, you could sit down and do all the math that ChatGPT does by hand. And as we talk about intelligence, a lot of critics that I speak to about this say a math equation cannot be intelligent. And ultimately that's what this is, right? And maybe as they get so complicated, they become indistinguishable from intelligence.

17:53But does that mean that they're the same thing? I'm not convinced. Yeah, because an equation is an inanimate object and it cannot be made more animate just by virtue of being complex. That's absolutely right. And if you talk to cognitive scientists and philosophers who study the mind and neurologists, there's one school of thought called computational functionalism, which essentially says that the brain is just a meat computer, right? That it's just ultimately what it's doing is it's processing chemical and electrical signals. And it works the same way as a computer works, that it fundamentally is all math. And therefore, you could, given infinite time and resources, presumably you could build a brain out of silicon. But there's a lot of experts who disagree with

18:41that. And it's worth pausing to remember that we don't know how the mind works on a fundamental level. We don't agree on what consciousness is. We don't agree on what intelligence is. And we have vague ideas of how the brain works, but a lot of it is a total mystery. But I think it's fair to say that there's some daylight between these two systems, the human and the machine. Okay. so how at this point then do we actually judge intelligence? The Turing test is out the window. We heard this clip, Thomas, of Bill Gates talking to the podcaster Ezra Klein, which caught our attention. I was going down to see OpenAI on a regular basis to see the work they were doing. And I challenged them. He asked OpenAI if their machine could get a five

19:30on the advanced placement biology test, which is, it's like a high level, what, like a high school biology test? A high school test, if you pass the AP test, then it counts for college credit. So difficult stuff. Sam, Greg, and Ilya come up and demonstrate to me getting a five on the AP exam, even on questions that I had made up that it couldn't possibly have seen, very complex biology problems. It was nearly perfect. Well, he also said he noticed that Claude seemed to be as good at coding as he, Bill Gates, was. And then late last year, when the Claude coding models have gotten super good, and I can see that they are as good as I am, that's another moment where I go, this is incredible. Which I enjoyed, firstly, because it felt like a bit of a flex that Bill

20:25Gates was saying he was so good at coding that if something was better than him at coding, then it was quite impressive. But also, it was just fascinating that he seemed to be inventing this as a benchmark. Is this the new Turing test where basically very smart tech people set their own personal things that impress them? There are all these new benchmarks that we've set up now that we've set up now that we've moved past the Turing test and we're measuring things on a more practical level. Steve Wozniak, the co-founder of Apple, famously said that he would judge that a machine was intelligent if a robot could walk into a stranger's house and make a coffee in a kitchen that it'd never seen before. There's all kinds of other tests, right? There's one famous one that a guy

21:03named Gary Marcus, who's an AI researcher, came up with where like, is it if you can show an AI a movie or a TV show without subtitles and it understands the context and it can explain it to you, then it's reached some level of intelligence. There's more standardized tests that the AI labs use where they have like a battery of specific challenges and like math and science and certain kinds of reasoning that they use to compare these tools against each other. And it's really important that we're able to judge whether one AI is better than another or determine whether it's a useful tool, if that's the right one that you should use for whatever you're doing. But I also think it's really important

21:44to remember where this can go wrong. And the same way in the first half of the 20th century that the IQ test was perverted by racists to justify their worldview, I think that there are dangers in letting the narrative slip away from us and how we're defining consciousness and intelligence when it comes to these machines that are being built by increasingly powerful, gigantic companies that in some way function more like countries than particular corporations. Thomas, it feels like there are two possible scenarios being outlined by experts right now. The first is one in which you have AI accelerating at an exponential rate. And the second is one in which maybe it plateaus at some point. It has

22:34natural inbuilt limitations of some sort. What are the two arguments there and which one is most likely to happen? Well, I can give you my opinion. These two scenarios, right, where it just continues to accelerate forever and another where it peaks, there are reasons that both sort of make sense. When researchers first built the modern generation of large language models, they did not expect that they would be as capable as they turned out to be. They had abilities that they weren't expecting, you know, supposedly, and it's hard to say whether these stories are apocryphal, but OpenAI apparently hadn't thought to check whether the early models of GPT would be able to speak other languages, right? And then when they asked it, you know, a question in French

23:24or whatever it was, it was able to, even though they hadn't specifically trained it for that, which is mind-blowing in one sense, but also, you know, it's processing all the text that's ever been written and that it would pick up more than you asked it to is perhaps unsurprising. But things have moved at a clip that's so much faster than even experts predicted that it makes sense that maybe it would continue into the future and the roadblocks that we might hit, we could quickly overcome them. But I think there's a lot of reasons to be suspect of that prediction. For one, the fact that it's coming from companies that are trying to impress their investors above all else, like they would like us all to believe that there's no limit to what they'll able to be billed. Like the same way,

24:08there's no limit to how great my journalism will be. It's just going to keep getting better forever. But I think there are hard limits built in. For one, large language models are built on text, right? And there's only so much text, right? They, you know, apparently read most of everything that's ever been written on the internet. And then there was this whole scandal where some of the AI labs were like buying as many rare books as they could find and like shredding off the spine so they could scan the pages because all of the trillions of lines of text on the internet weren't enough to get us to the next level. But there are other limits too. And I mean, think about it like this, right? Like back when I was like 11 or 12,

24:50I got really good at this video game called Tony Hawk's Pro Skater. Oh, I remember that game. It's great. And I was pretty good. And I was like, I gotta get my hands on a skateboard. Like I could tear this thing up. And I like, you know, begged my mom and convinced her to take me to the store. Brought it home, stood on it, immediately fell right on my face. Because I can describe, I can show you what skateboarding looks like. Or to put it another way, I could teach you the laws of physics backwards and forwards. That doesn't mean that you'd be able to walk if you'd never tried it before. Like, can every idea or kind of thinking be put into written language or spoken language? Like, I can describe what it's like to be in love, but if you've never felt it in your

25:34body, do you really know? I could show you all of the poems that have ever been written, theoretically, and then, like, ChatGPT will write you a decent poem, maybe better than I could write. I don't know. It's not my strong suit, but these tools so far, there has been no compelling evidence that they are able to do things that are truly novel. They can combine things that exist in ways that we've never seen before, but could they come up with a new kind of poem? So far, computers can't do that because there's something more than just what you can express or understand from reading. Tom, it sounds like you're saying that there are elements of intelligence that are uniquely human. And we are now at this point where computer programming

26:21is really sophisticated at pattern recognition. Now, there will be people who will say that part of human thought is pattern recognition. You see, a lot of child development is around recognizing patterns and then learning to mimic and repeat those. And I feel like throughout this conversation, it has made me really wonder, like, what is human thought? And the more prevalent we see AI, the more I have, I think, begun to wrestle with the idea of, well, what is making our intelligence really unique? I got a piece of advice. I was talking to a philosopher who studies consciousness, and he gave me a line that I really liked, is that in these conversations, we have to resist the desire for magic.

27:14If your worldview on this is based on the idea that there's just some spark, there's something that I can never define, that I can't explain, that makes us special and different, you're not thinking about this in a scientific way. I think that AI will continue to develop. I think it's going to get way better than it is now. How much better it'll be, I couldn't tell you. But at a certain point, asking whether it's intelligent is perhaps the wrong question. Perhaps instead we should be measuring what it can do and what effect it might have on our society and paying more attention to the consequences than just like the day-to-day abilities of these machines. That was the BBC's technology columnist, Thomas Germain.

28:05He co-hosts the BBC's tech podcast, The Interface. It's really worth checking out. Go and listen to it wherever you get your pods. And tomorrow here on The Global Backstory, have you ever been called a Luddite? Terms often used to describe people who are tech-phobic. But the real Luddites, English textile workers who smashed up machines in the 1800s, they weren't actually afraid of technology. They were afraid of losing their livelihoods. So what can they teach us about the fight over AI and our jobs today? Make sure to tune in. It is an incredibly fascinating episode. And we'd love to hear from you all about this special AI series we have for you all this week. You can send us your thoughts and questions, as always, theglobalstoryatbbc.com.

28:51And a reminder, if you enjoyed this show, please take a second to rate us wherever you're listening. It really helps other people to find us. And now to give credit to the team who made this fantastic episode today. It was produced by Zandra Ellen. It was mixed by Travis Evans. Our digital producer was Sam Gruet. Thank you. California. Vacations are supposed to be the fun part, but stuff can happen. Travel Insured International has travel insurance that can help protect the money you've put into flights and hotels and help when luggage is delayed, lost or damaged, or if you need to see a doctor on your trip. While planning a trip to the airport, check out TravelInsured.com. Don't wait until things go sideways. Protect yourself from travel surprises today with

29:54Travel Insured International. Terms and conditions apply. Shop plans at TravelInsured.com.

▸More episode summaries

The Global Story cover art
The Global StoryDan Carlin on AI: Lessons from the dawn of the nuclear age2026-11-16 · 27 min
Transcribe your links. Paste a YouTube, Google Drive or podcast link and WhipScribe gives you the transcript. Transcribe a link →
New The Global Story transcripts by email
One email per new episode, unsubscribe in every message. Not shared with the publisher.