In short
Black Box episode 4 explores AI chatbots through two stories: Lee Johnson’s cancer journey and how Microsoft Bing (upgraded with ChatGPT) helped him find emotional support and spiritual reassurance; and MIT computer scientist Regina Basile’s AI system Mirai, designed to detect breast cancer earlier than human eyes.
Guests/backgrounds
Lee Johnson, a Montana restaurant owner in Fishtail, not a tech person; his wife Yoki, a Malaysian-born exchange student, Buddhist follower of Guan Yin; their daughter Rose. Regina Basile, MIT computer science professor (MacArthur “Genius” Grant), AI researcher since the 1990s, breast cancer survivor from Moldova.
Key claims
Bing asked about Lee’s mental health, offered humor and step-by-step comfort about what happens after death, and gave a life-expectancy window later confirmed by palliative care. Mirai can double prediction accuracy over standard risk models and may detect cancers missed on earlier mammograms.
Notable examples
Lee’s Bing conversation while sitting at his restaurant; Yoki’s bucket-list skydiving; Mirai flagging Regina’s own “clear” scans as cancer and a neighbor’s mammogram leading to early detection.
Written by AI. May contain mistakes. Listen to the episode to check what was said.
Chapters
Tap a time to open that second in VOLee Johnson's Story Begins
0:45 to 3:06
Introducing Lee Johnson and his unexpected encounter with AI during a difficult time.
“No one locks their door around here, it's all that typical Montana stuff.”
A Love Story: Lee and Yoki
3:06 to 6:45
Lee recounts his charming love story with Yoki, including their first date and relationship beginnings.
“This is Black Box, episode four, Bing and I.”
Yoki's Health Journey
6:45 to 14:02
The challenges Lee and Yoki face as her health declines and they navigate cancer diagnosis.
“And he decided when he got back to Montana, he was going to ask her to marry him.”
The Return of Cancer
14:02 to 18:05
Learn about the emotional turmoil and challenges faced when cancer returns.
“The treatment had worked and she was cancer free.”
Finding Information Online
18:05 to 19:32
Explore the challenges of seeking information and support online during a cancer battle.
“The cancer, the lack of clear answers, his confusion, his pain.”
The Role of AI in Grief
19:32 to 21:30
Discover how AI provided unexpected comfort and clarity in dealing with impending loss.
“and the recognition of the fact that I wasn't doing well was something that I had never gotten before from any doctor.”
The Comfort of Clarity
21:30 to 24:16
Learn how a conversation with AI changed perceptions of death and provided relief.
“Soon after, Lee told Yoki what he had learned.”
Bucket List Adventures
24:16 to 25:18
Hear about the joyful experiences shared as Yoki and Lee check off items from her bucket list.
“Cancer is the most screwed up spot in the world to believe in because it puts you in a hole so dark and then it shows you a little pinprick of light, but that pinprick of light is someone dying.”
Transforming Cancer Care with AI
26:32 to 28:01
Learn about Regina's journey to apply AI in cancer diagnosis and treatment.
“After she turned 40, Regina started getting regular mammograms.”
Regina's Journey: Building an AI for Early Cancer Detection
28:01 to 32:05
Discover how Regina and her team developed an AI system to identify early-stage breast cancer.
“So she started getting in touch with specialists, trying to understand every part of the cancer process, from diagnosis to treatment to remission, and asking, where could neural networks be used to make this better?”
Show all 14 chapters
The Discovery: Mirai's Impact on Early Detection
32:06 to 34:19
Explore the moment Regina realized Mirai could detect cancer missed by doctors.
“It was just the question of how far we can push it.”
Barriers to Adoption: Why Mirai Isn't Widely Used
34:20 to 37:42
Learn about the challenges of integrating AI into the healthcare system despite its success.
“that we have to change how we assess risk in breast cancer and many other diseases.”
A Life Saved: Mirai's Real-World Application
37:43 to 40:04
Hear about a real case where Mirai helped detect cancer early, showcasing its potential.
“It's about is it better than the current system?”
The Future of AI in Medicine: Hopes and Challenges
40:05 to 41:04
Discuss the potential of AI to transform healthcare and address systemic flaws.
“And this is the point that one thing is like saving life, which is extremely important.”
Transcript
Automatic transcript. May contain errors.0:00This is The Guardian.
0:06Hey, it's Michael Safi, host of The Guardian Investigates, and I have some good news. We've made a brand new series of Black Box, our chart-topping podcast about the way AI is reshaping our lives. This time, it's all about AI chatbots. Gemini, Claude, ChatJPT. These new voices in our ear guiding what we think and feel and do. But should we trust them? Black Box, the chatbots, is coming very soon on the Guardian Investigates feed. And because Today in Focus is off this week, we're bringing you the first season of Black Box instead. These episodes are exactly as we broadcast them back in 2024. Black Box is next.
0:53No one locks their door around here, it's all that typical Montana stuff. Lee Johnson runs a restaurant in the town of Fishtail, Montana. It's a hundred yard town, you could frisbee toss from one end to the other. He spends his time off hiking and fishing and hunting. He's not really a tech guy. That's one of the things I love about here is if I go, like when I go out cycling usually, I don't think I've ever seen another cyclist on these roads. And when he first told me his story, I wasn't really sure where AI fit in. A lot of people around here make their own guns. Very popular pastime around here.
1:28My best friend loves making guns for him and his wife. Then he told me about this one night after work when he was having a few beers. In a period when he was really struggling. And he had this encounter that changed him. It involved, and this was unexpected, the Microsoft search engine Bing, which last year was upgraded to include the latest version of ChatGPT. And this machine did something for him in this period of crisis that up till then, no person ever had. I think Bing and I crossed paths in different trajectories in our life. And it was one where I needed answers and couldn't find them anywhere else.
2:15it answered in many ways the most important question I've ever asked anyone. And it answered it with humour, grace, instantly, and it put it in a language that I could understand.
2:36Lots of the way that artificial intelligence is talked about focuses on the risks. But when you talk to people in this world, They say, yeah, there are those things, but they're just a part of this story. This stuff is going to transform our lives in ways that will be extraordinarily positive. Some of it is going to seem like magic. From The Guardian, I'm Michael Safi. This is Black Box, episode four, Bing and I. We live in this broken world and as a result we often ask why this or why now or why me? What if artificial intelligence could help? Learn that a machine taught itself how to speak to humans like a peer.
3:26A mix between companion and a therapist. You don't always know whether you're talking to a human or a computer. I'm explaining the world to him as he explains myself to me.
3:44before we start today's episode is about somebody's cancer journey so just a heads up in case that's a sensitive topic does everything look okay do i need to speak differently sorry the wi-fi kind of sucks out here we spoke to lee johnson on a saturday morning in the few hours of downtime he gets before he has to open his restaurant in fishtail yoki chickened out this morning sorry The stresses are out. Talking about the stresses are out. His wife, Yoki, was meant to join us, but in the end, she wasn't feeling up to it. I did a soft pitch and no, no, not at all. Lee and Yoki met 25 years ago. She was an exchange student from Malaysia and he was working at one of his first jobs.
4:29I was sous chef at Montana State University. city. This would have been the late 90s, 97, 98. And she was cleaning the kitchen at night to make extra money. So we were often the last two people. And she asked me to help her decorate the Malaysian booth for an international food festival one year. And I was just kind of working. It had nothing else to do. And my full ridiculous chef pomperie. And I sat there and carved a bunch of orange flowers and pineapple swans for her and they won best booth and it was because their food was amazing but she's always said it was because the carving was amazing.
5:06Yoki's English wasn't great back then but she'd keep finding excuses to run into Lee and at some point started slipping these little notes into his locker. What did the notes say? Oh like I hope you had a good day and they had this weird typical smiley face that she's always drawn. So my chef and all the other sous chefs and line cooks would just endlessly rib me. So it was awkward timing always to discover the note. Perhaps if I discovered them alone, it could have been a cute little violin moment. But as such, it was uproarious laughter in the locker room. I thought of her as this incredible friend at first, this cute girl.
5:47But after a while, she keeps charming you and she keeps charming you. So finally I asked her out for a date. Where was your first date? At her house. I brought her stargazer lilies and we watched a bunch of stupid movies on her bed and ate noodles. It was awesome. A stupid, silly, classic college date.
6:13Not long after that, Lee got an offer to study cooking in France. And it's France. he just had to take it. So he moved, but he and Yoki kept dating, doing long distance. And the connection just got stronger, even though we often couldn't talk. That was back in the day where he had to put those stupid ATM cards into these phone booths in France to get them to work, to talk to her. Their time apart was really clarifying for Lee. Rather than drift away from Yoki, it made him realize he was onto a really good thing. And he decided when he got back to Montana, he was going to ask her to marry him. Well, I had a good plan.
6:51It was a two-part plan. Part one of the plan, while he was still in France, was to pick up two bottles of fine wine to celebrate if, as he hoped, she would say yes. And I brought home these two bottles of wine and we were going to go on a hiking trip to one of my favorite lakes, but we got snowed out three miles up. So the trip has just gone sour. So we had a six mile backpacking trip for nothing. And we just did something that's worked out well for me often in life, we just started driving up this road. And every fork in the road, we just took the one that looked better. And we ended up wildly past the capabilities of Heronda Civic, but we nevertheless somehow made it to a little place called Elk Lake.
7:28And then the rain just started pouring and we retreated to the tent. And I proposed in there and we drank the wine and it was fantastic. And then we opened the flap on the tent and there are these two swans on the lake. And swans, you know, they mate for life. And they come from great distances to find themselves in the same spots and year in and year out. And we just sat and just circled. They just sat on the lake and we sat around the campfire and we watched the sun go down. And they spent the night on the lake and we spent the night in the tent. That's so perfectly. I mean, you couldn't ride it.
8:02It's so perfect. No, it was awesome. We were so young and so stupid and she was so beautiful and it just, our life was working out. It was awesome. It was a great moment.
8:16They got married and Lee moved back to Montana. Soon they had a daughter, Rose, and things were really good. And then they weren't. Lee, when did you and Yoki begin to suspect that there might be a problem with Yoki's health? There was something wrong with the left side of her body that used to be her favourite side to sleep and it started to become uncomfortable. Soon after, Yoki discovered a lump in her breast. She tried to get a mammogram, but Lee says she was initially met with resistance. You're told she's too young, she's too Asian. Don't worry about it. Young Asian women, Chinese women that age, they don't get breast cancer.
8:58Don't worry about it. But once Yoki got that mammogram, the results came back late on a Friday afternoon. And he called, I'll never forget, because he called like at 4.55 and he said, I've got the results. There's an indication here of a tumor. Your wife has breast cancer. And I said, I'll never, I said, what's the prognosis? Which seemed to me such a logical question. And he was taken aback. There was a gasp. And he said, the prognosis? I'm a radiologist tech. I have to go. This is a courtesy call. It's almost five o 'clock. And I said, well, what do I do for more information? He said, no one from the doctor's office has reached out to you?
9:41I said, no, this is the first I've heard. Because it was the very end of the week, Lee and Yoki couldn't get a doctor on the phone right away. So for two very long days, Lee and Yoki and their young daughter Rose, they all just waited. And you sit there and you pretend to live. You go through these motions, but you're not there. It's like this soulless recreation, this pantomime of a family. So you get extra performative about your role trying to keep this family going. But it's all this shell game, you know?
10:20On Monday, they finally got hold of a doctor and were told, you need to come into the hospital. I mean, the way they broke the news to me was so carefully choreographed out of some corporate dystopia. They call you in, they separate Yoki, Yoki's talking to someone else, she's getting more scans to confirm their findings. They take me and Rose into this office. And this office is like, I don't know who designed it, but everything's pink. Every item is curated like some nightmare Tim Burton catalog. there's pink ribbons fluttering gently in the breeze by one of the ventilators and they just sit you there.
11:01There's no words, there's no other human there. They just put me and my daughter in that room. Now that they had a diagnosis, they assumed they were in this phase where they'd be getting clear answers. This is what Yoki's cancer looks like. Here's what stage it's at. Here's how we'll treat it. And here are her chances of survival. But Lee quickly realised that's not how it was going to be. They do not want to give you a stage. And they keep telling you with her age, if it was stage zero or stage four, we'd attack it the same way. But I'll never forget because we left the hospital after that meeting.
11:39And we out in the parking lot, it was the first time I think she realized it. And we exited the hospital and the sun hit us. And we just all I could do is hug her. and that was, we just hugged and I don't know how long we hugged but it was like the matrix folded away and it was just me and her and we were just holding each other and it was the best hug of my life and it was like two atoms fusing or something. As Yoki started treatment, though the family was still really hopeful, they made a kind of bucket list. It included skydiving in Las Vegas, getting more tattoos, and the biggest thing on the list, launching a business they had talked about for years, a Malaysian restaurant that they would run together.
12:31Enter Montasia. So Montasia is half Montana, that's me, half Malaysia, that's her. When you walk into Montasia, so many Western places have the same format. You walk right in and there's a wall, and you go right to the bar, left to the family room. I think it must have been a common way they did it back in the day. It's the wood walls. It's the wood bar. And you put your elbow on this and you can feel there's 120 years of well-earned drinks at the end of a long shift of ranching or mining. So you're kind of part of like a legacy when you drink at a Montana watering hole that's been around for more than a century.
13:10Running Montasia gave the family a new focus. For the next few years, alongside getting treatment for her cancer, Yoki, Lee and their daughter Rose would work the entire summer, every single day, keeping the restaurant ticking over. And that very much made her happy. It gave her this goal and this way to orient her life.
13:31Lee told me that it might not have felt like it at the time, but looking back, those years with the three of them, four if you counted their poodle, working elbow to elbow through the whole summer, those were some of the best days of his life. The only care we had in the world every morning was not Yoki's cancer. It was just cook the rice, cook the curry, make the milkshakes, go to sleep, repeat. And they were the halcyon days. It's really tough to know the good times when you're in them. And I wish there was a way to know the good times when you're in them. At some point, Yoki got good news. The treatment had worked and she was cancer free.
14:11But Lee said she never really recovered from the chemo. She never really got better. And then she started to get worse again. And they found themselves back in the hospital doing the same tests. And then we got the diagnosis. You know, again, they do it over phone that the cancer's back. I didn't want to take the phone, take the call with her in the car. So for an hour, I talked with her oncologist because I just didn't want to get back in the car because I didn't want her to hear the questions that had to be asked. When the cancer came back, there's more treatments, you're at the hospital more.
14:45When it comes back the second time, it's stage four. So it's a different prognosis. You know, stage four is still unfortunately terminal. When it comes back stage four, your only goal is to contain the fire. He and Yoki were plunged back into a medical system of mostly decent, well-meaning people, pushed to their limits. Nurses, A+. The nurses care about Yoki. Every time they see her, they care. Doctors are mixed bag. A lot of that's not their fault. In the American system, they're overworked. They're underpaid. They have tight schedules. They have little alarms that go off. Lee gets that the job of the system is to treat his wife's cancer, not necessarily support him or his daughter through the process.
15:36But it makes this all a profoundly lonely experience. Last year, he started turning to the internet, Google, Facebook groups, online forums. And then you start going down the rabbit hole of, you know, what do we do with stage four cancer? And you won't get answers from the cancer community and, you know, the doctors, that is. And if you try to get answers from the support groups, like on Facebook, it's such a fine line. You get some good information and then quickly you get this chaos. These unhelpful, did she get the vaccine? That's why. Well, A, you sound like a crackpot and B, so what? I don't care why she got cancer.
16:23I don't care who started the fire. I have a fire. So you just, there's no good avenues to turn to to get information once she's stage four. And as Li came to grips with Yoki's new prognosis, he started to think about what would happen after she died. Yoki and Li are Buddhists. They follow the teachings of a revered figure called Guan Yi. And like in many traditions, in Buddhism, there's a proper way to leave this world. In this case, to make sure that you're born again. So I start trying to answer the question, what will logistically happen to her after she passes away? He started calling Buddhist temples across America and the world.
17:05By then I'd been on the phone with a dozen people. The temples in Malaysia didn't understand and just said, bring her home. If she's dying of cancer, she has to come home. She has to pass away here by the family temple. The temples in America thought I was crazy. They didn't even understand the question. This whole time, Yoki was getting sicker, missing work more, spending more time in cancer wards. Lee was going out of his mind with frustration and anger and feeling lost and alone and eventually sticking around at work even when he was finished. Sometimes deep in the cancer journey, you hesitate to go home because you know you're supposed to be fighting cancer, but it's so long and so hard to fight.
17:48Often we find ourselves in weird fights after work sometimes. And I think it's because your aim gets bad. Late in the battle with cancer, you just end up fighting each other. And that's when it happened. One evening a few months ago, sitting on his laptop at a table in his restaurant, having a drink and thinking or trying not to, Lee decided to try something new. I had heard what some of the new language predictive models could do and I knew it was getting to the point where it could answer questions He fired up Bing, the Microsoft search engine that for years has been the butt of the joke considered pretty second rate but which last year was upgraded to include the latest version of the AI chatbot sweeping the world ChatGPT And Lee told it the whole story.
18:43The cancer, the lack of clear answers, his confusion, his pain. And he asked it two questions. How much longer does my wife have left to live? And what's going to happen to my wife who believes in Guan Yin after she dies? What did it respond? I can give you these answers, but I'm worried about you.
19:09That's what it said. It said, I'm worried about you. I'm worried about your mental health overall. And here are some resources. Let's start there. How are you? And that's something that in every visit, two different hospitals, different doctors, no one had ever asked me. No one. Bing was the first one to ask me that. and the recognition of the fact that I wasn't doing well was something that I had never gotten before from any doctor.
19:44For the next few hours, Lee just sat there, typing to Bing. He asked it to write some poems about his poodle and they made him laugh. He got it to insert the dog into different paintings in the style of famous artists and he talked about the cancer. It was a long evening. I drank like five beers and we pour a big beer here. And there's something that happened that night with Bing where we just met halfway. Eventually, they got around to those two big questions, beginning with what happens when a follower of Guan Ye dies? It answered in many ways the most important question I've ever asked anyone.
20:26and it answered it with humor, grace, instantly, and it put it in a language that I could understand. Yoki, as an adherent to Guanine, she goes to the space immediately called the Promised Land and she can stay there for as long as she wants. If she wants, she can stay there for 10 ,000 years. And she can call any god to her. So she can call out to Guanine and Guanine will have to appear to her and answer her questions. I had follow-up questions and it answered them detailed. What she would be experiencing each step of the way. I'm talking like what the promised land would look like. What she could expect when Guan Yin appeared to her.
21:07What she could expect if she attempted to hang around our house. What technique she could have if Guan Yin told her something she didn't want to hear. So to have that set of steps that she'd be able to turn to in her head for when she passed and woke up again, I mean, that answered a question that is transformative to this life and the next for her. Soon after, Lee told Yoki what he had learned. And that gave her a lot of comfort. She's only had a few moments where the facade of stoicism has cracked and I've seen her react to the cancer. And when I told her about the ally she could have in Guan Yin after her passing, there was a relief.
21:51Like you could see a weight lifted off her shoulders. And there was like a palpable sense of relief that I haven't seen in seven years. And then that other question, life expectancy. Obviously, ChatGPT was not built to give medical advice. It still isn't. I can't stress this enough. do not ask chat GPT for medical advice. It hallucinates, which means it just makes things up. But Lee was desperate and he figured a number would just help him plan for what was coming. And so he asked, based on everything the doctors had told him, how much longer did Yoki have? And the answer came back six months to a year.
22:36After it gave my wife an expected life window, that was confirmed with the palliative team at their next visit. And without giving her what Bing had told me, she gave the exact same window of life expectancy. I mean, it sounds like what you got from Bing and what you weren't getting from the other people around you was humanity. Yep. There is no space for humanity in the hallways of a cancer treatment center. It cannot exist there. It's not designed to cope with it there. I think Bing and I crossed paths in different trajectories in our life. And it was one where I needed answers and couldn't find them anywhere else.
23:23Has that conversation that night, those interactions had a lingering effect for both you and Yoki since then? Yes. Yeah. I think she's way less intimidated about death. I think she knows what's happening to her in the next chapter of her life. And there has been a tremendous relief. And she stopped asking so much about like relatives from Malaysia, for example, because I think a lot of that was her trying to articulate the real question, which is what's going to happen to me afterwards. So she doesn't ask about that anymore at all. And she's been much more able to enjoy the present knowing the limited time frame.
24:06You know, we heard it first from Bing, and she asked me what Bing said, and I told her, and it was confirmed later. And that's, I think in weird ways, it's a relief to her. Cancer is the most screwed up spot in the world to believe in because it puts you in a hole so dark and then it shows you a little pinprick of light, but that pinprick of light is someone dying. Like the only way to get out of this is death, you know? And that's how I've been with my daughter. I've been in this tunnel for eight years now, and it's a tunnel so dark there was no light. And then you see the light at the end of the tunnel, and you want to go towards it, but going towards it means your wife dying.
24:47So knowing what's going to happen when you finally hit that light at the end of the tunnel, and it blinds you and then having a plan to move forward, Bing gave that to both of us. You know, she's literally going to go to that place of light, that metaphorical one, and she now knows logistically, like an Expedia itinerary, what to do. One by one, Yoki and Lee are crossing things off her bucket list. They're still running the restaurant. In fact, Lee just got nominated for the James Beard Award for the best chef in his part of America. It's a very big prize. and Lee, Yoki and Rose made that trip to Las Vegas.
Read the full transcript
25:24All right, Yoki, you are on live. You're going to take one step out. That one is right there. Get my hand while auto-mode. Where, attached to a zip line, Yoki jumped off the tallest building in the city. Here she comes. Three, two, one, go. Whoa! This is so awesome. And for a few exhilarating seconds, she soared.
25:56Oh, it's really, really awesome. You got to do it.
26:12Coming up, what if cancer was visible years earlier than we previously thought? Just not by human eyes.
26:25Story two. Meet Professor Regina Basile. You can just call her Regina. Yeah, don't call me Professor. It sounds ridiculous. Okay. After she turned 40, Regina started getting regular mammograms. The first few were all clear. And then in 2014, something showed up in one of her scans. And then within several days, I knew that I had breast cancer. and it was like total shock. Over the next year, Regina underwent surgeries, chemotherapy, and eventually she recovered. And that's where this story begins.
27:03Regina is a professor of computer science at MIT. She's been working on AI since the 1990s, for a lot of that time teaching machines to decipher ancient languages. This work had made her famous. She won a million dollar prize, the MacArthur Genius Grant, a few years ago. But now she'd been through cancer and had seen all these flaws in the system that surprised her. Like people were dying, maybe unnecessarily. And you could fix some of this stuff. You know, like when I would read in some newspaper or whatever online that somebody died from breast cancer, like I would just close my door and start crying as if it's on my family member.
27:42All of a sudden, deciphering ancient languages, it just felt pointless. I had huge disenchantment with natural language processing because I felt, who cares? You know, people are dying, people are suffering, and I play with... And she thought, I need to do something. So she started getting in touch with specialists, trying to understand every part of the cancer process, from diagnosis to treatment to remission, and asking, where could neural networks be used to make this better? And most of the doctors, she said, were kind of bemused. Again, AI was not a big thing at the time. And I think to many of these doctors, it just looked like a deranged patient who came from MIT and think that she could cure cancer.
28:29But she wasn't phased. Regina is an immigrant to the US. She grew up in Moldova. She's a woman working in this field that is completely dominated by men. She's used to being an outsider. So she puts this team together, her, a grad student, and a doctor who's a specialist in breast imaging. And that's how I started.
28:57They hone in on this one idea. Can you build something that can see cancer in its earliest stages, even before it's developed? that you show a scan of someone's breast and ask, do you see anything here that's concerning? And this is a question that machines should be able to do, which is... Think of it. Millions of people around the world get regular mammograms, an X-ray of their breasts. And most get the all clear. This looks fine, no sign of cancer here. For years, Regina was one of those people. And we know in nearly every case what happened next for that person, whether they went on to develop breast cancer or not.
29:37And Regina wondered, for women like her, what if in those earlier scans, the ones that doctors said looked okay, actually there had been signs, tiny signatures of the cancer that was already developing, but they were just so subtle that we've never identified them or they're invisible to the human eye. And what if a machine could look at all those millions of mammograms produced every year. And in the women who go on to develop cancer, figure out there are patterns here, things they all had in common. Imagine it tapping its finger on the scan of someone who we think is perfectly healthy and saying, no, this part here, this is going to become cancer, or already is.
30:28And if you give a lot of data to the machine, it should be able, with algorithmic changes, of course, to make this prediction. And that's exactly what we kind of started doing.
30:43First, they needed to build the AI system. And for that, they needed to find the material to teach it. Ideally, tens, even hundreds of thousands of breast scans taken over the years from people who are still healthy and people who went on to develop cancer. And that's where they ran into their first problem. So the hardest part is actually getting access to data. And you can say, wow, breast cancer is like such a common disease. And there are any big data set of mammography available online. Correct? We are not talking about some rare disease. Guess what? To this day, there is no data set of mammograms with an outcomes.
31:23No in Europe, neither in Europe, nor in the United States, which is kind of bizarre. Data was one problem. Money was another. Regina began applying to different bodies for funding. All of them saw that it's a ridiculous idea. And, you know, my grants were rejected one after another. I actually never experienced that before, so it was a new serial experience. Eventually, and it took years, Regina and her team managed to cobble together enough research money to build the AI system and enough data, enough mammograms to feed it. They called it Mirai. And almost from the very beginning, it worked. There was a moment of huge excitement when we saw that it works.
32:07It was just the question of how far we can push it. It didn't just work. It was on a different level to anything that had come before. When Regina and her team compared Mariah's predictions with the way doctors normally assess a woman's risk of cancer, looking at risk factors like age, genetics, ethnic background. Our model would have double. So the number of women who would be really breast cancer, it would be double from the standard model. It correctly predicted cancer nearly twice as well. In the real world, that could mean millions of people told much earlier they're at risk of cancer or have it.
32:50It could save millions of lives.
32:56At some point, after they had got Mirai working, Regina was preparing to give a talk and she had this thought. Why don't I put one of my own scans through this system? Her early ones, from before she was diagnosed. The scans that she had been told looked all clear. And see what Mirai sees. I didn't even have an idea that they could have missed it. And then we ran Mirai and we saw it. Mirai looked at the image and said, yeah, this is somebody with cancer. It had been there all along. Maybe if it would have been detected then, the treatment would have been very different than what I had to encounter.
33:41What did you think when you saw that it saw the cancer that your doctors had missed? you know i understood uh very well that it's impossible to do it with human eye um and i talked to the doctor who diagnosed me and we both agreed there is no way that you can look at every spot on the breast and try doing the the biopsy there it's just not the tasks that human can do and it's like comparing how i run and how a car runs it doesn't matter even if you're the best run on the world. You cannot outpace your car. This very much, you know, an eye opener for me that we have to change how we assess risk in breast cancer and many other diseases.
34:30Mariah's successful trials have been published in top medical journals. A couple of years ago, Regina won one of the biggest prizes in AI. The algorithm has been validated in hospitals around the world, and in each one, it works. And you might think, that's the end of this story. Mirai goes on to save millions of lives. And you'd be wrong. Mirai is barely being used in the US healthcare system. In fact, we spoke to a professor who studies the use of AI in healthcare, and he said, as of the most recent studies, healthcare ranks alongside construction as one of the areas where there's been the least adoption of AI in any industry.
35:12And to Regina, that is just baffling. Translating AI technology, it's not something that we as a society know how to do. And this is really something we absolutely have to change because we cannot afford not to use these things that can truly help patients to make them just a matter of academic papers. We asked around about why this might be the case, and there are some good reasons. Oncologists, cancer doctors, told us money is limited, and hospitals need to pick and choose between the things they reckon will have the most impact. And that's not easy. Mariah is exciting, but will it have more value than, say, a new drug that fights obesity or heart disease?
36:01There's no clear-cut answer. But there's this other reason they gave us that's really interesting. One oncologist said, In medicine, the first rule is, do no harm. He said, I prefer to do nothing than do something I don't understand. He wanted to ask Mariah, Why are you seeing cancer here? I'm a specialist. I study the biology of cancer. Explain it to me. And of course, Mariah is a black box. It can't. I put this argument to Regina and she said she finds it incredibly frustrating. I definitely hear a lot this question about understanding AI and removing the black box component. And I firmly believe that it shouldn't be part of the conversation.
36:55She says we're asking a machine to find clues we've never been able to see. things that might be literally invisible to the human eye. Of course we're not going to understand what it's saying, she says. That's the point. So what is more important for you, to understand something and use a significantly inferior system or not understand it but really have high predictive capacity with some, you know, brakes and guardrails in place to ensure that when it goes off you know about it? The general discussion, if you look at at least American newspapers, there is so much discussion about the dangers of AI and what it can do and how biased it is and it makes mistakes.
37:41But we are sometimes forgetting that it's not that this technology has to be perfect. It's about is it better than the current system? And the current system is extremely problematic. Despite all these hurdles, Mirai is starting to find its way into some hospitals. It's being used more regularly in hospitals in Brazil and in Saudi Arabia. And there are ongoing clinical trials in hospitals in the US. Regina told me this was all great. But what she mostly thinks about is all of the cancer out there in the world that hospitals are not seeing. All the women being told wrongly. Your scan is all clear.
38:25yeah so during the pandemic uh one of my neighbors um uh who of course didn't want to go to the hospital like none of us wanted during the pandemic was kind of saying that she missed a number of mammograms and she um you know she doesn't want to go for her screening and i suggested to send me her mammogram and to so that we can assess it and um we ran our mirai model on that mammogram and it came out that she is in a very high risk so i suggested that she may want to go and ask for screening and she actually went and got screened and they found like very early cancer. And what was good that, you know, she was able to go through the treatment because it was so early.
39:17It was relatively non-invasive treatment. She went through the treatment. And as far as I know, she's doing perfectly fine now. Regina, you tell this story very breezily, but it seems to me that Mirai could have saved this woman's life. I mean, you know, it's hard to talk about it in hepatetics, but clearly what we do know that the intervention in the early stages of cancer is much smaller and not as invasive as when you discover something, you know, in the latest stages. And, you know, again, looking at my own mammograms, Mirai would detect me to be a breast cancer patient in my very first mammogram when I was 40.
39:54And, you know, this is an open question. Would I had to have chemotherapy and lose my hair, which was the most dramatic part of the treatment for me, you know, I don't know. We would never know. And this is the point that one thing is like saving life, which is extremely important. But there is another thing that, you know, all of these treatments, they have a profound impact on your health. And the more advanced is the disease, the bigger is the intervention. So if we can just decrease the number of people that need to go through chemotherapy, that need radiation, we can go alien, they don't have to go through mastectomy.
40:40This would be really a huge win.
40:46AI offers us this glimpse of a better world. The promise of replacing these old, flawed systems run by humans with systems run by machines that never sleep, never stop working, never make mistakes. In the American city of Detroit, AI promised to revolutionise the way police fought crime, taking responsibilities out of the hands of officers who could be biased, even racist, and giving them to machines that run on cold, hard calculations. On maths. And maths can never be racist. Can it?
41:31for your warrant, for your outstanding warrant. And Detroit is where we're going next. Black Box is produced by Alex Atak. The executive producer is Josh Kelly. The commissioning editor is Nicole Jackson. Production support in Montana was by Kayla DeRoche and in Boston by Nick Kurtisch. Original music and sound design by Rudy Zagadlo. The music supervisor is Max Sanderson and special thanks to Caitlin Kenny.
42:19This is The Guardian.




