Instant Genius - What’s the deal with algorithms?

Episode Date: July 18, 2018

Algorithms are everywhere. They can make our lives easier, by curating our Twitter feeds and Netflix suggestions. But they can also be bad. They lack empathy and we can become too reliant on their logical abilities, putting ourselves and others at risk. Here we talk to mathematician Hannah Fry, who tells us all about the good, the bad and the downright ugly of the algorithms that surround us.   Hosted on Acast. See acast.com/privacy for more information. Learn more about your ad choices. Visit podcastchoices.com/adchoices

You said this place was steps from the water. We just haven't found the steps yet. How much did we save? Enough. Enough to get lost. Or you could book a stay with Hilton. Welcome to your oceanfront room. Just steps from the water.

The Hilton sale is on now. Book on Hilton.com or the Hilton app and save up to 20% to get the stay you expected. When you want savings, not surprises. It matters where you stay. Hilton, for the stay. When you need to build up your team to handle the growing chaos at work, use Indeed sponsored jobs. It gives your job post the boost it needs to be seen and helps reach people

with the right skills, certifications, and more. Spend less time searching and more time actually interviewing candidates who check all your boxes. Listeners of this shell will get a $75 sponsor job credit at Indeed.com slash podcast. That's Indeed.com slash podcast. Terms and conditions apply. Need a hiring hero? This is a job for Indeed sponsored jobs. Introducing the new best skin ever ultra slim precision concealer from Sephora Collection. It's full coverage with a matte finish and perfect for any look, whether you're building it up for a full glam moment or targeting correction for a more natural vibe. At only $12, it's great for affordable touch-ups on the go.

Get this new must-have concealer at Sephora or at Sephora.com today. This podcast is sponsored by name, audio and focal. Streaming has made music more accessible than ever, but true listening is about more than ease. It's about quality. British audio experts name audio, alongside French acoustic specialist focal, combine handcrafted tradition with cutting-edge innovation

and high-end materials, delivering digital precision with analogue warmth. So you can experience exceptional sound at home. Music just as the artist intended. Visit name audio. to learn more. Ultimately, all of these algorithms are based to some extent or another on the data that we've recorded about the real world, the sort of the snapshot of the real world as it stands at

the moment. And actually often, the real world, as it is at the moment, isn't how we necessarily want the real world to be. You're listening to the Science Focus podcast from the BBC Focus magazine team. With the UK's best-selling sites and technology monthly, available in print and in several digital formats throughout the world. Find out more at sciencefocus.com or look out for us in your app store. Hello and welcome to the Science Focus podcast. I'm Jason Goodyear, the commissioning editor of BBC Focus magazine. Algorithms are everywhere. They can make our lives easier by curating our Twitter

feeds and Netflix suggestions, but they can also be bad. They lack empathy and we can become too reliant on their logical abilities, putting ourselves and others at risk. Mathematician Hannah Fry is written a brilliant new book, Hello World, which takes us on a tour of the good, the bad, and the downright ugly of the algorithms that surround us. Here, she talks to BBC Focus production editor, Alice Litscomb Southau. So, yeah, first of all, if you just want to tell us a little bit about your book? So the idea behind this book is that all around us, computer algorithms are making decisions on our behalf without us necessarily realizing that that's the case. And there are some slightly dark arts going on behind the scene. So Cambridge Analytica is the sort of most

visible example of this. You know, of course, that we found out earlier this year that our data is being harvested and manipulated and then used against us. But Facebook's really not the only place where this happened. So you find these algorithms, you know, in our hospitals, you find them in our courtrooms, in our police stations, in our supermarkets. Basically, everywhere you look in modern life, you have these things running. behind the scenes and silently making decisions for us. So, I mean, for any people that don't know what an algorithm is, can you sort of sum up in a few sentences what actually is an algorithm?

Yeah. One thing I have discovered in writing this book is that people hate the word algorithm with a passion. So I really need to learn to not use it as much. But essentially, I think one of the reasons why people hate the word is because it's sort of a word that doesn't really mean very much. It's this big blanket umbrella term. officially an algorithm is just it's just something that it's it's like a recipe essentially it's a series of logical steps that takes you from from one from a starting point to an end point right and it's you know that's that's that's a that's a word that conveys almost no meaning whatsoever um but really the way that I think people use the word computer algorithm is they mean little bits of computer code little bits of software essentially um computer decisions essentially where

where computers are taking your data, running some kind of clever stuff on it, and then spitting out an answer. So an example of that might be, you know, your Twitter feed, taking all the people that you follow and everything that they're tweeting about and then spitting out a timeline

that you get to see what they're talking about. But another example of this might be, say you have a defendant who goes into a courtroom, it might be taking all of their history of all of the crimes that they've committed of every time they've ever had contact with the law and then spitting out a prediction about whether that person is then going to go on and commit more crimes. So they range from the very trivial to the very, very serious.

And is that why we should care about them because they're making these decisions maybe without any human input or any sort of empathy there, is that way? Yeah, completely. Completely. They're totally incapable of empathy. And I think that there are some situations, a courtroom being a really great example,

where sometimes actually having empathy is really important. I think the other thing is that actually these algorithms or these sort of machines that we've built, they're not perfect. They make mistakes. And, you know, anyone will know that if they've ever tried to talk to Siri. How, you know, how even something that's actually a very good piece of technology can get things wrong.

And I think that when we're asking machines to make these really big decisions, for us, you know, about our future, about who gets certain treatments in hospital, about where police are sent to sort of focus their attention, you know, to combat crime, about who gets sent to jail. I think, you know, we really need to have these conversations more out in the open about whether this is something that we're comfortable with and whether it's a future that we want for ourselves. So do you think there's a danger that we could become too reliance on them?

I think we already are. I think we already are. I mean, I think there's all these, I was looking for stories of where people have sort of, you know, relied on algorithms a bit more than they should have done, relied on computers more than they should have done. And I came across some amazing stories about people like blindly following their GPS over cliffs and stuff. There's one of my favorite stories, this group of Japanese tourists who were on a road trip in Australia. and they decided they want to go to an island just off the coast of Brisbane. So they put it in their sat-nav,

and the sat-nav said that essentially they just drive straight there. And like, okay, maybe, you know, they didn't have access to a map, their tourists, that's fair enough. Maybe they didn't speak English particularly well and didn't notice that the word island was in the destination that they were going to. But you would have thought

that when they actually started driving on water, they decided to overrule the machine. But unfortunately, no. Unfortunately, they ended up sort of, you know, a fair way out in the water before having been rescued and then a few minutes later a ferry sort of sail past them

so it's some very funny photograph but I think that we're seeing this across the board you know that's sort of again a very silly example of where people are trusting what a computer tells them to do over their own insight over what they can see with their own eyes but there are lots of examples of this

so in Idaho quite recently there was a the council essentially the sort of The local council got a new piece of software that was going to help them decide how much funding certain disabled residents of Idaho should get every year. And these are people who need a lot of support and a lot of care. Otherwise, they'd sort of have to be moved out of their homes and go into special institutions and be cared for full time. So these are people who need a lot of support. and this new piece of software that they got,

the residents just didn't understand. I mean, it seemed to be giving out, deciding these numbers of how much money they should get every year, almost completely at random. So some people ended up with more than they had before, and some people ended up with sort of tens of thousands of dollars less, and it didn't seem to make any sense as to why that might be the case. So they filed a class action lawsuit against the group in Idaho who represented this.

They sort of, first they asked, them to explain what was the decision process. And they were told that it was a piece of software that they got in and they couldn't explain the workings and they sort of had to accept what this software was telling them. So eventually after this class action lawsuit revealed what was going on inside this machine. It turns out it wasn't some clever sort of artificial intelligence. It wasn't some genius piece of, you know, amazing insights that was coming up with the cleanest and most logical way to distribute the money. It turns out it was just a really crappy Excel spreadsheets, and the formulas were so messed up that, I mean, essentially it was just

dishing out, dishing out money at random. And I think that, you know, it took a lot of people to have faith in the machine before that kind of situation arose. And I think we're seeing this more and more that, you know, people quite like the idea of signing over responsibility for decision making to a faceless object that just, you know, that just, tells them what to do. I think that's something we need to be a bit careful about. But there's some cases, like you say, we should be a bit more careful. And there's some cases where we need to just trust the algorithms that they're doing the right job? Yeah, I mean, that's the serious flip side to this story, really. I mean, is that actually, if you think humans are making decisions, you'd be

wrong. So there's some amazing studies of judges in courtrooms, human judges in courtrooms, where, I mean, if there is one job where you want a human to be consistent, to be impartial, to be immune to bias, it's in the situation where someone's judging, a judge presiding over someone's future. So, I mean, all hope is lost, really, when it comes to you. So, I mean, they're incredibly inconsistent. So you can come up with sort of fictional cases and you can test judges on how they would decide on what they would do in those fictional cases. you can ask a series of judges and make sure that they're all coming to the same decision. And they never are, right?

Like, there's been studies done like this and almost, I mean, it's shocking. The difference, the range of answers you get from different judges is kind of shocking on the same cases. But I think the most shocking of all is that when you trick these judges by secretly giving them the same case twice, so the same judge seeing the same case twice, just the names have been changed so the judge can't tell that it's the same case. days. The judge doesn't even agree with themselves, right? So it depends on the day that you get the judge, depends on how they're feeling, it depends on their mood. I mean, there's studies that show that if a judge gives out bail too many times in a row, then they'll refuse bail the next time, even if that case is particularly strong. There's studies that show that judges who have

daughters tend to be much stricter in cases that involve crimes against females. Humans can't make good decisions for Toffee. So in some ways, actually, having this kind of impartial, clean, logical assistant is a step forward. But is it right that sometimes algorithms can reflect the biases of the people who built the algorithm? Because there was a thing with Google Translate, wasn't there, recently? Where if you translated it, it said, oh, if it was a doctor, it would be a man. But if it was a nurse or a hairdresser, it would be a woman. Yes, exactly.

Yeah, so that's if you type something in in English and then translate it to something like Turkish, a gender, a language doesn't have any genders. and then translate it back, then, yeah, it switches the genders to sort of what you might expect, having read all literature up until this point in history. Yeah, that's a really good example that the data that we have, ultimately all of these algorithms are based to some extent or another on the data that we've recorded about the real world, the sort of the snapshot of the real world as it stands at the moment. And actually often the real world, as it is at the moment,

isn't how we necessarily want the real world to be. So, you know, if you type in math professor into Google images, then only one in 20 of the top 20 images is when I did it anyway. Only one in 20 of the top images will be a female. And that might be a fair reflection of how many female professors there are in British universities, female math professors there are. But it's not a fair reflection of what we want society to look like. We don't necessarily want our computers to hold up a mirror to us.

us. We want them instead potentially to hold up a picture of what we want the world to look like rather than what the world is at the moment. So while you were researching your book, what facts that you came across surprised you the most? Oh, that's a good question. That is a good question. I think it is about, I just like the idea that there's teeny tiny clues to our future that are hidden in very, very, very small little fragments of data or, you know, fragments of stuff around us. So the best example of this, I think one of my favorite stories is it's something called the Nunn Study. And essentially it's an epidemiologist called David Snowden. And unsurprisingly, he was studying some nuns. So he had, I think it was 678 nuns involved in this study. And he was

looking at their cognitive ability in older age, right? So when they started this study, the nuns were between about 75 years old and I think 101 or so. And every year of their life, he would test them on how sort of sharp they still were. So things like, he'd ask them, how many animals can you name in a minute, that kind of thing, and test them on that sort of stuff. And he would look for how they sort of kept up their mental abilities they got into older age. But they also, incredibly in this study, the nuns donated their brains to the project after they were dead. So they could then look for kind of physical signs of dementia,

as well as the signs that were there in life. So it's sort of signs in life and signs in death. But sort of what's interesting is that it's not straightforward, right? It's not that the people have the biggest sort of physical symptoms in their brains, once you cut them open, were the ones who were worst affected in life. it's not kind of a one-to-one thing. But it turns out that there might be some clues as to your cognitive ability in older age that happened decades before anybody ever develops in dementia.

Because these nuns also, it turns out, they all wrote a little essay when they entered the sisterhood when they were sort of 18, 19, 20 years old. So David Snowden and his team have analyzed the complexity of the language used in those essays. So how long the sentences were, how many ideas were packed into a single sentence, and found a connection between that and the chances of someone developing dementia in older age. Now, there's a lot more to come on this, right? So it's not just a straightforward, oh, well, that's the answer. Look at who writes complicated essays when you're a teenager, and there'll be the people who are

safer in dementia. It's not like that. But what it suggests is that for starters, there's so much more about our bodies that we still don't know and don't understand. But secondly, I think the most powerful thought there is that there might be clues to our future hidden in these seemingly insignificant pieces of data like those essays, tiny, tiny clues as to what our future has in store for us. And I think that's really what we're seeing now in medicine a lot, actually. So cancer diagnosis has always been looking at what state your tumour is in now and trying to make a prediction as to what will happen. to you in the future based on what your situation looks like now.

But the latest papers that are coming out are suggesting that actually maybe the best clues to whether you'll live or not, to whether your tumour will end up being life-threatening or not, aren't in the tumour themselves, but in the surrounding tissue, these teeny tiny clues that are kind of hidden. And I think that's really sort of both my favourite fact and also my favorite, the reason why I'm so optimistic about the future. I think we've got a lot of kind of good stuff and good discovery ahead. Now, that's fascinating about the nuns. You just wouldn't have thought, would you, back when you were sort of 20 or something? No, not at all.

I was tempted to dig out my essays from when I was 20 because I'm pretty sure I didn't use very complicated language. So after you're writing your book, have you started applying any algorithms to your own life at all? Ah, well, have I? That's a good question, actually. I mean, to be honest, I think I've always been like. like that really. I've always been very sort of methodical and logical. Like, for example, when I was moving house, me and my husband were trying to decide where to move to. And we were kind of really struggling between different areas. So I made him sit down and write a series of equations, construct a series of equations to analyse which area would be best. It's actually a utility

function for those who've worked with those in the past. And I think he's seen as sort of as I presented the idea to him who was rolling his eyes. But by the end, he agreed with me that actually it was an extremely efficient method to decide on where you want to live. So I've always done this. I've always been quite, you know, try and use these mathematical ideas in everyday life. Well, I was thinking that. I'd love it if someone would build an algorithm telling me what to cook for dinner every night

because I always just get home and go, I don't know what to do. I know, that's true. And actually, to be honest, you know, the recipe, that any recipe book in some sense that is containing algorithms for how to cook food. So, you know. So with your job, you're an associate professor in the mathematics of cities. Is that right? Yeah, that's correct.

So can you tell us a little bit about that role? What does it involve and what are you studying and researching in that job? Yeah. So it's a lot about looking for, it's about looking at humans through this kind of logical lens, really. So it involves collecting an. awful lot of data and trying to find patterns in human behavior. So there's no real limit to the kind of situations that this can apply to. So I've worked in the past with sort of supermarkets. I've worked with police forces, with governments. You know, you can, you see these ideas in education, a lot in transport as well,

especially in transport. But really, it's about how people will behave, how people will move, how people will spend money, how people will react. And it's looking at them from this scale where you're looking at large groups of people and trying to predict what they'll do in the future. With the book, one thing I thought was quite interesting, you're talking about the China credit scheme, where they're saying, you're using algorithms to rate citizens on how good they are, and that helps them apply for credit cards or loans. And so they're bad citizens, they might not get those credit cards, but if they're a good citizen, they'll get like the extra good one. Do you think that's likely to be rolled out across other countries, or do you think it'll stay in

China and... Oh, God, I hope not. I think it's really, I think it's probably one of the most evil uses of technology that I've ever come across, really. I think it's really, yeah, pretty horrendous. I mean, they're also talking about, you know, sort of restricting travel visas and stuff for people who get low citizens' scores. I think the thing about China is it's always been quite a special case. They've had ID cards for a very long time, something that we have, you know, rejected more than once in this country. I mean, it's a danger, right? It's a danger. But I also think that we're naive if we think that these kind of big, vast databases of everything we've ever done, connected databases as well, of everything we've ever done,

aren't being collected on us already, because there are these companies called data brokers where essentially that's their job, right, is they have files of people who don't know that they're in there and it's, you know, everything from, I mean, some of the stuff in these people are collecting and inferring from your data is really terrifying. So it's things like whether or not your parents got divorced, whether you were younger. I mean, standard things like your declared gender, your age, you know, those kind of things, standard stuff, right? But then it's also things like your true sexuality and your professed sexuality, whether you've had an abortion, you know, whether you've got HIV, you know, whether you're a rate victim, all of these different kind of things. Because inevitably, if any of those things apply to you, you know, you will have made some kind of a search online at some point that relates to that situation, right?

And that data is being harvested and collected. Now, GDPR is supposed to help with that. That's sort of one step in the right direction. whether it will or not and whether it has or not, I mean, I don't know about you, but I feel like I've gone through death by terms and conditions in the last couple of months. And the thing is,

is that we don't know what people are hiding in their terms and conditions. We don't know what we're clicking agree to when we're just like, oh, just get ready if I just want to look at the website. So, yeah, we kind of have to wait and do how this stuff unfolds. But certainly, Sesame Credit is the scariest example of what can happen. And as well, it's like you say, if they're harvesting data and all sorts of things, and especially somewhere like America where you have a medical insurance,

then you think if they find out that maybe you've got HIV or some other illness, then that might restrict whether you can actually be insured or not, surely. Yeah, so there are, thankfully, there are strict rules in terms of, as far as I understand, and I'm not going to expect this, but as far as I understand, there are strict rules about what you can be denied insurance, health insurance, but there are no such rules for life insurance, right? So, and actually if you have, you know, those DNA tests, those kits that you kind of send off and they send you back and tell you, you know, whether you've got the gene for a monobrow

or whatever, those kind of things. So within those, you can also pay for tests, which will tell you whether you have a propensity for breast cancer, for Parkinson's disease, all of these different kind of illnesses. And you can be denied life insurance if your genetic test. test comes back saying that you have those genes. And the only way to ensure yourself against that is to never have the test. So yeah, I mean, I think this is a worrying point in the future. But, you know, we've already seen the NHS denying people, denying smokers, things like knee operations, sending them to the back of the queue. And, you know, do you really, I don't know,

there's a dystopian future that's just within touching distance that I'd rather avoid. Yeah, there's a concern. Because I actually did one of those genetic. tests and I was, oh, three percent Neanderthal and all this, it's all good fun. But equally, I think I clicked the right turns and conditions so it wouldn't get shared. But equally, you do worry and think, oh, if someone does get hold of that or, you know, if someone just to track you down and try something out. Yeah, yeah, completely. And I mean, you know, you can cut your fingerprints off, right?

You can sort of wear a mask, but you can't get rid of your DNA. It's you and it's yours forever and there's no denying it. I'm sure you'll be fine. I'm sure it'll be fine. Let's be optimistic. I'm sure it'll be totally fine. It's going to be fine, yeah. Definitely fine. I was also wondering, when you were in your book, you talked quite a lot about driverless cars. And I got the feeling from it that you maybe weren't the biggest fan of them at the moment, being sold as like a driverless car and all of this, but actually, they're not really. So I think it's very exciting. Don't get me wrong. I love the idea. I mean, I buy into the idea just as much as anyone else does. But I

also think that there's a real, it's like going back to that thing about following Sandnav off a cliff, right? If you tell people that a car is driverless and that you're living in the driverless dream, driverless for me evokes the idea of not needing a driver, almost by definition. And the thing is, is that the actual technology, there's a big gap between that image and what the actual technology itself can do now. So at the moment, the best technology out there is like a fancy cruise control. A cruise control where you can take your hands off the wheel, right? But you absolutely categorically need to still be driving the car, even if you're not physically touching the wheel. And using words like driverless and self-driving and autopilot, I just think create this

image in people's minds where people are already prone to overtrusting machines, right? And I really don't think it helps the language that's being used around it now. Yeah, because you think people sort of think, oh, you can just climb in the back, have a little sleep. Which they do. There's videos of people doing this. And yet, and yet, you know, there are examples of these cars actually killing people when people aren't paying attention. You know, there's people sort of, there was a particularly famous death where someone was watching a Harry Potter video while, you know, as their car careered into a lorry.

You know, another example of a car that killed a pedestrian when the driver was a, you know, was looking down, not looking at the wheel. I mean, this stuff is really worrying. So you co-present a long-running podcast with Adam Rutherford, the curious cases of Rutherford and Fry, which I'm a big fan of, to be honest. It's really good. That's where I'm going today. That's where I'm off to you.

Ah, I'm going to film the next one. Yeah. So you also answer questions from listeners, and I was going to say, what's been your favourite question that you've answered so far? Oh, good question. I like the ones that come in from kids. Those are always my favourite. not just because they're from kids,

but also because they have genuinely the best questions because I think they're just sort of unrestrained by worrying like they're going to sound silly. So you get questions in from adults sometimes and there's a good question in there somewhere, but it's about 14 pages long. I don't have to read through it. Whereas a question from a kid, probably my favourite,

is what's the tiniest dinosaur? What a great question. What a great question. And actually doesn't have an easy answer. And to dig into that answer, you have to go all through, you know, talks of paleontologists. You have to kind of dig through archives. It's just a great question, very clever question, an insightful question. And I also extra love it that it comes from a kit.

That was Hannah Fry talking about the world of algorithms. Her book, Hello World, is out in September. Did you enjoy this podcast? If you liked what you heard, then why not subscribe and leave us a review? You can find us on iTunes, ACAST, Stitcher and many of your favourite podcast apps. This podcast was brought to you by the team behind BBC Focus magazine. In our summer issue, which is on sale now, we dive deep into the science of laziness.

We also talk to some experts about the threat of space war, and we meet two men trying to create an Ice Age Jurassic Park in Siberia, and much, much more. Thank you for listening to the Science Focus podcast from the BBC Focus magazine team. best-selling science and technology monthly, available in print and in several digital formats throughout the world. Find out more at sciencefocus.com or look out for us in your app store. This podcast is sponsored by Name, Audio and Focal. The texture and emotional depth of music can be lost through digital sources or poor signal. Name Audio believes you can have digital

precision with analogue warmth. Alongside French acoustic specialist focal, Name creates high-end audio systems combining innovation with craftsmanship so you can listen to music just as the artist intended discover more at name audio.com ambition comes in all shapes and sizes at first citizens bank we roll with your goals because we're built for what you're building fit for your ambition for citizens bank wireless can feel like a world of traps but not with visible it's one-line wireless with unlimited data and powered by Verizon for $25 a month. Taxes and fees included. Plus, for a limited time, new members pay just $20 a month for one year on the Visible plan.

Using the code Fresh Start. Refresh your wireless with Visible. Tap the banner to switch today. Terms apply. Limited time offer subject to change. See Visible.com for plan features and network management details.