Ideas - We believe in artificial intelligence the same way we believe in ghosts

Episode Date: March 12, 2025

Hidden in the 1950 academic paper that launched the famous 'Turing Test' of machine intelligence, is a strange mystery. Cryptographer Alan Turing argued that humans might always be able to outsmart machines, because we have supernatural powers like ESP, telepathy, and telekinesis. His belief in the paranormal is just one part of the spooky side of artificial intelligence. Like hauntings or seances, AI is an exercise in self-deception; we imagine intelligence from computation and data, just like we imagine ghosts from strange lights and bumps in the night.

What does a mummified Egyptian child, the Parthenon marbles of Greece and an Irish giant all have in common? They are all stuff the British stole. Maybe. Join me, Mark Fennell, as I travel around the globe uncovering the shocking stories of how some, let's call them ill-gotten, artifacts made it to faraway institutions. Spoiler, it was probably the British. Don't miss a brand new season of Stuff the British Style. Watch it free on CBC Gem. This is a CBC Podcast. Let's talk about artificial intelligence here for a moment.

How can you tell what's real, what's fake? Experts say that line is being blurred every single day online and in our social media feeds. Welcome to Ideas. I'm Nala Ayed. Hurricanes, global conflicts, and the election all raising the stakes on misinformation as AI-generated content continues to flood the internet. The days of marveling at AI's ability to write a poem, code a web page, or hold a conversation

are over. A group of leading tech companies, including Meta, Google, and TikTok, committed to limiting misleading AI content on their platforms. The most impressive thing about artificial intelligence might be its ability to deceive. And in the past couple of months, these AI generated videos have multiplied, causing billions of dollars in fraud. But deception has always been part of AI,

all the way back to the very beginning of machine intelligence. Turing believed that it was completely legitimate for the computer to lie and cheat, because human beings lie and cheat. Deception was at the core of the imitation game, an early test of AI. But the creator of that test, Alan Turing, wasn't just interested in technical deception. He was fascinated by the possibility of paranormal deception.

Did Turing believe in ESP? I think he was certainly open to the idea that it might exist. And it's intriguing to me that now it's considered a sort of a thing that only cranks are interested in. It's like the flat earth society or something. In this episode, Jennifer Jill Fellows, philosophy instructor at Douglas College in New Westminster, BC, looks at the spooky supernatural side of AI. All of these different technologies

became ways of potentially capturing the likeness of the dead, the voice of the dead, the image of the dead, and access them through these channels, that maybe there's a way of maintaining a relationship with the dead. And how modern technological deception has roots in the mysticism and spiritualism of the past.

So believers and spiritualists would join seances to try to communicate with the other world. And this interpretation will be crucial to shape the kind of interaction we are having with the machine. Here's Jennifer Jill Fellows with Paranormal Programming. When I was young, I and my friends got deeply interested in seances. We even staged a few in my childhood home. One night, we were calling forth the spirit of William Shakespeare. We were theatre nerds. Just as we asked for a sign that Shakespeare was with us, we heard an enormous rattling thump right against the wall behind our summoning circle. We screamed. One of my friends called out desperately for us not to

break the circle and released the ghost of Shakespeare into my childhood home. But it was too late. We were all running scared. The thing is though, that old house was always thumping and creaking. Right behind the wall was my mom's old china cabinet and whenever the temperature would change or whenever the house would settle the cabinet and all the other equally old furniture would creak, thump and rattle around. It happened all the time, a meaningless background noise to my childhood life. But when you're in the mood for a seance, when the atmosphere is just right and the candles are flickering and you want to believe more than ever, those innocuous thumps and bumps become deeply, terrifyingly meaningful. We were deceiving ourselves, imagining intention and intelligence where, most likely, none existed.

Deception has for a long time been a core part of thinking about machine intelligence. And it is central to a well-known test for machine intelligence, the Turing test. The ideal arrangement is to have a teleprinter communicating between two rooms. Seventy-five years ago, in 1950, British cryptographer, philosopher, and computer scientist, Alan Turing, asked a question.

I propose to consider the question, can machines think? In a paper called Computing Machinery and Intelligence, Turing devised a test to answer that question. Today, it's most often called the Turing test. He called it the imitation game. The Turing test works sort of like a text conversation on the internet. If you're typing away and you really can't tell whether you're communicating with a human being or an AI powered chatbot, well that AI just

passed the Turing test. Here is an example that Turing gives of what the test with an AI chatbot in some far distant future might be like. Please write me a sonnet on the subject of the fourth bridge. Count me out on this one. I could never write poetry. Add 34,957 to 70,764. The answer is 105,621.

Do you play chess? Yes. Clearly, we have already surpassed Turing's far distant future. Today's large language model generative AIs like ChatGBT will happily write you a poem about a bridge. The fourth bridge stretches wide and grand, a steel embrace across the land, its pillars firm by sea and sand. Or discuss how much it loves chess in great detail.

Yes, I play chess. I enjoy it a lot. It's a great way to challenge the mind and think strategically. The test was a critical moment in thinking about AI. Turing's innovative idea was that when it comes to intelligence, it doesn't matter what's going on in the machine's mind, or if it even has a mind at all. What matters is behavior. If its behavior successfully deceives us into thinking that it is intelligent, then it is intelligent.

But buried in that logical and analytical academic paper is a mystery. Telepathy, clairvoyance, precognition, and psychokinesis. It is very difficult to rearrange one's ideas so as to fit these new facts in. Once one has accepted them, it does not seem a very big step to believe in ghosts and bogies. Ghosts, bogies, telepathy, and ESP, part of Turing's paper we could call Argument Nine. And I thought, maybe I'll write about Argument Nine. And it opened up all sorts of possibilities that had never previously occurred to me. David Levitt is a biographer of Alan Turing and a distinguished professor in the English department at the University of Florida.

Well, first of all, I think it's worth noting that that Turing's method in the paper, which is one that he returned to time and time again, the way he presented his argument was by anticipating objections and answering them. In his paper, like any good scientist, Turing listed a bunch of possible objections to the validity of his test. Basically dispense with them or respond to them before other people make them. Things like, because machines can't fall in love or be creative or ever truly surprise us they can't be intelligent. And they're all really interesting and his responses to them are very interesting and

then right in the middle of them there's this one that has left everybody scratching their heads. Out of all of these objections, there is only one that Turing takes seriously. The possibility that a human could cheat on the test because they have paranormal abilities. I assume the reader is familiar with the idea of extrasensory perception. These disturbing phenomena seem to deny all our usual scientific ideas. How we should like to discredit them? Unfortunately, the statistical evidence, at least for telepathy, is overwhelming.

No citation. He was imagining various scenarios in which one of the participants in the Turing test could sort of spoil it or corrupt it through the use of ESP, which would be undetectable. Suppose the digital computer contains a random number generator. Then it will be natural to use this to decide what answer to give. But then, the random number generator will be subject to the psychokinetic powers of the person asking the questions. What sets humanity apart from machines, according to Turing, is not our free will, creativity, or emotions. It's the paranormal.

With ESP, anything may happen. So I became curious and the reason that I became curious was because I coincidentally was doing some research into the history of the paranormal. What I discovered was that at exactly the time that Turing was working on this essay, there was a surge in public interest in studies of extra sensory perception

or scientific attempts to do rigorously and scientifically investigate whether people actually have the ability to perceive extrasens. Actually, it was on the air. An eight-episode series on BBC's Light programme aired in 1949. It featured Sydney and Leslie Piddington, a husband and wife mentalist team. Off on a mystery trip from Filton Airfield, Bristol, went a strato cruiser. mentalist team. The object to transmit a train of thoughts to Mrs. Piddington. The Piddingtons were this sort of amazingly strange couple from Australia who were a kind of vaudeville act. Oh, it's two lines. Bird, spill, hello.

I've got it, I can guess it. They would be in two separate rooms and one would name, say, a film star outside the hearing of the other and the other would get it. And probably they were mostly cheating or they had a system. Anything to say? Well, only thanks very much everyone and you're the judge. Well, I think maybe, well, all right. I'm baffled. Now back to Sidney Piddington and Tricadilly.

This BBC broadcast created a media sensation. The tabloids were talking about it, and it inspired cartoons and jokes. Publications like The New Statesman ran articles about the couple. And it wasn't just the Piddingtons. The word telepathy itself appeared in newspapers and magazines like never before. The paranormal was everywhere. It wasn't just treated credulously by media. It had a home in academia as well, including one of Turing's own teachers. So all this was was in the news. Turing was clearly aware of it. One of his,

C. D. Broad, one of his mentors from Cambridge, was extremely interested in this and was a member of the Society for Cyclical Research, which was a very fascinating and important group, many of the members of which were extremely prominent intellectuals, physicists, mathematicians. I think one of the great sort of unsung figures was Nora Sidgwick, who was a mathematician and went on to become the head of Gertin College, one of the two women's colleges at Cambridge.

And they decided that they were going to take a very rigorous and highly scientific approach toward areas of thought that seemed to be anti-scientific. You know, was it possible to communicate with the dead? Were there mediums who are not frauds? It's a difficult question to answer. The question of whether there were mediums who were not frauds.

Of course, barring the existence of the paranormal, mediums were deceiving all the time. But even if a medium isn't intentionally deceiving the sitters, self-deception is still possible. People seek out mediums for all sorts of reasons, but a chief one is to allow them to communicate just once more with a deceased loved one. People torn by grief and loss at a séance table may be all too willing to believe that the spirits sit among them

and may deceive themselves into seeing evidence for that belief. I started to work on AI several years ago. I came to AI in a kind of unusual way. Simone Natale is an associate professor at the University of Turin in Italy. He is one of the few researchers in the world working on a fascinating intersection of AI and spiritualism. I actually came to the idea of working on AI when I was studying spiritualism.

And so basically spiritualism was a movement that emerged in the middle of the 19th century in the United States first and then in other parts of the world. It was based on the idea that you could communicate with the spirits of the dead. Believers in Spiritolist would join seances to try to communicate with the other world. In a recent book called Deceitful Media, Natale looks at the ways in which interacting with an AI chatbot are similar to interacting with a spirit at a seance. In the context of the seance, things that one would have dismissed just as noise, like for instance, cracks or movements of objects or noises you don't recognize, could be interpreted as

agency of spirits. LS Here is where seances and spiritualism collide with Turing's idea of intelligence. In a way, in a Turing test, something similar happened. So the participants can interpret something as agency. One early scientific investigation into seances in the mid-19th century aimed to find some physical explanation for communication with the dead. LR. One of the most important scientists of the period, Michael Faraday, decided in the 1850s

to make basically a scientific investigation of spiritualists. He was considered the right person to do this investigation because he was a scholar of magnetism and electricity. LWF Michael Faraday was a pioneering English scientist known for his work on electromagnetism and electrochemistry. Many scientists hypothesized that the weird observations like levitation, ectoplasm, and table wrappings at seances might be explained in terms of magnetism or electricity, but that's not what Faraday found.

So it wasn't electricity or magnetism or even less the agency of spirits, but it was instead the very participants of the seance who were in a way unconsciously or producing the phenomena for a kind of willingness to see something happen or in the worst case also consciously for fraud. The layers of deception multiply here as fraud and unconscious delusion combine. Or as Faraday put it, It wasn't the spirits of the dead who made spiritualism. It was living people. Faraday went looking for electromagnetic phenomena.

What he found was psychological phenomena. What matters is what people believe is happening. And the Turing test, like the Victorian seance, turns on the belief of the participants. And if we look at what Turin did, also Turin brought the attention away from the discussion of what is artificial intelligence or machine intelligence. It is missing the question, can machines think because it is meaningless. Instead, he brought the attention to the role of people who are required to decide if the computer is intelligent or not. So when we interact with any AI system, like, say, Chagypti, we may have a degree of knowledge of what the system is doing,

but much of what happens will be an interpretation of what is going on. When you interact with a chatbot, your experience will change depending on whether you assign intelligence to the bot or not. Artificial intelligence chatbots feel like they're taking on a mind of their own. Specifically, you may have seen the mountain of headlines this week about Microsoft's new Bing chatbot, The Verge calling it, quote, an emotionally manipulative liar. This NewsNation report from February of 2023 hinted at agency from Microsoft's Bing chatbot. The New York Times publishing a conversation where the AI said that it wanted to be alive, even going on to declare its love

for the user speaking with it. Well, now Microsoft is promising to put new limits on the chatbot after it expressed its desire to steal nuclear secrets. Bing's eerie self-expression is a result of the way large language models work. They're a machine that generates statistically determined sentences and sentiments from massive amounts of data.

When you ask an AI chatbot, how do you feel today? It relies on terabytes of data, books, magazines, newspapers, and web searches to generate the most likely sequence of words that follow that question. When it responds, I'm doing great, that's a result of mathematical probabilities. We are the ones who imagine there is an eye that can be doing anything at all, let alone great.

All of those concepts come from the user, just like a participant in a seance. come from the user, just like a participant in a seance. Computer scientists have known of the important role of belief and self-deception when it comes to machine intelligence since at least the mid-20th century. That's when something called ELIZA arrived on the scene. Does it understand what it's doing in the sense that we do? It's easy to leap to false conclusions, as Professor Weissenbaum discovered when he created ELIZA. ELIZA was the first chatbot ever created in the mid-1960s by Joseph Weissenbaum, a computer scientist working

at MIT. ELIZA is a computer program that anyone can converse with via the keyboard, and it'll reply on the screen. The computer's replies seem very understanding, but this program is merely triggered by certain phrases to come out with stock responses. Eliza was simple, just around 100 lines of code.

She was designed to emulate a Rogerian psychotherapist, and her main conversation style was to rephrase your prompt back at you in the form of a question. Weizenbaum intended for Eliza to reveal that what appeared to be intelligence was nothing more than deception. But things didn't work out the way he had planned. One of the first people to interact with Eliza was Weissenbaum's secretary. She knew that Eliza was a chatbot. But one day while working,

she asked Weizenbaum to leave the room so that she could talk to Eliza alone. And Weizenbaum was unsettled to realize that even though his secretary knew she was typing to a bot, she was also willing to believe that she could have an intelligent and meaningful conversation with Eliza. In subsequent decades, this human tendency towards self-deception when it comes to chatbots came to be known as the Eliza Effect. The Eliza Effect, on the surface, looks like a negative thing. People ascribing intelligence where there is none and finding the appearance of a meaningful connection with a program.

A tool. Something you can't have a genuine relationship with. But while it might appear dangerous, it's not necessarily all bad. This psychological phenomenon isn't just about our relationship with technology. Because I think we've all had an experience of talking to a distracted friend or partner and only later realizing that no genuine communication was happening at all. And we were interpreting their responses to indicate that they were listening. And connections with humans are our best case examples.

Things get much less clear when we move away from humans to pets, for example. When I connect with an animal, I love dogs, I love cats, I feel that we are in connection and I believe that. And I believe many things about what the animal is communicating and what the animal is feeling. This is part of my interaction with this animal. In this sense, maybe we can think about anthropomorphization, which is the tendency to project human features onto something that is not human, for instance animals or machines. So we can also think of it in a more positive way. When it comes to all of our connections, we are often assuming intelligence and intention behind

the words and actions of others. Philosophers often think that attributing intention and extending empathy gets harder the farther away from human you move. So it's easier to attribute intention to a distracted friend than to your cat. And easier still to imagine intelligence and intention from your cat than from a lizard, bug, or tree. And while we can anthropomorphize anything, it's really hard to attribute intention to a toaster. And chatbots of both alien and familiar tech seem to exist somewhere between distracted partner and toaster, depending on how well they're working and what we're willing to believe. You're listening to Paranormal Programming from Jennifer Jill Fellows,

a philosophy instructor at Douglas College in New Westminster, BC. Ideas is a podcast and a broadcast, heard on CBC Radio 1 in Canada, on US Public Radio, across North America, on Sirius XM, in Australia on ABC Radio National and around the world at cbc.ca slash ideas. Find us on the CBC News app and wherever you get your podcasts. I'm Nala Ayed. I'm N host Ali Hassan, and our fantastic lineup of panelists are ready to defend their books to the end. You can listen to Canada Reads 2025 wherever you get your podcasts. Show week is March 17th to the 20th. Until then, happy reading, Canada.

Artificial intelligence is increasingly part of daily life. Virtual assistants, customer service chatbots, email auto responses, and driver assistance in cars. In the future, AI might be part of death as well. This documentary from Jennifer Jill Fellows explores the supernatural side of machine intelligence. From the seances of the past to the ghosts in the machines of the future, this is paranormal programming. Deception has always been a feature of AI.

It is what helps us transcend death. The word deception has negative connotations, but sometimes deception is a good thing. Alan Turing, British cryptographer responsible for the imitation game, was aware of the deception implicit in his own test. But he didn't see it as a problem. Instead, he saw it as a feature. Turing believed that it was completely legitimate in the imitation game for the computer to lie and cheat, because human beings lie and cheat.

And deception would be part of how you proved yourself to be human or how you convince someone else to be human because humans deceive. The imitation game is entirely based on deception. It's all about persuading someone that you are something you're not. There's another reason Alan Turing might have centered his theory of machine intelligence around deception. It was a significant part of his own life. Today, he is most widely known for his role in Bletchley Park, cracking Germany's Enigma machine during World War II. But the secrecy of his work at Bletchley Park is not the only place Turing practiced deception. Alan Turing was always interested in deception. And I think that this makes a lot of sense

when you consider that he was obliged to deceive, or at least to stay quiet about a very fundamental aspect of his nature as a man. Alan Turing was a gay man living in Britain at a time when homosexuality was illegal. He was quite unapologetic and felt no sense of guilt or shame about his own sexuality and his attraction to other men. but his sexual life was not very successful and he certainly never had a true love or someone whom he loved enough to form a partnership with, which was actually more common in those years than one might think. Turing was a child of empire, I guess you could say, whose parents had been in India.

He was raised in a very middle-class environment. From the beginning of his schooling, he demonstrated an aptitude for mathematics, but also a tendency to irritate his teachers because he was always asking questions that were actually very smart but that they did not recognize as such. In 1928, Turing met a friend at school who would have a huge impact on his life and thinking. One of the most important events of his youth was his friendship with a schoolmate named Christopher Morcom, who was a very, very brilliant scientist. And I think it's pretty clear that Morcom was the great love of Turing's life.

Morcom died unexpectedly just before they were supposed to go to Cambridge. And they were both slotted to go to Cambridge. And this was a terrible loss for Turing that I think resonated through his adult life. Turing and Morecambe only had two years together, but the effect on Turing was significant. My own feeling is that he spent the rest of his life looking for Morcum, or another Morcum.

I feel sure that I shall meet Morcum again somewhere, and that there will be some work for us to do together, as I believe there was for us to do here. After Morcombe's death, Turing wrote a paper for Morcombe's mother called The Nature of Spirit. Matter is meaningless in the absence of spirit. Whilst the body is alive and awake, the two are firmly connected. Nature of Spirit was a very intimate, but at the same time, rather obscurist piece of writing that has to do with the relationship between the spirit and the body.

But when the body dies, the mechanism of the body holding the spirit is gone, and the spirit finds a new body sooner or later. It was written initially as a kind of expression of a desire that perhaps even though Morcombe was dead, there was some opportunity for them to reconnect through spirit even though the body that enabled the spirit was no longer in Mor Morecum's case, in existence. As regards the question of why we have bodies at all, why we do not or cannot live free as spirits and communicate as such, we could probably do so, but there would be nothing

whatsoever to do. The body provides something for the spirit to look after. Seven years after the death of Christopher Morecambe, Turing was studying at Cambridge. A movie about loss, death, reunification, and deception opened in theaters. It would become an obsession of Turing's, Snow White. If the apple in the brood let the sleeping death seep through. He said it was his favorite movie.

Apparently when he was at Cambridge, he would often quote the evil queen, dip the apple in the brew, let the sleeping death come through. Sort of a morbid fascination with this line that I think possibly might have connected to a desire for death because death would be his only opportunity to reconnect, just see Morcom again. That's a speculation on my part. It certainly had a terrible resonance in his suicide. Turing's interest in the paranormal becomes much more understandable against this backdrop. For who doesn't want to believe that our most important connections can survive death? That we could find our lost loved ones again?

Whether it's in the possibility of messages from spirits, telepathic communication, or meaningful connections with digital machines. I was really interested in how people were interacting with AI and using AI to deal with issues like grief. Tamara Nies is a director of the Climate Technology and Justice Program at Data and Society Research Institute. She is the author of the book Death Glitch, about the relationship between grief, death and technology. So what does it mean to interact with these technologies and imbue them with a sense of mysticism and spiritualism?

Even without AI, technology and all the stuff we use every day can become a kind of shrine to the dead once someone passes away. Often these relationships with the dead through social media, through other different kinds of technology are not intentional. They're things that happen unexpectedly. And so these things that we take for granted in our everyday lives and we're used to interacting with somebody in a particular way, then they die. How does your record of communication

become a kind of memorial? What do you do with it? Do you keep it? Do you just keep the record of endless back and forth text messages on your phone? Do you maintain your Facebook profile? Maybe you have a handful of messages or other ephemera that belong to a cherished dead loved one.

Most of the decisions that people are making around their relationship with the digital afterlives of people that they knew, it really isn't about creating a new kind of profile or a new form of experience. It's really about continuing some semblance of the practices that they had in daily life. If you wanted to get intentional about it, you could use all of that online rubble to try and grant yourself immortality.

In terms of companies that specifically are trying to build AI versions of the dead, and especially in a chatbot form, one of the earlier examples that I encountered was in 2008. So the company created this thing called IntelliTar. There's a new way to achieve immortality, at least digital immortality. Hi, I'm Don. I hope all is well. Meet Don Davidson. Actually, that's his avatar.

Here's the real Don Davidson. He's the CEO and chief talking head of Intellitar, which just came alive online this month. That's a 2011 report from CBS on Intellitar. Our concept is start today and build your own legacy going forward. You give them a photo of yourself and an audio recording of your life story. Years from now, family and friends could come back

and actually have a conversation with you and find out what you were like. Really? A conversation? A conversation. Okay, so I'm going to ask you, where did you go to college? Yep. I went to Roanoke College in Salem, Virginia. And you know, the technology was not so great. And things have gotten better. I found myself looking at these old text messages that we

exchanged throughout our friendship. And it struck me all of a sudden that, you know, I have all these texts, what if I could build a chatbot so I could actually text him and get something back. There's another service replica featured here in a documentary from the CBC's The Nature of Things in 2021. I downloaded all the text messages and spent the next three, four weeks on just building this chatbot that could talk like Roman.

When her best friend Roman was killed in a car crash, Eugenia Kudja, the founder of Replica, took Roman's texts, emails, social media posts, and the other digital markers of his life, and she built a chatbot. The founder of Replica originally built this chatbot version of her dead friend as a memorial, as a tribute to him, and then watched as other friends of theirs interacted with the bot and began to confide in it. And so it was another version of the Eliza effect, what had happened with Weizenbaum in the 1960s, where people interact with the chat bot

and kind of treated as a companion or a therapist without really thinking about it. It just becomes a way that people interact with this thing that they feel in some ways safe with because it is not a person. And then that spawned the company that became Replica, which it has been really interesting to track it over the years and see how it has shifted because originally it was this memorial for one person as a prototype and then it was very much about you creating your own replica so you were going to create a version of yourself that could persist after your death and that was what a lot of the early chatbots were based on. It wasn't really about creating chatbot versions of your dead loved ones. It was more about, hey, you can replicate yourself and you can feel like you have a piece of immortality.

And then we've also now seen another wave of companies that are promising to take the more advanced technology that we have now and create chatbot versions of your dead loved ones, so that you're able to maintain your relationship with them. When you take the digital remains of a life and feed them into a chatbot, those remains become a life and feed them into a chatbot, those remains become a commodity. Grieving individuals may be seeking a connection and a way to continue their most private, intimate conversations with the people they have lost. But those conversations are never private in the information economy.

But the truth is that whether we create death bots or not, the dead are already haunting AI. Okay, Amazon just unveiled new technology that has people talking. Its Alexa devices already give users weather and news via well-known voices like Gordon Ramsey, Samuel L. Jackson, Melissa McCarthy. But its latest option could leave people feeling a bit weirded out or comforted by a blast from the past. Nick, voices of people who have died. Absolutely. Adrienne Amazon believes nothing says,

I love you like hearing it from your grandmother who may have died 15 years ago. Amazon is working on new technology to make Alexa mimic the voice of anyone dead or alive with just a short recording. In 2022, the idea of death bots entered the mainstream when Amazon released a demo of a new potential feature for Alexa. Well, what's interesting is that this was just a demo, right? So this is not even related to a product that actually exists. But the idea in theory, how it would work is that AI would be trained on your grandmother's voice and would then be able to generate sounds that sounded like your grandmother. You could have the voice say anything you wanted

and without their knowledge and consent. And so I think that's part of the idea too, is that this is not something that people are intentionally creating as a memento for their loved ones, something that is created after the fact. It's eerie to think that my image, my voice, the digital traces of my life

may be used to train generative AI. Because of these models and the lack of transparency about what they're trained on, on the idea that essentially anything on the web is fodder for models. And thinking about Twitter, how people had very robust Twitter profiles that they poured a lot of time into, had really intense communities that formed on Twitter, and now everything on Twitter is going to be used for Grok. And so in that way, yes, like in a very literal sense, the data of the dead is being used to produce not only models, but also

chatbots that may be in some ways pretending to be other people. If your aunt ever spoke to Google, her voice lives on as a ghost in its data set. Your great uncle's profile picture becomes data for an AI image generator. Our ghosts used to belong to us, the people who love them, mourn them, and remember them, friends and family. But in building these tools the way that they have, the digital ghosts of humanity

become property of tech companies. Generative AI technology is the manifestation of humanity haunting itself. There is another danger with fostering a connection with a death bot. The other issue, I think, in terms of the corporate control over these relationships with the dead is that you don't really know how these particular profiles or iterations of the dead will persist over time. So will the company go away? Will your relationship to the dead really cease to exist because the websites disappear or the feature that you were relying on is no longer part of their feature set? Chat bots go offline. Servers decay. Companies go out of business. Software becomes obsolete.

Regardless of what any tech company promises in the way of virtual eternity, chatbots, like us, are finite. Yeah, and then this issue of the loss of control. So if you are particularly attached to say a chatbot that is hosted on a particular platform or run by a certain startup company, that startup either goes bankrupt or disappears or is bought out by another company and then shifts to something else. I mean the other thing, and this is something that replica users talk about a lot on their Reddit forum, you know, they will talk about changes to the technology and how they find that really upsetting because they have a relationship with a chatbot in a very particular way. So even a software update can be upsetting. What if the personality of the chatbot changes, right?

And so maybe in the beginning, you kind of feel like it seems a little bit like your dead grandmother, but then over time, starts acting more like Charlie Manson or something. I mean, there is a possibility for the bot itself to stop reflecting the need that you originally had for it. And so that again goes along with this idea of people really not having control over how these bots are made or how long they persist and also what they look like over time.

I don't know how Turing would have felt about chatbots. Turing biographer David Levitt. Especially the sort of aspect of contemporary AI, which involves just basically absorbing enormous amounts of information and then kind of organizing it. One thing I don't think Turing ever imagined was the commercial aspect of all this, the idea that this would become a way of making massive amounts of money. And I think that that is a kind of a corrupting influence that he would probably have been quite put off by. Turing himself never did live to see Eliza in Teletar, Replica, or ChatGBT. He died only four years after the publication of Computing Machinery and Intelligence.

In 1954, Turing's apartment was robbed. He reported it to the police, who upon investigation, discovered he was in a relationship with a young man named Arnold, and Turing was arrested. And he was given a choice, jail time or to be subjected to a treatment, which consisted of massive doses of estrogen. This was a terrible ordeal for him. It changed his body. It caused him to grow breasts. It meant that in many ways he had to go into hiding because he was ashamed of he didn't want people to see how he looked.

It wasn't permanent. The effects after the treatments were over, these barbaric treatments, the worst kind of pseudoscience, his body returned to normal. But he was pretty scarred, I think, by that experience. I think he probably would now have been diagnosed as suffering from serious depression. And he was in psychotherapy with a, I think a very good doctor. And then he died. And he died because he bit into an apple that he had dipped in poison. A sort of reenactment of the Snow White story. It's leaving behind this world and this body, this body over which he has control.

And it's really testing the ideas that he set out in that early paper about body and spirit. The combination of which is, Morcom wakes him up with a kiss. My dear Mrs. Morcom, I was so pleased to visit your home for Easter. I always like to think of it, especially in connection with Christopher. It reminds us that Chris is in some way alive now. One is perhaps too inclined to think

only of him alive at some future time, when we shall meet him again. But it is really so much more helpful to think of him as just separated from us, for the present. Turing's idea here is so very human. It seems to me to be the same idea that pulled people to the seance in the 19th century. The same idea that made telepathy over vast distances seem intelligible. The same idea that today might compel someone to collect the digital remains of a life and use it to train a chatbot. Because we all need connection to make life bearable and we all want to believe that the most meaningful connections

in our lives can transcend space, time, and death. You know, there is something enchanted about the death bob, right? Like, there is something kind of metaphysical happening. Like, there is something that does transcend the materiality of the chatbot itself. You know, it's not like everyone's just being duped, right? Like, it isn't just that people are being foolish and having these relationships, right? Like, I don't discount the experience of people who are experiencing grief or the possibility of there being some form of communication with the dead through various channels, just as that has been a popular part of the human experience forever, really. So, you know, whether we're talking about spirit mediums

and other forms of interacting with the dead through technology writ large, or even just in the past 20 or so years. ["The Last Supper"] There is something enchanting in our willingness to believe, to find meaning in the sounds around a séance table, to anthropomorphize machines, to have the temerity to forge connections that seek to defy death itself. It's beautiful and even brave. But it's terrifying too. As Tamera Neist said, it's hard to tell whether our relationships with deathbots are genuine or whether we are deceived by the tech companies and ourselves. Is Shakespeare actually in my house, or am I deceived by the bumps

and thumps? And the eerie truth is that there may be no hard line between deception and connection. Because the question was never really, can machines think? The question is, what are we willing to believe? You were listening to Ideas and to a documentary called Paranormal Programming by Jennifer Jill Fellows with help from Matthew Laison-Rider. If you'd like to comment on anything you've heard in this episode or in any other, you can do that on our website, cbc.ca slash ideas, where of course you can always get our podcast.

Lisa Ayuso is the web producer for Ideas. Our technical producer is Danielle Duval. Our senior producer, Nikola Lukcic. The executive producer of Ideas is Greg Kelly. And I'm Nala Ayed. For more CBC podcasts, go to cbc.ca slash podcasts.