Rerelease: Should We Be Worries About Artificial Intelligence? With Torunn Jegleim S1 | E8

This episode was first released in June 2021. 

India and Torunn discuss Artificial Intelligence and the effect it's having in today's world.  During the episode, they consider themes of: Technological bias; The data gender and race gaps; Whether we should be worried about technology now and in the future;  How we can better understand the system, and How we can change it.

Guest: Torunn Jegleim 

Logo Art: Catrin Harrison 

Music by: Aleks Filipiak

Transcript

SPEAKER_00 0:07

Welcome to another episode of Blurring Boundaries, the podcast where we talk about anything and everything avoided or controversial in society today. And this time is the very first time on the podcast that we have a returning member. Welcome, Torrin. Thank you so much. I'm very honoured. Yeah, you should be actually.

SPEAKER_03 0:26

Was it audience demand?

SPEAKER_00 0:28

Um, well, I wouldn't go quite that far, but you know, host demand. Yeah. Well, thank you, thank you. Everyone loved your episode. I think. I think they did. The stats don't lie. Exactly. So we decided to get you back, you know, under protest. And um I can feel the sarcasm coming out just like the last one. Sorry, we don't miss each other. Yeah, no, I'm happy to be back. I'm really glad you're happy to be back. And today we're going to be talking about AI. Something that I feel like you've talked to me a lot about, and something is really interesting. I think actually you've probably informed me the most about AI and technology and all that kind of thing. So thank you very much. And it's a really interesting thing when you suggested doing this, I think it's really interesting because I think a lot of people don't really know as much about it unless um you've researched into it or watched doc documentaries, which I don't think is common at the moment. I think it's becoming more common. I think lots of people still don't really understand the depth of it, if you know what I mean. And do you kind of want do you want to explain like why you know so much about AI?

SPEAKER_03 1:34

Oh, well, um I I wouldn't say I know that much about AI to start there. But yeah, I think um I have always found um technology in general interesting, and I'm generally interested in anything that happens in the world that is going to be or is already a big disruption. So anything that kind of changes how we how we do things, how we live our lives, how systems work, um, and nowadays uh technology is you know uh the cause of many of those types of changes. And I think AI at the moment is a very yeah, it's very topical, it's very um, you know, present. It's it's something that is increasingly a part of our lives, but something that is still not very well understood, um, both from the general public as well as to some extent the people developing it, as I'm sure we'll we'll get more into. And um yeah, I I study politics with quantitative research methods at uh at uni, and I really feel like these discussions about AI is literally those two combined, where there is the societal aspect of it and how yeah, how it is going to influence how we live. But AI and artificial intelligence is basically just stats. It is just statistics. Um so so for me, understanding more and more about statistics and understanding more and more about well, becoming better at thinking about the world around me um allowed me to think about AI in a way that I hadn't been able to before I became aware of certain things through uni. And then I had a few um units at uni where I was able to uh to dig more deeper into AI and write some some assignments on it. And yeah, I just find it really interesting because it is here to stay. I think I think it is going to be uh increasingly present in all of our lives and potentially very influential in good and bad ways, and I think it's really, really interesting for that reason.

SPEAKER_00 3:54

Yeah. So I kind of wanted to start actually, I think that's a really good point to start on. That when we talk about AI, I don't actually think I've just been saying AI from the start of this, but obviously, like you said, AI means artificial intelligence. I think when people think of that, they think of like robots. Um, like what's that that series called Humans, where they they're like robots that genuinely start to have a consciousness or whatever, and and they start like feeling oh Westworld, that one too, but I haven't seen it, Westworld, don't kill me. I know you love it. Jesus Christ. Um but there's there's one called humans as well. I don't know if it's similar. I thought you just describing humans. You know what humans are? Um no, there's a series called Humans, and it's basically that where the like the robots look like people so much, and then at one point they become they gain consciousness and they are basically humans. It's just that they don't um they were they obviously don't have blood and all that kind of thing. But um yeah, I just kind of wanted to start out from the start, like to explain what AI is and what kind of like give some examples of what AI are apart from like robots that can talk, you know what I mean?

SPEAKER_03 5:04

Yeah, yeah, no, I I think I definitely think uh that's a good idea. And I think going off from that, um a nice place to start is between AI and what we call generalized AI, which is what people typically I think think of when they hear AI, and that's what is kind of like say Westworld or humans, where you have a robot that kind of looks like a person and they can do everything that a person can do, uh, and whatever, and they can just kind of yeah, they've just gained some level of intelligence that allows the machine to just kind of act as a as a person or have human-like intelligence. That's that's kind of what we think about. Um but typically AI at the moment is specialized, it's not uh generalized, meaning it's made to do a very specific task.

SPEAKER_01 5:58

Okay.

SPEAKER_03 5:59

So for instance, um IBM made the the chess playing uh machine, and there was all this stuff about how it beat um the the reigning world champion or whatever, um, but that machine can't do anything else. It can play chess and it can play chess really well, but that machine can't then go around and do other things. It's not generally intelligent for the lack of a better phrase. Yeah. But it's it's trained to do something very specific and it does that very well. And those are the kind of things that we are surrounded by today. We're surrounded by machines that can perform very specific tasks very well. And yeah, just to take a step back, AI, meaning artificial intelligence, there is no agreed upon definition. There is no kind of this is definitely what it is, but it generally refers to a machine having the ability to do something um similar to what a human can do. So it relies on algorithms, which are basically a recipe or a code telling telling the machine to do this and then to do that and then to do that. And with the emergence of big data, so having just tons and tons and tons of data available from all kinds of sources, we have so much data and we have the computing power to allow these algorithms to very, very quickly process all this data. And through that we get what we call machine learning, which is when the machine uses those algorithms and it uses all the data that currently exists, processes all of that, and kind of learns from that. So you have an outcome that you want the machine to get to. So say you want to predict you want to predict what kind of products I'm gonna buy, so that you can decide what kind of adverts you're gonna show me online. Um, you would give the machine loads and loads of data about me, but about anyone, about any person in particular, and then once it had data on me, it could then predict, okay, what is this person likely to like, what is this person going to buy. So AI in general is just statistics with a shit ton of data and really, really, really fast computers. And yeah, it's kind of statistics meets computer science.

SPEAKER_00 8:35

I think it's really interesting because you mentioned um then about how they can learn by themselves, and my understanding after kind of like reading or and seeing lots of things about this is that this is kind of where some certain problems arise because these computers are programmed to learn, like you know, however they're programmed to learn, it gets to a point where they've taught themselves so much that people the people who've created them no longer know like what they're actually doing and how they're doing it, because they uh they've kind of gone so far into like teaching itself that nobody knows how it works anymore. And that's I mean, I don't know about you, but that makes me like quite scared because it's like, well, if it can teach itself that, like what the fuck is it gonna teach itself next? And like, you know, like yeah, it's just yeah, I don't know if you know what I mean, but I think that's one of the scariest things that I've heard that nobody knows what it's doing, even the people like at the top of I don't want to say Apple, but it's probably not IBM or you know what I mean, the big companies. I know Apple's it's not Apple, but yeah. Um, you know, the biggest people in tech can't even understand what they've created. You know, it's like when in those horror stories, you know, when they like create a monster and then the monster like goes off and kills loads of people. You know, that's not the same thing. But you know, I feel like we're leading to stand kind of thing. Yeah, exactly.

SPEAKER_03 10:02

Yeah. Um, so I think yeah, that is that is absolutely like a big factor and it comes into so many discussions. Um, I think it might be helpful to think about for a second what uh in general terms AI can be used for.

SPEAKER_01 10:18

Yeah.

SPEAKER_03 10:19

Um so if you give a machine loads and loads of data about something, about people, or for example, image recognition is potentially a good example, where say you have a machine and you want it to be able to identify a dog, you just give it loads of images of dogs, and then some which aren't, and it looks at those and it tries to establish what a typical dog looks like. It tries to establish what kind of features it might have, what would generally uh characterize a dog and make it different from you know other animals and whatever. So AI can learn what is kind of typical, and if it knows what's typical, it can then also predict, like we talked about a second ago, if it knows who I am, then it can also predict things about me, what I'm gonna buy, what I'm gonna do, stuff like that. So it's very, very often used in prediction because it is able to identify what the kind of the normal is. Um but there is also something that is referred to as anomaly detection, which is kind of the opposite. So instead of trying to figure out what the normal is, we care about the exceptions. So for instance, um you might have a bank with thousands and thousands of financial transactions going on in the bank, and the machine knows what a normal transaction is, which means it can then identify what a you know a transaction that doesn't look normal is like, and to do it can use that to then identify fraud, to identify criminal activities, it can be like, hey, this is a pattern in the financial transactions that aren't that aren't deviating from the norm. The same can be done with um, say credit card spending, or if they're trying to identify a person that is acting strange, they know what the norm is. And by they I mean like the machine or the institution. They have data on what the normal is, and then they can identify like outliers in statistical terms. And that's also how they use it often in healthcare. If they know what a normal level of blood sugar is or whatever, they can then identify abnormal levels of something. They can identify cancer, identify, you know, dangerous levels of any type of substance in your body and use that to make advances in healthcare. So AI is such a broad topic, and it can be used in so many ways, but often it is the problems stem from wrongfully determining what the kind of the normal is and how some people then don't fall into the normal and hence they're not picked up when you want them to be picked up, or they are classed as an outlier or an anomaly or something that is weird when they shouldn't be.

SPEAKER_00 13:24

I think that's a really great point. I think you're probably going in that direction anyway, but to talk about like bias in tech, because this is one of the things I think is just so interesting, and I think probably what got us on to talking about AI and and tech in general is that people, I think in general, like the idea with these with robots or with computers or whatever, because they're not a person, because they're made of whatever metal or plastic and wires and stuff, people think, right, okay, well, it's not a person, so it can't have bias, it can't be like prejudice against people because it doesn't have a mind, it doesn't have emotions, it doesn't think like that. But obviously, not obviously to everyone, but like if something is created, it's created by somebody. And the coding or the I suppose if you're writing something within the code and you're trying to determine something at the end, it's still the way it's written is still inherently related to the society that it's written in, or the person who's written it, or you know, the culture that it's being written in, or whatever like that, things like that. So actually, and what I think everyone's starting to realize, or it's coming into the kind of like the general consciousness now, is that AI isn't unbiased, and in fact, it is very biased, because the people who are creating it are biased, and because the society we live in in general is biased. And I think talking about like I think is a really good one to like, you know, you were saying about facial recognition and stuff like that, how it's inherently racist and I think sexist as well, right? Because of the data it's being fed.

SPEAKER_03 15:07

Yeah, I think uh this whole discussion around bias is so important, and I think it is one that people are increasingly becoming aware of um as we're becoming more educated on the subject of AI. And it is, as you say, the the tech industry is um unsurprisingly, or as we all know, not 100% representative of the general population, of course, uh is is largely white, is largely male. And that has implications. Um, mainly I would say because of what you mentioned at the end with the kind of data that is being fed. Um, as my uh professor said, it's it's very much a matter of shit in and shit out. So if if the data you put in is unrepresentative, then you are going to have unrepresentative results. And a good example of that, I think, is is detecting skin cancer in uh in people, where it's becoming really good at it. And I think it's one of the, you know, these are one of the areas of AI that are really interesting and that we kind of want to see more of. But it's been found that they are much better at detecting skin cancer on white skin because the people that have participated um in the studies, the people they have data from are white, which is probably not a conscious decision, but it's just the fact that no one clocked that our entire sample is white or too much of it is white. So it's not good at detecting um yeah, darker skin. That's to say that when the data we're feeding the algorithm is unrepresentative, then the outcome is going to be biased. It can also be that the data we're feeding the algorithm is very much reflecting the actual world, but it's also reflecting the biases and stereotypes that are already present in the world.

SPEAKER_01 17:11

Yeah.

SPEAKER_03 17:12

And I think um we've talked about this before, the example with Amazon and Google also had this problem where they used an algorithm to decide which people they were gonna hire and to teach the algorithm what kind of people they wanted, they gave them the profiles of the people currently working there. So unsurprisingly, the algorithm then learned that white men is uh those are the people that we want to uh employ. And and obviously when that was discovered, it was crapped and it was a big uh yeah, it was not a great moment uh for the leading kind of tech houses. Um in Google's case, women were given less uh or lower paid jobs. So yeah, there were some examples where we kind of like bake the biases currently existing into the machinery, into the into the algorithms, and then we get the results that would naturally kind of stem from that. And those kind of biases we see all the time in skin cancer, in facial recognition, where they are typically worse at predicting uh or identifying people of darker skin. Uh hiring processes are another big example. Um AI is also, for instance, used to predict which um people are going to re-offend if they've committed a crime, which uh typically is a huge disadvantage to black men who are then stereotypically um, you know, incarcerated at a higher rate than than white people, they serve longer sentences, they might face, you know, they do face so many injustices on the way there and then on the way out, it's also predicted that they are the ones that are the most likely to re-offend, and then they get kind of the tournaments and conditions thereafter. So I really think it is it is about being conscious of what kind of data that we are feeding it. And often that's automatic. Often those data processes, um, those data collection processes are automatic. So it's drawing data from users on Facebook, or it's drawing data from kind of automatically generated processes where we don't have much insight, where the developers don't have much insight, and they don't see the data going in. Um and I think that is something that we really need to be aware of. And I think there are so many places to begin, kind of thing. There are so many things to do. One is to be aware of what kind of data we're giving it, like in the skincare example, or being aware that, okay, if we're giving it old profiles of people currently working, let's have a look at how our workforce is currently looking and have a think about how might how might that affect the results. And then there is having the diversity within the tech sector. There is teaching people about AI enough so that people can participate in these debates and question it and and criticize.

SPEAKER_00 20:37

Yeah, I think that's the the question that like springs to mind whenever I hear things about this is like, why don't we just use like or create new data? Like I know that's not as easy as it sounds, but to me, if you're using like you know, um like you said, the profiles of people in Amazon, obviously it's not gonna be like it's it's got the historic problems that are in the society and the culture, like at the moment, or for instance, with I think I don't know if it was you who told me, I think I'm pro I'm sure it probably was, that like obviously with facial recognition or just recognition in general, like um self-driving cars, I think, were better at recognising white people than they were at recognizing black people. And to me, it's just like just be like you said, aware of your bias and then create new like data, if you know what I mean. If that just means, you know, getting a load of paying a load of people to take photos of their faces or their bodies or whatever to feed that into the machine to make sure you've got a diverse group of people, well, there we are. But I feel like, like you said, part of the reason is people not under not even seeing it. Like it's kind of like privilege. Well, I mean it is a privilege in itself, but that you don't see it unless someone points it out. And also, I can imagine it's surely got something to do with economics because. Like obviously, if you're gonna go and do that for every single project you start, it's gonna cost a lot of money. Whereas if you're just taking the data that you've already got, obviously that's free. But um, interestingly, I think I don't know if it was in that documentary that he sent to me. Was it um uncoded coded bias? Uncoded bias, coded bias, coded bias. When uh one of the women who was talking in it was saying, I think it was the main woman in it who was saying like she contacted all the main, I don't know if it was the big nine or something like that, about the fact that like white men were like 99% likely to be recognised by facial recognition, and women were like, I don't know, less, white women were less, and then black women were less, and obviously like black men were the like least or something like that. Don't quote me on that, but you know what I mean. But she she sent this information to them, and they actually changed when they when they saw it, they changed they put inserted more data into it so that they their technology did work better. But I feel like that to me it seemed like the an unlikely event, like an unlikely case. Whereas she'd clearly sent it to all of these and only like one or two of them changed how they worked, because I feel like it would be very, you know, costly in terms of finances to change the system, which is obviously like a problem with a lot of a lot of companies in general, really.

SPEAKER_03 23:22

Yeah, and I think I uh there are so many questions here, and I I there are so many answers that I wish someone would give me, especially as you say, in terms of how come the data that we're feeding it is so unrepresentative. And in some cases, it just seems so obvious. And I guess perhaps part of the explanation is if you have a workforce that is, for example, made up entirely of white men, they you know, not enough people will be conscious that okay, we are a very, very um homogenous group of people. Yeah, maybe we should get some other people in. We're just using the people we already have, um, and whatever. Um I think the case of the hiring algorithm where it's like, okay, uh let's let's use the profiles we currently have to decide what a good candidate is. I guess the problem with that is that you still need to kind of uh define what a good candidate is. The machine itself doesn't know when it gets a C V what a good candidate looks like. So it needs to learn that from something. So you need to give it some kind of data. So you need to pre-define what kind of things it should look for. You need to give it some examples of people, you need to give it, say, you know, here are a hundred CVs that we would have picked, try to find similar ones. And then obviously, even if that's the way you go, someone is defining what a good um CV is going to look like. So the person or the people making it are making those kind of decisions still. But I think the the the point of women being unrepresented in data used for scientific progress hasn't started with AI. And I think I haven't read too much into this, and I yeah, don't quote me on this. But I know there are quite a few debates on how, for example, male animals are used in in studies because they just don't want to deal with the the variation in in their kind of biology or whatever uh when they're doing those kind of studies, or I think it was um this thing about seat belts being tested and men. Yeah, yeah. Where they don't have to be able to do that. And cars and steering wheels. Yeah. Which is just the same thing, where just they have probably just tested with the people they have working on it, yeah. Without thinking, actually maybe this is going to be different for certain kind of people. And I imagine it's the same kind of the same kind of thing happening, which is why it's so important that we have people taking part in these processes, whether it is as an audience, as an active audience, or it is as developers and scientists.

SPEAKER_00 26:20

Yeah, I hadn't even thought of it like that before. I I had heard of the um gender data gap, which in itself is like so problematic, and it's just like, wow. And I'm assuming probably, I mean, I don't know this exactly, but I'm assuming there's probably a race data gap as well, probably in a similar way. But just thinking about like um a while back, there was a big thing about how women just generally, like you were saying about male animals being used, same with humans, like a lot of medicine isn't tested on women because it could affect their fertility, and therefore, like obviously, we've got different hormones in our bodies and things like that. It affects women a lot differently than men in in a lot of different ways, but they're just not being tested, and therefore, medicine often isn't as effective on women or it has different side effects or blah blah blah. Um, I mean, I know we're not talking about that really, but just thinking about in general how how the gender gap, whether that's a data gap, pay gap, you know, information gap, or you know, scientific gap, is just crazy. 100%. It actually makes me think. I just keep thinking, I I keep thinking about this, like, you know, you're saying about adverts and stuff, and how when I first a few about a year ago, I just kept getting bombarded with um adverts for pregnancy tests. And I just kept thinking, like, what part of the AI has like, or whatever, this technology is is like taking me as a person and saying, like, she needs a pregnancy test. Like, what part of them, or not them, but the tech is doing that? And I wonder, and again, I thought this is hilarious. Recently I've been bombarded with like um not actually super recently, maybe a couple of months ago, there was a period where all I was um highlighted was lesbian dating apps. I just thought it was so funny because I thought, like, I don't know what it is, is it the fact that I have like quite a lot of left-leaning things on my um, you know, on my foot, on my on Instagram, or is it like you have too many gay friends? I have quite a few gay friends, but how would they know that? I mean, I suppose on Facebook maybe or Instagram, I don't know. I suppose I I um I follow quite a lot of um kind of like feminist things, but like to me, that's such like a if I'm like looking at it, like that's such like a white male's perspective, like, oh, she's a feminist, she's left leaning, she must be lesbian, like and it kind of I don't know if that's anything to do with it, but it it would make sense in terms of the gender gap, because I could imagine that maybe if you were a man in a similar situation, you might get a lot more targeted advertising because they've kind of like you know profiled you in a much deeper way and have more information on you, which I'm happy for them not to have information on me, but it's just funny.

SPEAKER_03 29:15

I think yeah, it's as you kind of mentioned in the beginning, part of the difficulty here is that we don't really know how the AI makes a decision, which is um let me see if I can think of a good way to explain. If I want to predict uh what kind of um people are going to buy my I don't know, frying pan. And this is what I'm gonna promote. And I have uh I have loads of data, but I only have their gender and their age, and I'm telling my model when I'm doing kind of a normal statistical model, use these two uh variables to try and predict who are gonna buy my frying pan based on who's bought it in the past. That's kind of how statistics works without the AI. Imagine instead you have just loads of data and some of them bought a frying pan and some of them didn't. But you're not telling the machine which variables to look at. You're not telling it to look at gender or age or anything else. You're just giving it all the data and you're saying, you know, predict for me which people are more likely to buy this product. And this is where the algorithms just go on an iterative process of testing everything. So they're testing loads and loads and loads of correlations to see if anything can predict who is going to buy this product. Is it the case that, you know, does gender have an impact, age, but anything else as well that isn't predefined? And then it might be like, okay, I tried this now. It makes me, well, as if the AI can like think it can't. But it kind of goes on this process of being like, okay, this works, let me do more of that. This didn't work, less of that. And that's how it's kind of learning. So it's learning to reprogram itself to get better and better predictions. And once it's going on and on and on doing that, the developers don't have insight into what it decided to do in terms of what data it decided to look at, what correlations it found, how it's weighting those, how it's using those to create predictions. So we just gave it loads of data and we said, this is what we want to know. And then we don't know what's going on in between.

SPEAKER_01 31:40

Yeah.

SPEAKER_03 31:40

And if what we're trying to do is predict who are going to buy my saucepan, it doesn't matter that much. It's not that it's not that important if it doesn't, you know, if you don't get it right. Yeah. But say you're trying to predict if someone is going to be a member of a terrorist organization or whatever, and that isn't very accurate, or you are trying to predict who is going to fault on a loan and you don't want to give them a credit card. Those kind of situations. If you don't know how that decision was made, if you can't tell the person who was rejected from getting a credit card why, you just know that they got a very high likelihood of failing to repay their loan, then that becomes a really serious issue for for everyone, for society, for democracy, for anything that has to do with with justice when we can't really pinpoint how a decision was made. And so I think we should be very, very careful about using and relying solely on AI for those kind of decisions.

SPEAKER_00 32:50

Well, that's really interesting. It's made me think about that documentary again that you mentioned, Coded Bias, and how they are now in the US, they have been using AI to decide who are good teachers and who are not. And there was a teacher who was basically fired, or or thousands of teachers who've been fired. And one case study was this guy who'd like got exceedingly had exceeded in every single test he'd done for the past 20 years or whatever, and suddenly the AI came along and he was fired. And when he said to the state, like, Why have I been fired? Can you please give me the evidence like why I'm a bad teacher? Because I have literally put my heart and soul into all of this work for like 20 years, and they couldn't tell him. And then he obviously found out that the AI was doing it, and the A they couldn't tell him because they didn't know. They didn't know how the AI had reached that conclusion. And it's like, wow, if we're getting to the point where the government is using it for things like hiring teachers or whatever, you just kind of despair because it's like, wow, it's not something that you can decide. To me, it's not something that AI can decide. Um yeah, 100%. And actually, that goes quite well on to I wanted to just talk about a bit about CCTV, and I know we're kind of jumping back to facial recognition a bit, but this is another thing that kind of really scares me because my my mum always goes on about like, oh, you know, the UK is like one of the countries with the highest number of cameras in the in the country and blah blah blah, and that could be a real problem in the future, and I've always been like, oh yeah, but to me it makes me feel safer actually, you know, like for instance with Sarah Everard, like they were able to find her really in a really short space of time because they found saw her on CCTV cameras, and in terms of that, I feel safer in a way, but learning a lot more about this kind of like facial recognition and how other countries use it and how it is being used and how it could be used in the future, wow, like that really is like oh my goodness. Because I think at the moment, well, I think I think it's the best part to start off is like, for instance, with your DNA. If you give your DNA, if you give your fingerprint or whatever, you have to give permission for whoever it is taking that to keep that information about you. And if you don't want them to keep it, you say no. Whereas there is nothing to do, my understanding is that there's nothing to do with that with facial recognition, which is basically the same thing, it's a type of profiling. And they can keep your facial, what would you say, facial profile? Yeah, yeah, yeah. And they can keep that on the system and use that in the future for whatever reason. And one example is that where they're doing it on a massive scale, is China.

SPEAKER_03 35:39

And I'm guessing you know more about this than me, but yeah, I think I think for many people China's social credit system is like a dystopian like warning from many people here. They're looking at China and being like, shit, shit, shit, let's not do that. Um, the way it works is that they have everyone's profile. Uh, so they know your face, they know who you are, they have your identity, and then they can through CCTV, there are other types of data gathering. So when you use your uh credit card, for instance, they know what kind of things you're doing or whatever. But they can, for example, see that you are crossing the street on a red light, for example. And they can see that and then be like, hey, that's not good. We're gonna we're gonna reduce your social credit score a little bit. Or if you're doing something nice, you're picking up litter or whatever, like something like that. You get some points. Um, that's generally how it works. So they're trying to um nudge people to do good things. They're trying to incentivize people to be good citizens. And if you have a low social credit rating, you don't get access to certain things, you don't get access to certain uh certain jobs, you don't get to go on certain trains to certain places, you don't get to be on certain dating apps. So they're really trying to like engineer the social kind of makeup of the population. Whereas if you do have a really good credit score or social credit score, then you do get access to those things. You get discounted, I don't know, mortgages, I'm not exactly sure what the exact benefits are. But yeah, so that is that does partially rely on CCTV to see what everyone is doing, to see if anyone is doing something that's out of line. And if you see someone doing something out of line, you get a picture of their face, you know who they are, and then that's kind of that's it done. You've you've now lost your points.

SPEAKER_00 37:40

Yeah. This really reminds me of I think I was telling you about it earlier, that that um Black Mirror episode, which Black Mirror, for people who don't know, is the series on Netflix where they kind of basically predict the future with what technology can do. I mean, it's really scary, but it's actually you look back at old episodes and you're like, oh my god. And there's one where yeah, it's the same system, you have a social score depending on what you do, who you're with, blah blah blah. Um, again, on this documentary, they were talking about this, and um, one of the girls or women they were they were interviewing said something along the lines of, yeah, well, if I meet a guy, um I'd be more likely to consider dating him if he had a higher higher um social score than a guy who didn't have a higher social score because you know it would show that he's a better citizen and he's a better person, all that kind of thing. And I'm just like, oh my god, I can't believe you're I think it judging someone on that.

SPEAKER_03 38:37

I think part of the scary thing is also the highest room it's working really well. Yeah. Like it is, it is, you know, providing very strong incentive for people to behave in certain ways. And having seen various documentaries on Netflix and also Al Jazeera and other newspapers doing um kind of stories about this, speaking to people there. I don't know how representative of a of a crowd they were interviewing, but a lot of people don't view it as that bad. And I think um I know nothing about China or Chinese culture, but I would perhaps guess that here we have a lower tolerance towards state uh involvement in our kind of lives on average. Um, but yeah, I think I think a lot of people don't necessarily mind that much. Um which is yeah, it's I I don't think it's necessarily seen as you know, so illegitimate as we might do.

SPEAKER_00 39:43

I think as well, because it's kind of like using apps in general. In China it seems to be like you genuinely use your face for everything, like you use it to like go onto the train, you use it to buy foods. Instead of like using your credit card, you'll have your face scanned and they'll just use they know your credit card is connected to you and they'll use that. Um, vending machines I saw and things like that. Um, and it reminds me kind of like of like with the apps that we use in the UK, that we kind of accept the terms and conditions like Facebook, Google, and all those things where we're happy for our data to be shared, or even if we're not specifically happy, we kind of go along with it because we want to use the apps. Um, we're actually they have a lot of data on us, and the same thing with you know, China. And I kind of think that maybe we're all being like, oh, what does it matter? What does it matter? And then maybe in like five, ten years' time, everyone's gonna be like, fuck, this really matters, and it's gonna be way too late, and they already have all the data, and they've we've already agreed to it, and like what are you gonna do? And I can really feel that like kind of happening already. It kind of yeah, it's it's quite scary. And one of the things I saw was about like the protests in Hong Kong. I thought it was quite, it was quite like it was really ingenious, like the way that they were um covering their faces and using gas masks and using what was it, lasers, because they knew that they would disrupt the the the C C D fee cameras, and I kind of think like, oh my god, like we're in this world where you can't even protest without like being screened and having a worse social score and blah blah blah blah blah. And even though we don't have it in the UK at the moment, I feel like it really could go that way, and actually it is starting to go that way in the way that police are starting to use facial recognition technology. Um again, I keep quoting this documentary, but I feel like it was so informative. Like one of those they had like a facial recognition thing on the street, but as we've said, generally white men are the ones that get recognised easier, and black men are the ones that get le recognised less easily, and therefore people are getting profiled wrongly. So, in this thing, often people were getting profiled on the road and getting pulled pulled over, and at one point this I think it was a 14-year-old boy was pulled over by the police and interrogated for like 15 minutes. Obviously, that's very traumatic, especially I think of somebody a person of colour who already has to go through all that kind of thing in general in a society that is obviously racist, and then for that to be like another level is just I can't imagine like when or I would I hope it wouldn't happen, but when or if it happens, you could just kind of swip flip a switch and send this technology to every single CCTV camera in the UK, whereas, as I've mentioned, is one of the countries with the highest amount of CCTV cameras everywhere everywhere, and you could have this happening like I was gonna say tenfold, but a lot more than that, a thousandfold, you know, like a millionfold so many times, people getting recognized in the wrong way. And I just feel like I can't, I find it hard to picture the pros and the cons and how the pros could outweigh the cons, and I really don't really know if it's worth it.

SPEAKER_03 42:57

Yeah. I think I think the CCTV and facial recognition technology is a very good case study if we're talking about those things we did talk about in terms of bias and how it's um inaccurate with certain people and it's fueling certain injustices or certain, you know, stereotypes that we we want to get away from. And I think another big big topic that we have now also touched on is the surveillance state, the surveillance society where, as you said, you can't choose not to be seen in the way you can choose not to give your fingerprints or whatever. You haven't agreed, you haven't signed the terms and conditions when you entered the public space that had CCTV in it. So you haven't agreed to those um yeah, those terms and conditions. You read can't really opt out. And I think increasingly that sort of surveillance is also found online where personally I don't think it is reasonable to. Say to people that okay, if you don't want any data to be collected on you, you need to just kind of go off the grid. Like don't have a phone, don't have a don't have a laptop, don't step into like public places where there is CCTV. I I really don't think that can be the solution to avoiding this if someone is saying, actually, I'm not happy with this. I I really don't think just being like, okay, I'm gonna stop taking part in society should be the way to do that. And I really, really think that these debates need to be had more and that we need to be had publicly. And I really think there is an education project to be done, starting with people at an early age, making people informed, not to the level at, you know, we don't all need to be engineers, but we should all know enough to question these things, to be aware of how it's affecting us, to talk to the politicians, to make our elected officials address this. And I think this is being attempted at many levels, at local levels, at state levels. Um, the EU is quite uh outspoken on the on the topic in trying to regulate these massive tech companies and in trying to enforce various rules and regulations. And the regulations and governance aspect of AI, I think, is the main challenge at the moment. Where, as you say, there are so many potential negative consequences, and the most scary ones, I think, are the unintentional consequences that we're just not aware of. And I really think that, you know, we we can definitely question whether it's worth it in terms of having this technology, but say that it's gonna be around, and I think it is going to be around. We need to regulate it and we need to take ownership of it and be active participants in deciding what's gonna be allowed or not. And there are so many challenges to that in terms of regulating. One is what we already talked about, the black box problem where we just were not sure how the AI got to the conclusion it got to. Another thing is the knowledge gap between the people enforcing these rules and the people making the technology, where most, you know, judges don't have, you know, they're not computer scientists uh working in the field. And another thing is definitions where what we consider to be autonomy, what we consider to be AI, changes all the time. 30 years ago, we would have been quite impressed that certain technologies that we take very for granted now and wouldn't really care about. Um, and so there are a lot of practical challenges to regulating the technology. And I think that is yeah, really where we need to focus our our efforts a bit more. I think that's more urgent than advancing the technology at this stage.

SPEAKER_00 47:10

It's really interesting just thinking about like how like in terms of regulating and laws and things like that. I suppose with most industries up until like like you said, 20 or 30 years ago, they kind of move slowly, and if a problem rises, like you change the law and it kind of all moves at a similar pace. Whereas obviously we're at this point where technology has just like, wow, like picked up the pace, it's moved so fast that nobody is able to keep up with it. And I just am like, how can we even regulate something that by the time we've regulated it, because let's be honest, the government and all these kind of things is not fast enough, and in fact it's very slow with all these kind of things. How can we like keep up or even like overtake technology and kind of predict what is going to be a problem in the future when we're not even able to do it now? If you know what I mean. But I suppose that's not a problem that we can solve exactly like you said, talk to your politicians, talk to your or talk or send emails, you know, if it's something that's really important to you and make sure that people know about it. I think that is again one of the most important things. Recently, I've like, you know, with cookies and things like that, if they say we use cookies, either just deny if you can, or just like instead of going yes, you can like be like amend cookies or whatever and only agree to the ones that are absolutely necessary and cancel all the others, um, to try and reduce the amount of data or the amount of whatever information that these kind of companies or whatever it is are gathering on you to try and kind of reduce the impact. And I do wonder actually if that's when going back to the advertising thing, why I get some really random advertising advertisements sometimes go think, where have you got this idea from? Like, really, that's not the kind of thing I would buy.

SPEAKER_03 49:02

But you know, yeah, no, I think it feels like a very open-ended question, where it's like, how are we gonna do this? And as you say, it's moving so fast, it's like, how the fuck are they gonna keep up? But I think there are a couple of um a couple of points. One being, yeah, try for people to try and educate themselves about this and to just talk about it, to ask questions and to put it on the agenda, to put it on the agenda of you know decision makers at any level, be that local authority, national, EU, making it something that we demand to be discussed, to put it on the agenda as something that is increasingly important. Uh and I think a really big um point that I'm very keen to argue whenever I get the chance is that this shouldn't be for the the developers to decide. Um I think sometimes AI is made to sound very mysterious, very mystical, very scary, very dangerous, very like sci-fi looking. And if there is any, you know, when we talk about it sounds like it is so far away from anyone who's not got a PhD in computer science to understand that people are just like disengaging and just feeling like, you know, this is all too much for me to engage with. But it is bringing it down a bit from that, it is just statistics with loads of data, and that is quite, you know, manageable to get on board with. It's quite manageable to to get the the basics right. It's all about having lots of data, drawing correlations, making predictions, and using that to make decisions. And we need to be critical of all of those steps and how those are used and how we rely on those decisions. And I yeah, I I really, really would hope that people increasingly feel like this is a debate that they should and can join.

SPEAKER_00 51:09

Yeah, definitely. Well, I think we've kind of we're running out of time, unfortunately, although we could probably talk forever about this, and we have in the past. But um, was there anything that we didn't talk about that you feel like you'd like to be like, you need to hear this?

SPEAKER_03 51:28

Was there anything Well I think I think we have covered uh covered most of it, but I think another important point to remember is that AI can make predictions and it can kind of show us a correlation, but it can't make ethical judgments. It can't tell us what we should do, or as I read in this article you can't get a should from an is. So we can say that this is, you know, this person has these and these characteristics, they're this and this likely to fail to pay back their credit card loan. That we can choose not to base our decisions 100% on that. We can choose not to make that decision automatic. In deciding, you know, with automated vehicles, what decisions are they going to make, those are decisions that are for us to decide. So we need people from the social sciences, from philosophy, from people studying ethics and morality and any kind of any kind of field to engage in those things. I do fundamentally believe that AI as as any technology is just a tool and it can be used for for great things and for really, you know, bad things, and it can be attempted to be used for like good things and then it can be done badly. So there are there are many options here, but I really do think that is it's something that people can can help kind of direct in the direction that we want it to go. Um and I hope that people won't feel like they are forced to remain passive bystanders to something that is just kind of happening and we're just looking at it being like, what the fuck?

SPEAKER_00 53:12

Well, there goes the shit and there goes the fan, and that's how it is.

SPEAKER_03 53:17

Literally, and and I really, yeah, would want people to remember that it's not simply uh a computer science or like a statistician's job to work with AI. We really need people to to help decide how we're gonna use it. It is just technology, it's just a machine, and when we can decide where it's going to go.

SPEAKER_00 53:40

Well, I think that's a great place to end the episode. Thank you so much, Torin, for your wisdom and your knowledge and for teaching me and for teaching all my listeners, all five of them. I would hope by now there's more than that. Um we're gonna we've gotta start somewhere. We've gotta start somewhere, exactly. You know, it'll be interesting to see the the amount of listeners compared to the I was gonna say the first, but it was the second, in fact, the second episode of the podcast, which was your podcast on dating. Feel free to listen to it, guys, if you haven't listened to it before. And we already have another one planned, and I will drop a little hint. We're going to do an episode on investing. That is the plan, and specifically women in investing. But I think probably that won't be until you know a few months' time, maybe a year, who knows? But just watch out for that one because you haven't seen the last of Tauren. Just prepare. Just prepare yourselves. Count down in the calendar, even though I don't even know what date we've got yet. But anyway, thank you so much. I hope you've enjoyed. Thank you so much for having me. You're welcome. It's great. It's great to be on. And we'll see you soon. See you soon. Once again, we've come to the end of a very great episode. Thank you so much, Toronto for joining us. I hope you all enjoyed and learnt a lot from the episode. As usual, thank you very much to Alex Philippiak for making the awesome intro and outro music that you can hear on this track. And if you're interested to find out more or to hear about our next episodes, don't forget to follow us on Instagram at BlurringBoundaries1. Or if you'd like to get in contact, feel free to send us an email at blurringboundaries1 at gmail.com. Until then, we'll see you soon.