Sarah: Hi, welcome to Perception. Today I'm here with Stephanie Antonian, who is founder of Astora, which is an innovation consultancy. Hi Stephanie, how are you?
Stephanie Antonian: Hi, I'm good, thank you. Thanks so much for having me.
Sarah: It's good to have you. Just before we crack on, just want to tell everyone how we met. So we met at a conference called Evolve in Brighton. Shout out to Evolve. Fantastic day for the Brighton ecosystem. And we met because you were handing out booklets. And I was really intrigued by this because it's not often you see people handing out little paper booklets anymore. it was an essay titled AI is Irrational. So I went away and actually I put it
Stephanie Antonian: Okay.
Sarah: desk and I didn't read it for a couple of weeks, I'll be really honest, but I sort of looked at it and thought I really want to, you know, I'm going to read that and as soon as I read it I thought wow this is going to be a great chance to talk about this on Perception. So thank you for agreeing to come on, I'm looking forward to the conversation and let's get straight in, okay?
Stephanie Antonian: Well, thank you for reading it.
Sarah: Well, I mean, if someone gives me an essay to read and a booklet in the way that you did it, I'm going to read it, But first of all, I just want to get an intro from you and what under First of all, what's a store rare and what brought you to build it and doing what you're doing now?
Stephanie Antonian: Okay. Okay, great. So I set up a store a few years ago, because I spent most of my career in big tech. Although I started with my degree in Bible studies thinking I'd go into the church and then I was like, no, I'm gonna go into consulting and went to Accenture and then joined Google and their business strategy team. And at the time, this was a while ago. So I was like, data privacy is going to be a really big issue went to policy school. AI was just emerging. And so there weren't a lot of people with business tech policy and like divinity studies together. So then I helped DeepMind set up their ethics team, left that and then worked for Google X for a bit. when I was there, you know, obviously life happens as well. And I was just like, this is all just a bit mean. Not the people that I was working with, they were actually absolutely amazing. And I was like, We just, can't really talk about whether something is kind. Like it's become so sophisticated. Like the analysis has become so sophisticated and so complicated. You couldn't just say, well, is this a kind way to treat people? Is this a loving way to treat people? And then I remember at Google X, was like, love is the answer. we're missing love and this is all built on fear. And they're like, that's correct. But obviously that's not going to come from Google. So do your own thing. So then I set up a store. and I do contracts through it and then write about where love is missing really and what it would look like if we were focused on love, which doesn't sound like it's linked to rationality, but it is. And then design products and just, it's really just my own way to be more hopeful and have some joy left really.
Sarah: Wow, mean, it's the first time I've heard of anyone approach it in that way. And I think your background's really interesting in the sense that, you you started off. doing Bible studies and ended up in tech, which is a bit of a juxtaposition. But I think I would love to understand whether there was, was there one sort of experience where you thought that the story that we, as in people who might not work for the big tech companies, are being told about that rational objective AI that doesn't match what you were seeing and the technology that was in front of you.
Stephanie Antonian: Yeah, so actually are when I left deep mind my dad became very sick and so I decided to be his carer. And that was the career break. And that really, that really changed my perspective on everything. And it's, it's so difficult to believe in the AI hype when you're, you know, in an NHS hospital every single day, chasing and project managing and like just for sort of the basics, like the basic things that we know. And so it was It was so abstract and it was really like humbling, like shaming basically that I'd been all like, NHS, you should spend all this money on AI and like AI is the future and it's going to help on healthcare. And then like just a few months later to be on the other side of it and you know, the money's gone to AI. So even though they have the medicine and they know the solution, there isn't enough and there aren't enough nurses and He was just running around like a headless chicken. And that was like... the most shocking thing that I had experienced and like the most shaming thing that I had experienced because it was like, don't want to, I don't want this for my dad. And I don't want to inflict this on anybody else.
Sarah: Yeah, well, it's kind of that perception versus reality piece, isn't it? And I guess if you're sort of in something and you don't have a kind of front row seat to actually what's going on in real life, I guess that is... That's sort of what came up for me. yeah, sorry to hear, but also it sounds like it was, you know, I always say every negative comes positive. So it sounds like there was something that flicked and it's taken you on a better direction for you personally. So that's what it sounds like.
Stephanie Antonian: Yeah, and it's a luxury to be able to be with someone in their final months and there's lots to be grateful for. So it was difficult, but it was a luxury. And it definitely changed the way I viewed the world and tech.
Sarah: Yeah, yeah. And it goes back to the human and love again, those two words stand out for me when you're talking. It's kind of, you know, we need that human element in life and we need love in life. yeah, I mean, that's thanks for sharing that as well. Like it's a very good way to understand how you pivoted.
Stephanie Antonian: Yeah.
Sarah: can you just give a brief overview of what the essay is about? Because I think that's gonna really unlock some of these questions that I'm gonna ask.
Stephanie Antonian: Okay, I'll try to do it justice. But basically the essay is about AI and rationality. And what it talks about is how we imagine AI as classical AI, which was this tech that once existed. And we had this big culmination when the tech beat, when Deepu beat Garry Kasparov, and that was this big moment of tech taking over the world and surpassing humanity. And actually, nobody really cared. And it didn't, it didn't generate the returns that people expected and IBM had been doing well, but they're not at the top anymore. Garry Kasparov sort of became immortal, chess became more popular than ever. And so it didn't have this like doomsday effect that people imagined. And we say AI has come back now, but actually, it's a different type of AI. It's called machine learning. And that's a fundamentally different vision to classical AI. Classical AI was all about rationality and machine learning is all about irrationality. And so because they're both in computer science, the same way, you know, the falsism and atheism are both under religion. We put them together and we merge them together and we don't quite understand the differences and we don't know how, whether we're using one or the other and that's leading to loads of confusion and like a really big mess basically overall.
Sarah: Yeah, yeah, it's the juxtaposition and kind of two different parts coming at it from people's not understanding that it's actually two very, very different things. I think that's something that definitely needs to be educated. When
Stephanie Antonian: Yeah.
Sarah: you see people who are describing AI as rational, first of all, who is talking about AI being rational? And what do you think they think that is happening inside the machine, as it were?
Stephanie Antonian: Yeah, I mean, I think a lot of the big tech companies are describing AI as rational. And I think it's very difficult to differentiate between AI and machine learning because these big tech companies are also the biggest advertising companies in the world. And, you know, it's been made, it's been muddied very, like very, very well and very intentionally. So unless you were working closely on it would be, it's quite difficult to even think about that being a difference, or unless this was your field of expertise. And I think the reason why it's pushing this big rationality piece is marketing really to sell this tech, to get it into all the businesses. And it's really just a sales play. mean, if you were like, this is This is different from classical AI and it's really just about merging patterns and actually the best use case is targeted ads, which conveniently helps our business model. You wouldn't really sell so much, you wouldn't quite get the hype that it's got now.
Sarah: Yeah, yeah, well, I mean, everything's a sales tactic in this world, isn't it?
Stephanie Antonian: if
Sarah: I think we have to, it's good to think like that because otherwise we can get very sidetracked by it. I guess, do you see the parallels between what happened when social media was in its advent then in terms of what happened with the story that they were telling versus what stories sort of the big frontier lads are telling?
Stephanie Antonian: I think what's different on the story is when social media came out, although I was quite young, but not that I don't remember, it was more positive. It was like we're going to connect everybody. You can be friends, you can find out who's single. I was like 17 at the time when I got Facebook. you know, that was really valuable information. Whereas now the narrative is very depressing. It's like, these machines are more rational than us. And so they surpass us. And so we're rubbish. And we can't compete. And it's like, wow, this is all incredibly depressing.
Sarah: Yes, it is pretty depressing and there's lots of elements of it that are depressing. I do think there is some for good and we can maybe talk about that in a bit, but I wanna also understand, so your argument depends on a fairly demanding definition of reason, but I wanna understand what for you sort of separates the calculation, prediction, reasoning and judgment.
Stephanie Antonian: That's, yeah, that's a good question. I mean, the reason why I know about the rules of logic is because when I was an undergrad, you had to take extra courses and I was such a bullshitter and was like, you know, doing divinity and philosophy and thought it was so great. And I signed up to do logic and I didn't go to any of the classes and I just thought, at the end. I'm just gonna, I'll just write about Descartes or whatever, like, it'll be fine, I'll just do the exam, doesn't matter. And I failed because when I got to the exam, it's all equations. And I was like,
Sarah: no.
Stephanie Antonian: what? There wasn't even an essay question. So I was like, shit, then I had to do the whole summer, spend the whole summer studying to go and do the research, which I didn't do well in. But so it always stayed in my mind that you can't bullshit logic and reason, like it's like math. Like the laws of it are really set out and you can't escape it. So I was one of those people that thought you could just say something was rational or something was logical. But I learned the hard way that you can't. And I think that's the lesson we're all going through because the rules of reason and logic, I mean, they've been discussed for thousands of years and they're like really cemented. The basics are really cemented. And they have been how we've come to define intelligence for like refined over millennia. So that's not really something that any of us now can come up with. Maybe we could refine, maybe like there's a philosophy expert who can refine something here or there, but they're predefined. And so it was just going back to that. And I totally forgotten about it until I thought, This all seems a bit mad, to be honest. And then I said, I wonder if it passes that. And he was like, no, we've all made the same mistake. We all thought we could bullshit it.
Sarah: I think that the line that you said, you can't bullshit your way through reasoning. Brilliant. I love that. That's probably going to be the quote of the episode
Stephanie Antonian: I tried!
Sarah: unless you say any more gems like that. But it's good. It helps you understand.
Stephanie Antonian: Yeah.
Sarah: course. I also love the stuff that you just thought you'd just do nothing and turn up. Amazing.
Stephanie Antonian: Yeah, that you could argue it that it's like I think it's funny because the more you learn about logic the more you realize like it's specifically the guardrails to stop people thinking they can just push it stuff.
Sarah: Yeah.
Stephanie Antonian: But it's so natural and so I'm like I don't even I wouldn't even criticize people for even people who work in tech whose job it is to investigate this for thinking that yeah, okay, it's chat GPT add a reasoning button, it must be reasoning and that there's much more to reasoning than all this stuff. And like, I used to think that way too, but no, there is not.
Sarah: No, right, that's so fascinating. When you think of, so you've talked about the difference between classical AI and machine learning and. What do you, like what change do you think when the kind of field shifted from one ambition to the other? And why do you think that matters outside of computer science, right? What do you think that matters for society in general?
Stephanie Antonian: it's a good question because I think with the boom of classical AI and the deep blue moment, what we learned was that intelligence isn't profitable. And the money, it's not that there isn't money in it, but it's that it's in hardware. I think that's what IBM learned that actually we don't have an economy that runs on intelligence. Again, I don't mean that as a criticism, but like I wouldn't drink Diet Coke in the way that I do if I was truly intelligent. Okay, I drink water. I wouldn't I wouldn't do half of the things that I do if I was really intelligent. So like we built a system based on people being driven by emotion and things other than intelligence. So there isn't really money there. And what you see with machine learning is that it's really moving more towards the monkey brain, but also, and there is money there in like, you know, ads and manipulating and getting people to do things that are irrational. There's lots of money there, getting people to vote one way and another. But the other thing is that it requires a huge amount of hardware because you need the storage. It's like, get more data, get more data, get more data. Why? Because the money comes from you renting. the stories, the money's coming from cloud and the data centers, not necessarily from like the front end bit of AI. And so that has been a really big, that's a really big shift for society because we've now unleashed these technologies that are, that thrive at manipulating people. manipulating is like a really heavy word because it's not like it's even intentional in that way, but They understand how to influence behavior and whoever pays the most gets to influence it in the way that they want. And so that's the big crisis really in that this tech understands. We've come to understand that we don't have money in intelligence. We have money. There's so much money to be made in a rationality and in huge hardware. And so that's what's happening now. Everyone's being called to get more data. to use more tech, to need more data center space, in the hopes that they can manipulate their consumers to do what they want. But ultimately, Also, because of the wealth consolidation, unless people have money, it doesn't really work in the end.
Sarah: What's your in an ideal world? What would you like to change? how would you know what change would you like to see?
Stephanie Antonian: Another great question. What change would I like to see? I used to be anxious and scared about all this stuff, but now I'm quite relaxed because systems like this always destroy themselves by overplaying their hand. Like it's never really policy or... it's not policy or corporates and things that make change. Like they didn't bring down the Berlin Wall or give women the vote and bring, you know, end colonialism. It's always just when people decide they're not buying into the story anymore and just go another way that change happens. And so I think what I'd love is more on the individual level for more people to let go of the fear. of this technology and not buy into this damning story and just accept it as a product that can be used for some good things and can be used for some bad things and isn't an almighty basically. So I think I would start it much smaller with just people letting go of the fear of all of this stuff, seeing it for what it is, knowing that they don't have to engage in it in that way, they don't have to give their power away. And I think that is the foundation of all the times that things have really changed in history. But I used to also be like, I got to write a policy. I got to save the world from this. And now I'm just like, I got to save myself. Yeah, but it's on
Sarah: Yeah, I mean, it's hard to save the whole world. That's, know.
Stephanie Antonian: everyone's to do this now. I'm like, especially people who work on climate change, they're like, to do this, save the climate, like, oof. That's a big one. Or, you know, I meet so many people working in this space who are like, we need to come up with these policies, we need to come up with these laws and they are important. But it's like the, the, it's like, I've come to learn it's an inverted ego where you think that like, what you're going to do is going to save people and it's all gone a bit. it's all gone incredibly dramatic. again, I was that person because I was like, when I was on DeepMind's ethics team, I was like, the responsibility is on me. I got to do this. I got to get this right. And yeah, it's all bit silly in hindsight.
Sarah: But I guess that's why you were in that role, right? Like, I think, I don't know if you would go into one of those roles if you didn't want to make the change, right? But then I guess you probably, well, look what's happened in your world. saw that it probably wouldn't. Maybe I'm wrong, I don't know, I'd assume so.
Stephanie Antonian: Well, don't think it's bad. I don't think it's bad to want to help people. But I think what I wasn't willing to look at was my own inconsistencies and limiting beliefs. I wasn't willing to look internally first before helping other people. So it was a bit preachy. Because now I'm like, well, who was I to think that? There was so much in that I had given my power away to this big story of tech and then was like, now let me save everybody from it. Whereas it didn't, it never had to destroy anybody to begin with.
Sarah: Yeah, yeah, yeah, yeah. think you had a very big epiphany, didn't you? Sounds like a lot
Stephanie Antonian: Yeah, I know.
Sarah: of questions, but it's good because you've obviously, you you've got self-awareness in abundance, clearly. a lot of people might not have self-awareness in abundance. I think it's just interesting to me to hear you talking about it. I love self-aware people. It's great, but it's also a skill to be able to sort of, said, take yourself out of it and ask yourself the questions to think about what it meant for you. And I think that's an amazing way of
Stephanie Antonian: Yeah.
Sarah: starting it because if you relate it to anything in life, I used to say you can't be truly happy and give some else happiness if you can't be happy by yourself or you know that's kind one
Stephanie Antonian: Yeah.
Sarah: of the things I it's a similar sort of way of thinking about it but I think I think it's a good thing to be able to do that.
Stephanie Antonian: Yeah, yeah. Yeah, and that's why we're all here, like, how are going to make the world better? And it's like, well, how do we make the people who we sit next to, who's lives bearable? Like, let's start there. Like,
Sarah: Yeah.
Stephanie Antonian: you know, you would imagine that these teams would be the happiest people in the world. Like, you know, to do the job well, you'd have to be full of joy, really, to spread it around, not full of anxiety and panic and fear because that's just what moves forward. Like the nuts won't get shared, but I digress.
Sarah: Yeah, yeah And so it kind of leads me on to the next question, which is, So you're connecting the rise of AI with loneliness, falling life
Stephanie Antonian: Hmm.
Sarah: expectancy, carbon use and war. And while these things do happen alongside AI, how do you distinguish the effects of the technology from sort of business models, political choices and the institutions that are using it?
Stephanie Antonian: well the tech itself is is like neither here nor there like it's got no value or Like moral judgment to it. It can be used for great things it can be used for awful things and it's a bit like It's funny that I did Bible studies because it feels so relevant feels like the most relevant degree you could do for today. It's a bit like God, God can mean anything to people and you can use God as an excuse to start war, but you could also use God as your motivation to feed the hungry and do like wonderful things. So, know, AI now is sort of, it's this abstract term that can do anything and mean anything. And it's really the systems that we put it in that turn it into something really negative. But, you know, even when I say machine learning, it's not rational. It doesn't mean it's not useful. It can still be useful. There's still lots of cases where identifying patterns at a mass scale is really, really valuable. It's just knowing how and when to use it. But when we put it into these systems, it is going to cause catastrophe. we like to blame people like the tech leaders. And we're like, you're all so evil and blah, blah, But it's really not like that. If you create some tech. and you put it into the markets, the incentive structure is such that it will become exploitative, but that's not you, like that's a whole system. And there are problems with our system. We know that. And we know that we are heading towards catastrophe because of the way our system set up. So like we know we're heading towards climate catastrophe. We know we're looking towards a nuclear war or like we know we're looking at huge wealth inequality and so ideas that systems that will identify patterns and then put you on that track aren't particularly helpful because like we don't need what's probable like we need what's possible we need something totally like out of the blue like some miracle basically to come and set us on a new track so It doesn't really matter what tech we build and it doesn't even matter if we come up with the most amazing rational system or like AGI or whatever yet. As soon as it touches the current marketplace, it's exploited. It's it's baptized with oil. Basically it's like it just couldn't, it has no chance to do well.
Sarah: Yeah, well, baptised with oil. You're coming out with some absolute gems today, I love it.
Stephanie Antonian: Yeah, it's like that's why it's tough because people want to blame the tech leaders and they want to be like they're evil or they're whatever. And like they're actually quite nice people.
Sarah: Yeah, yeah, I mean, it's good to hear you say that because... For people who don't work in the industry or don't really spend their day to day learning about what's going on in the industry, which is having massive impacts on XYZ, people, it's like you can attribute this to lots of different things. People will look for the easy answer and the easy answer is, it must be the leaders of the tech companies who are the bad ones.
Stephanie Antonian: Yeah. Yeah.
Sarah: But it's good to hear that your rationale for why it's not necessarily them. And it's actually kind of a
Stephanie Antonian: Yeah.
Sarah: part of a much bigger systematic problem.
Stephanie Antonian: Yeah, like no one's that powerful by themselves. And the problem is, yeah, it's like, in life, you always think you're doing good. Like in the moment, you always think you're the good guy doing good and helping people. And then later you look back and think, shit, I was a bit of a dick actually. Like
Sarah: Ha
Stephanie Antonian: I should have messed that up. you know, and you see loads of people leaving the tech companies, but also people who are still in there who were like, really grappling with those things, because they genuinely do care. But
Sarah: Yeah.
Stephanie Antonian: like, you think you're doing good at the time. And, and then there's all these unintended consequences. And it's overwhelming, or you know, you do think you can change it or what have you. But it's like one mind can't hold like, the whole systems of how the world works. And Yeah, that's what I think is almost like even scarier is that it's not like there's this evil group of people being like, I'm gonna dominate the world. It's like people being like, wow, this is so cool. And I think it could be used for this and this and this. Wouldn't it be great if this and then the rest happens.
Sarah: Yeah, yeah, A story is old of time. But I want to talk about how sort of... Prediction shapes perception. So obviously this podcast is called Perception. And I like to speak to people from different areas around, you know, asking how people, organizations, and increasingly the AI systems decide what matters. So if machine learning systems infer what is probable from existing patterns, what happens when that starts to look like the most repeated or common account? And that's the truth that we get.
Stephanie Antonian: Room.
Sarah: like it seems it kind of blows my mind when we think of it in this way but what are your thoughts on that?
Stephanie Antonian: Well, I mean, I think we're sort of living through that. I mean, you just need to go on LinkedIn to think, what is going on? Yeah, like, can't, I can't live like this. Like, I can't, I can't, I can't read this stuff anymore. And I think it all converges to the same. And, you know, you see it happening with the media and with the news and it's like, everyone's sort of saying the same thing. You don't really get that much diversity because it's not just the tech. that's moving to it, but like the incentive structures are putting everything together. So even for example, like it is mad that I'm the person who wrote the essay on rationality and machine learning failing the principles of logic, like I had never read it before, couldn't find it anywhere. It is absolutely mental that that that I had the opportunity to write that because there's so many philosophers as people whose jobs it is to be philosophers in AI. And there's academics and thought in AI, but it's all funded by big techs. Of course you can't say that. Like, of course you can't say stuff like this. Like this is not what people want to read or want to hear in the intent structure. So the good thing is, is that I think it makes people just turn off. I think we just get so bored that people turn off and they look for more creative things. And so I actually think it's a bit of a good news story because it means new people have chances to say interesting things. And the ideas could come from anywhere. And people are looking more and more for different stuff. like, know, reconnected because I handed you a paper essay. You know, that that that's because like the reason why you read it is because there must be a party that's fed up of all this love and all the bullshit that's being said. So
Sarah: yeah. Yep.
Stephanie Antonian: when it converges, it's really it's, you know, a bit it's scary and it's depressing, but then on the flip side, we just all start looking for something else. We just go the other way, it swings on a pendulum. We're like, okay, I want something interesting. So I'm to have to go to another source and I'm going to have to find more original thinkers and maybe I do like artistic. And so I think it's a wonderful time for creatives now if you can hold your nerve, because it just means people will look for something else. And that's where it on the business side, that's where it's quite difficult because the more you embed these these tools, the more you're aligning to convergence, which means the less you the less competitive advantage you have actually in reality.
Sarah: Yeah, yeah, and you just become reliant on them. you handed me a piece of paper that was in itself interesting to me. And what else? So it's, yeah, I'm bored of the slop. That's absolutely one thing. But I'm also, you know, as a new business owner myself, I have never had to do what I'm having to do. All of the things that you have to do when you're starting a business, you have to be content creator, you have to do PR, have to do,
Stephanie Antonian: Yeah. you
Sarah: you have to sell your services, to everything as you know, but I guess one thing that I've sort of found very overwhelming, it's not just the slop, it's the, you know, sitting behind a screen, which is a very small screen, and the world seems infinite, and it's overwhelming, because actually
Stephanie Antonian: you you
Sarah: everyone's a creator these days, you don't actually know where to look, or it's just a sea of overwhelm, I think, online. So to have something physical, and like I said, I didn't read it straight away,
Stephanie Antonian: Okay.
Sarah: because I was, you I had loads of stuff going on, but I put it on my desk and I looked at it and I went, yeah, that's, I need to remember to read that. So it's having that physical, and it was creative and I really liked the way you thought and you positioned it. So it's kind of the, takes me back to human, I met you in person, we didn't speak for very long, but I remembered you. I had this physical
Stephanie Antonian: Okay. Okay.
Sarah: piece of content on my desk, which no, I've not had that for, I don't remember how long, and it was a different point of view and it was clearly not AI slot. So there were three things for me that made it really powerful. And I really wanna go back to, I just wanna go back to the humans. So I think conferences are very important.
Stephanie Antonian: love the human aspect too and what I'm noticing is just more and more communities building like in in person and more analog stuff and I see more and more people spending less and less time online and that's why I think it's like we don't need to be scared of this tech and anxious this tech because it kills itself. has its own. It has its own ending. Yeah, like, okay, amazing. You can make an app in the five minutes, but like, I can't have a million apps. I'm only interested in tech. So, you know, there's there is it, it's got its own endpoints and it destroys itself by itself. So focusing more on what's more fun and positive like building connections with people, building community, like doing things in real life. that's the way it's going. I mean, that's what I see. I don't even think this is a revolution. I think it's more of a renaissance because the number of communities that are coming together is really incredible.
Sarah: Yeah. I also see that. I've been saying that for quite a while whenever I'm speaking to anyone about it. I'm seeing a real shift back to human and human connection. I think, you know, we had the pandemic, it took a while for I think people to get back to that mentality of getting back to face to face. And
Stephanie Antonian: Yeah.
Sarah: I think people are just so fed up. Personally, I've been sat in a room for years now and on my own and I don't like that anymore. I don't want to do it. I was also wanting to ask you, so as the AI systems increasingly mediate the discovery and comparison and recommendation, what do you think credibility means for a business or a person? What does a machine need to establish before that interpretation can influence a human decision?
Stephanie Antonian: Just that it's correct, that the answer is correct.
Sarah: sound's so simple.
Stephanie Antonian: Yeah, does it give you the answer that you want? Like, because if it does, then it's great. That's fantastic. But like, if it's wrong, like three out of five times, four out of five times, one out of five times, like you need to know that because then you need to adjust that if it if it's correct, it's fantastic.
Sarah: Yeah. The question is, how do we ensure it's correct?
Stephanie Antonian: you just got to test how many times that you just do a test on it. I'm like, it's funny because people are like, we got to explain how the model works. We got to do all these difficult things. It's like, let me tell you, just is it just look at the endpoint and is it useful? Is it not useful? Like, you know, you can search on chat GPT, like ask all the things about your company 20 times and then look at how many times it was correct. Like, you can run small experiments to see how much you can trust it. And it's really interesting because there's so much talk about trust. But really, it's coercion because to be trustworthy, you don't have to talk about trust, you don't have to go and be like, I'm gonna build everybody's
Sarah: You
Stephanie Antonian: trust, you just need to do what you say. Like you just need to be like, like that's, that's it. That's all it takes to build trust. Like, but when you have these huge marketing teams on like trust and safety and, It's really coercion because you basically say I've not built something that's correct, but I want you to trust me anyway. It's like, well, just do it properly. Do what you say, follow through and that's it. People will trust it. I'm not saying
Sarah: Yeah.
Stephanie Antonian: you shouldn't trust a machine. There's lots of things that I trust that I don't understand. When I go to the doctors and I get prescribed medicine, I do not understand how the medicine works or whether... What ingredients in the medicine but I take it, you know because I trust them like or
Sarah: Yeah, yeah.
Stephanie Antonian: the building I'm not like always this up to spec. Is it gonna fall on my head? I'm like, okay, don't understand this stuff Trust that somebody does and that you know, it's fine. So I'm like you it's fine to trust machines as long as that crack
Sarah: got one more question for you before we go. And I just want to know what you think a genuinely responsible relationship between that prediction and judgment would look like. So like where should machines help us see patterns and where do you think humans still need to remain accountable? Like goes back into that thing about trust, doesn't it? Like how much should we trust machines versus not?
Stephanie Antonian: Well, So there's lots of use cases where looking at patterns at scale is really helpful. Like, you know, in medicine, it is helpful. And I'm not saying that AI won't come up with great or machine learning won't come up with new drug discovery or cures. I'm just saying, okay, tell me now once you've done it, I don't need to hear about it every day. I'm happy for you. That's not what I'm going to be working on because I don't have the technical skills for that. But so let me know when you've done it. But there's lots of different, there's loads of use cases where looking at patterns is really helpful. Personally, I don't think judgments about people should be made by machines, even really by people. I'm trying to let that stuff go. But that's where I think it gets really difficult because whenever it starts to become judgments about people, it's like it's only ever used on superficial layers again, because of the way the technology works. So like it, it will look, are you a woman? Are you Middle Eastern? Like, are you educated? Are you wealthy? Like it will look at superficial categories and then make its determination of yeah, you could, you should get alone. Like you're a white man who's rich, you get alone. Like, you know, you're a woman who's foreign, like you do not get a loan. And, that's using historical data, obviously, but the issue is like, it just keeps us entrenched and judging people at these really surface level factors. And that's what I think takes us away from intelligence. So at beginning, I said that love and rationality are actually linked. And what had, you know, before all this like, got me started on it is just We used to be really aligned on the philosophy of the love of mankind. Like we used to, like this, central principle of everything was this unstated, or actually sometimes stated view that to love mankind was intelligence. And that's what a lot of the Greek philosophers also talk about alongside rationality and like all of this only works when you value other people as yourself. And like even Adam Smith of capitalism, he's like, spoke a lot about how the invisible hand is the love of mankind, not necessarily only greed, but he's a Christian theologian as well. He talks a lot about love in his other works. But that's all got gotten, been forgotten, because machine learning cannot treat humans as equals. Because as
Sarah: Yeah.
Stephanie Antonian: soon as you bring judgment in it has to determine who's more worthy than others about other things. And so I think there's lots of amazing use cases for machine learning on identifying patterns. And where the red line should be is on making any judgment about humans. Like if we genuinely want to move to intelligence, enlightenment or something like that, I think we have to have the principles that, you know, all humans are equal. And so it becomes too difficult for it to make the judgments. But the problem with that
Sarah: Wow, yeah.
Stephanie Antonian: is that those are most of the uses that they're selling.
Sarah: Hmm, yep, yeah, well yeah. For now, we'll see, we'll see, because there might be a shift change at some point when people start to realise that,
Stephanie Antonian: Yeah, and I think, for example, like in recruitment, there's so many AI tools for recruitment. And now it's like, you know, AI read your CV, it decides who's going. But the reason why it's not like, it's not very useful is that most people don't really know what makes a good hire. So like, you give it all this date, when you give all the data to an AI system, it's going to look at like, what universities people go to, did they go to university? like where they work before, what are the brand names, like what's the nationality of the name, is it like, it's going to look at all these really superficial things and then it's going to say, yeah, they can have an interview. But none of those things are really what makes somebody a good hire. Like,
Sarah: Yeah. Yeah.
Stephanie Antonian: like often you just say, do you know what? I really like It's like something or somebody tells you a story like. met with this one girl and you know she was doing an evening degree and was a carer for her parents and I was like anything you need from me I will be it like I'm like I can read between the lines understand so much about you for what you're doing but all of that goes so the people aren't the best but because the core issue is we don't actually know what makes good hire like can't put it into words
Sarah: Yeah, yeah, yeah. Yeah. I mean, that's a great example of a use case that everyone will understand because everyone's been through a hiring process and recruitment process. Fascinating.
Stephanie Antonian: Yeah, yeah, which is also why they've just turned nepotism because I'm like, it's just so much stuff is coming to the top that then you're like, okay, well, my friend's child's reached out. So
Sarah: Yeah, yeah, yeah, yeah.
Stephanie Antonian: not good. Yeah, as someone who really depended on meritocracy, like,
Sarah: Yeah, yeah, yeah.
Stephanie Antonian: not good, that is not what you want to see. Yeah.
Sarah: No, I feel like that's a whole other podcast. That's like opening a can of worms. But I am going to wrap it up and I thank you so much I can't believe time has gone that quickly. I've really been engaged and listening to everything you're saying and I think it's fascinating and I'm looking forward to continuing the conversation. But is there anything else you just wanted to share before we wrap up or do you want to?
Stephanie Antonian: I
Sarah: Go back to your day.
Stephanie Antonian: you so much for having me on. was really fun talking to you. Thanks. Bye.
Sarah: Great, all right, well thank you so much and I'll see you soon. Bye.
Stephanie Antonian: Hi, I'm good, thank you. Thanks so much for having me.
Sarah: It's good to have you. Just before we crack on, just want to tell everyone how we met. So we met at a conference called Evolve in Brighton. Shout out to Evolve. Fantastic day for the Brighton ecosystem. And we met because you were handing out booklets. And I was really intrigued by this because it's not often you see people handing out little paper booklets anymore. it was an essay titled AI is Irrational. So I went away and actually I put it
Stephanie Antonian: Okay.
Sarah: desk and I didn't read it for a couple of weeks, I'll be really honest, but I sort of looked at it and thought I really want to, you know, I'm going to read that and as soon as I read it I thought wow this is going to be a great chance to talk about this on Perception. So thank you for agreeing to come on, I'm looking forward to the conversation and let's get straight in, okay?
Stephanie Antonian: Well, thank you for reading it.
Sarah: Well, I mean, if someone gives me an essay to read and a booklet in the way that you did it, I'm going to read it, But first of all, I just want to get an intro from you and what under First of all, what's a store rare and what brought you to build it and doing what you're doing now?
Stephanie Antonian: Okay. Okay, great. So I set up a store a few years ago, because I spent most of my career in big tech. Although I started with my degree in Bible studies thinking I'd go into the church and then I was like, no, I'm gonna go into consulting and went to Accenture and then joined Google and their business strategy team. And at the time, this was a while ago. So I was like, data privacy is going to be a really big issue went to policy school. AI was just emerging. And so there weren't a lot of people with business tech policy and like divinity studies together. So then I helped DeepMind set up their ethics team, left that and then worked for Google X for a bit. when I was there, you know, obviously life happens as well. And I was just like, this is all just a bit mean. Not the people that I was working with, they were actually absolutely amazing. And I was like, We just, can't really talk about whether something is kind. Like it's become so sophisticated. Like the analysis has become so sophisticated and so complicated. You couldn't just say, well, is this a kind way to treat people? Is this a loving way to treat people? And then I remember at Google X, was like, love is the answer. we're missing love and this is all built on fear. And they're like, that's correct. But obviously that's not going to come from Google. So do your own thing. So then I set up a store. and I do contracts through it and then write about where love is missing really and what it would look like if we were focused on love, which doesn't sound like it's linked to rationality, but it is. And then design products and just, it's really just my own way to be more hopeful and have some joy left really.
Sarah: Wow, mean, it's the first time I've heard of anyone approach it in that way. And I think your background's really interesting in the sense that, you you started off. doing Bible studies and ended up in tech, which is a bit of a juxtaposition. But I think I would love to understand whether there was, was there one sort of experience where you thought that the story that we, as in people who might not work for the big tech companies, are being told about that rational objective AI that doesn't match what you were seeing and the technology that was in front of you.
Stephanie Antonian: Yeah, so actually are when I left deep mind my dad became very sick and so I decided to be his carer. And that was the career break. And that really, that really changed my perspective on everything. And it's, it's so difficult to believe in the AI hype when you're, you know, in an NHS hospital every single day, chasing and project managing and like just for sort of the basics, like the basic things that we know. And so it was It was so abstract and it was really like humbling, like shaming basically that I'd been all like, NHS, you should spend all this money on AI and like AI is the future and it's going to help on healthcare. And then like just a few months later to be on the other side of it and you know, the money's gone to AI. So even though they have the medicine and they know the solution, there isn't enough and there aren't enough nurses and He was just running around like a headless chicken. And that was like... the most shocking thing that I had experienced and like the most shaming thing that I had experienced because it was like, don't want to, I don't want this for my dad. And I don't want to inflict this on anybody else.
Sarah: Yeah, well, it's kind of that perception versus reality piece, isn't it? And I guess if you're sort of in something and you don't have a kind of front row seat to actually what's going on in real life, I guess that is... That's sort of what came up for me. yeah, sorry to hear, but also it sounds like it was, you know, I always say every negative comes positive. So it sounds like there was something that flicked and it's taken you on a better direction for you personally. So that's what it sounds like.
Stephanie Antonian: Yeah, and it's a luxury to be able to be with someone in their final months and there's lots to be grateful for. So it was difficult, but it was a luxury. And it definitely changed the way I viewed the world and tech.
Sarah: Yeah, yeah. And it goes back to the human and love again, those two words stand out for me when you're talking. It's kind of, you know, we need that human element in life and we need love in life. yeah, I mean, that's thanks for sharing that as well. Like it's a very good way to understand how you pivoted.
Stephanie Antonian: Yeah.
Sarah: can you just give a brief overview of what the essay is about? Because I think that's gonna really unlock some of these questions that I'm gonna ask.
Stephanie Antonian: Okay, I'll try to do it justice. But basically the essay is about AI and rationality. And what it talks about is how we imagine AI as classical AI, which was this tech that once existed. And we had this big culmination when the tech beat, when Deepu beat Garry Kasparov, and that was this big moment of tech taking over the world and surpassing humanity. And actually, nobody really cared. And it didn't, it didn't generate the returns that people expected and IBM had been doing well, but they're not at the top anymore. Garry Kasparov sort of became immortal, chess became more popular than ever. And so it didn't have this like doomsday effect that people imagined. And we say AI has come back now, but actually, it's a different type of AI. It's called machine learning. And that's a fundamentally different vision to classical AI. Classical AI was all about rationality and machine learning is all about irrationality. And so because they're both in computer science, the same way, you know, the falsism and atheism are both under religion. We put them together and we merge them together and we don't quite understand the differences and we don't know how, whether we're using one or the other and that's leading to loads of confusion and like a really big mess basically overall.
Sarah: Yeah, yeah, it's the juxtaposition and kind of two different parts coming at it from people's not understanding that it's actually two very, very different things. I think that's something that definitely needs to be educated. When
Stephanie Antonian: Yeah.
Sarah: you see people who are describing AI as rational, first of all, who is talking about AI being rational? And what do you think they think that is happening inside the machine, as it were?
Stephanie Antonian: Yeah, I mean, I think a lot of the big tech companies are describing AI as rational. And I think it's very difficult to differentiate between AI and machine learning because these big tech companies are also the biggest advertising companies in the world. And, you know, it's been made, it's been muddied very, like very, very well and very intentionally. So unless you were working closely on it would be, it's quite difficult to even think about that being a difference, or unless this was your field of expertise. And I think the reason why it's pushing this big rationality piece is marketing really to sell this tech, to get it into all the businesses. And it's really just a sales play. mean, if you were like, this is This is different from classical AI and it's really just about merging patterns and actually the best use case is targeted ads, which conveniently helps our business model. You wouldn't really sell so much, you wouldn't quite get the hype that it's got now.
Sarah: Yeah, yeah, well, I mean, everything's a sales tactic in this world, isn't it?
Stephanie Antonian: if
Sarah: I think we have to, it's good to think like that because otherwise we can get very sidetracked by it. I guess, do you see the parallels between what happened when social media was in its advent then in terms of what happened with the story that they were telling versus what stories sort of the big frontier lads are telling?
Stephanie Antonian: I think what's different on the story is when social media came out, although I was quite young, but not that I don't remember, it was more positive. It was like we're going to connect everybody. You can be friends, you can find out who's single. I was like 17 at the time when I got Facebook. you know, that was really valuable information. Whereas now the narrative is very depressing. It's like, these machines are more rational than us. And so they surpass us. And so we're rubbish. And we can't compete. And it's like, wow, this is all incredibly depressing.
Sarah: Yes, it is pretty depressing and there's lots of elements of it that are depressing. I do think there is some for good and we can maybe talk about that in a bit, but I wanna also understand, so your argument depends on a fairly demanding definition of reason, but I wanna understand what for you sort of separates the calculation, prediction, reasoning and judgment.
Stephanie Antonian: That's, yeah, that's a good question. I mean, the reason why I know about the rules of logic is because when I was an undergrad, you had to take extra courses and I was such a bullshitter and was like, you know, doing divinity and philosophy and thought it was so great. And I signed up to do logic and I didn't go to any of the classes and I just thought, at the end. I'm just gonna, I'll just write about Descartes or whatever, like, it'll be fine, I'll just do the exam, doesn't matter. And I failed because when I got to the exam, it's all equations. And I was like,
Sarah: no.
Stephanie Antonian: what? There wasn't even an essay question. So I was like, shit, then I had to do the whole summer, spend the whole summer studying to go and do the research, which I didn't do well in. But so it always stayed in my mind that you can't bullshit logic and reason, like it's like math. Like the laws of it are really set out and you can't escape it. So I was one of those people that thought you could just say something was rational or something was logical. But I learned the hard way that you can't. And I think that's the lesson we're all going through because the rules of reason and logic, I mean, they've been discussed for thousands of years and they're like really cemented. The basics are really cemented. And they have been how we've come to define intelligence for like refined over millennia. So that's not really something that any of us now can come up with. Maybe we could refine, maybe like there's a philosophy expert who can refine something here or there, but they're predefined. And so it was just going back to that. And I totally forgotten about it until I thought, This all seems a bit mad, to be honest. And then I said, I wonder if it passes that. And he was like, no, we've all made the same mistake. We all thought we could bullshit it.
Sarah: I think that the line that you said, you can't bullshit your way through reasoning. Brilliant. I love that. That's probably going to be the quote of the episode
Stephanie Antonian: I tried!
Sarah: unless you say any more gems like that. But it's good. It helps you understand.
Stephanie Antonian: Yeah.
Sarah: course. I also love the stuff that you just thought you'd just do nothing and turn up. Amazing.
Stephanie Antonian: Yeah, that you could argue it that it's like I think it's funny because the more you learn about logic the more you realize like it's specifically the guardrails to stop people thinking they can just push it stuff.
Sarah: Yeah.
Stephanie Antonian: But it's so natural and so I'm like I don't even I wouldn't even criticize people for even people who work in tech whose job it is to investigate this for thinking that yeah, okay, it's chat GPT add a reasoning button, it must be reasoning and that there's much more to reasoning than all this stuff. And like, I used to think that way too, but no, there is not.
Sarah: No, right, that's so fascinating. When you think of, so you've talked about the difference between classical AI and machine learning and. What do you, like what change do you think when the kind of field shifted from one ambition to the other? And why do you think that matters outside of computer science, right? What do you think that matters for society in general?
Stephanie Antonian: it's a good question because I think with the boom of classical AI and the deep blue moment, what we learned was that intelligence isn't profitable. And the money, it's not that there isn't money in it, but it's that it's in hardware. I think that's what IBM learned that actually we don't have an economy that runs on intelligence. Again, I don't mean that as a criticism, but like I wouldn't drink Diet Coke in the way that I do if I was truly intelligent. Okay, I drink water. I wouldn't I wouldn't do half of the things that I do if I was really intelligent. So like we built a system based on people being driven by emotion and things other than intelligence. So there isn't really money there. And what you see with machine learning is that it's really moving more towards the monkey brain, but also, and there is money there in like, you know, ads and manipulating and getting people to do things that are irrational. There's lots of money there, getting people to vote one way and another. But the other thing is that it requires a huge amount of hardware because you need the storage. It's like, get more data, get more data, get more data. Why? Because the money comes from you renting. the stories, the money's coming from cloud and the data centers, not necessarily from like the front end bit of AI. And so that has been a really big, that's a really big shift for society because we've now unleashed these technologies that are, that thrive at manipulating people. manipulating is like a really heavy word because it's not like it's even intentional in that way, but They understand how to influence behavior and whoever pays the most gets to influence it in the way that they want. And so that's the big crisis really in that this tech understands. We've come to understand that we don't have money in intelligence. We have money. There's so much money to be made in a rationality and in huge hardware. And so that's what's happening now. Everyone's being called to get more data. to use more tech, to need more data center space, in the hopes that they can manipulate their consumers to do what they want. But ultimately, Also, because of the wealth consolidation, unless people have money, it doesn't really work in the end.
Sarah: What's your in an ideal world? What would you like to change? how would you know what change would you like to see?
Stephanie Antonian: Another great question. What change would I like to see? I used to be anxious and scared about all this stuff, but now I'm quite relaxed because systems like this always destroy themselves by overplaying their hand. Like it's never really policy or... it's not policy or corporates and things that make change. Like they didn't bring down the Berlin Wall or give women the vote and bring, you know, end colonialism. It's always just when people decide they're not buying into the story anymore and just go another way that change happens. And so I think what I'd love is more on the individual level for more people to let go of the fear. of this technology and not buy into this damning story and just accept it as a product that can be used for some good things and can be used for some bad things and isn't an almighty basically. So I think I would start it much smaller with just people letting go of the fear of all of this stuff, seeing it for what it is, knowing that they don't have to engage in it in that way, they don't have to give their power away. And I think that is the foundation of all the times that things have really changed in history. But I used to also be like, I got to write a policy. I got to save the world from this. And now I'm just like, I got to save myself. Yeah, but it's on
Sarah: Yeah, I mean, it's hard to save the whole world. That's, know.
Stephanie Antonian: everyone's to do this now. I'm like, especially people who work on climate change, they're like, to do this, save the climate, like, oof. That's a big one. Or, you know, I meet so many people working in this space who are like, we need to come up with these policies, we need to come up with these laws and they are important. But it's like the, the, it's like, I've come to learn it's an inverted ego where you think that like, what you're going to do is going to save people and it's all gone a bit. it's all gone incredibly dramatic. again, I was that person because I was like, when I was on DeepMind's ethics team, I was like, the responsibility is on me. I got to do this. I got to get this right. And yeah, it's all bit silly in hindsight.
Sarah: But I guess that's why you were in that role, right? Like, I think, I don't know if you would go into one of those roles if you didn't want to make the change, right? But then I guess you probably, well, look what's happened in your world. saw that it probably wouldn't. Maybe I'm wrong, I don't know, I'd assume so.
Stephanie Antonian: Well, don't think it's bad. I don't think it's bad to want to help people. But I think what I wasn't willing to look at was my own inconsistencies and limiting beliefs. I wasn't willing to look internally first before helping other people. So it was a bit preachy. Because now I'm like, well, who was I to think that? There was so much in that I had given my power away to this big story of tech and then was like, now let me save everybody from it. Whereas it didn't, it never had to destroy anybody to begin with.
Sarah: Yeah, yeah, yeah, yeah. think you had a very big epiphany, didn't you? Sounds like a lot
Stephanie Antonian: Yeah, I know.
Sarah: of questions, but it's good because you've obviously, you you've got self-awareness in abundance, clearly. a lot of people might not have self-awareness in abundance. I think it's just interesting to me to hear you talking about it. I love self-aware people. It's great, but it's also a skill to be able to sort of, said, take yourself out of it and ask yourself the questions to think about what it meant for you. And I think that's an amazing way of
Stephanie Antonian: Yeah.
Sarah: starting it because if you relate it to anything in life, I used to say you can't be truly happy and give some else happiness if you can't be happy by yourself or you know that's kind one
Stephanie Antonian: Yeah.
Sarah: of the things I it's a similar sort of way of thinking about it but I think I think it's a good thing to be able to do that.
Stephanie Antonian: Yeah, yeah. Yeah, and that's why we're all here, like, how are going to make the world better? And it's like, well, how do we make the people who we sit next to, who's lives bearable? Like, let's start there. Like,
Sarah: Yeah.
Stephanie Antonian: you know, you would imagine that these teams would be the happiest people in the world. Like, you know, to do the job well, you'd have to be full of joy, really, to spread it around, not full of anxiety and panic and fear because that's just what moves forward. Like the nuts won't get shared, but I digress.
Sarah: Yeah, yeah And so it kind of leads me on to the next question, which is, So you're connecting the rise of AI with loneliness, falling life
Stephanie Antonian: Hmm.
Sarah: expectancy, carbon use and war. And while these things do happen alongside AI, how do you distinguish the effects of the technology from sort of business models, political choices and the institutions that are using it?
Stephanie Antonian: well the tech itself is is like neither here nor there like it's got no value or Like moral judgment to it. It can be used for great things it can be used for awful things and it's a bit like It's funny that I did Bible studies because it feels so relevant feels like the most relevant degree you could do for today. It's a bit like God, God can mean anything to people and you can use God as an excuse to start war, but you could also use God as your motivation to feed the hungry and do like wonderful things. So, know, AI now is sort of, it's this abstract term that can do anything and mean anything. And it's really the systems that we put it in that turn it into something really negative. But, you know, even when I say machine learning, it's not rational. It doesn't mean it's not useful. It can still be useful. There's still lots of cases where identifying patterns at a mass scale is really, really valuable. It's just knowing how and when to use it. But when we put it into these systems, it is going to cause catastrophe. we like to blame people like the tech leaders. And we're like, you're all so evil and blah, blah, But it's really not like that. If you create some tech. and you put it into the markets, the incentive structure is such that it will become exploitative, but that's not you, like that's a whole system. And there are problems with our system. We know that. And we know that we are heading towards catastrophe because of the way our system set up. So like we know we're heading towards climate catastrophe. We know we're looking towards a nuclear war or like we know we're looking at huge wealth inequality and so ideas that systems that will identify patterns and then put you on that track aren't particularly helpful because like we don't need what's probable like we need what's possible we need something totally like out of the blue like some miracle basically to come and set us on a new track so It doesn't really matter what tech we build and it doesn't even matter if we come up with the most amazing rational system or like AGI or whatever yet. As soon as it touches the current marketplace, it's exploited. It's it's baptized with oil. Basically it's like it just couldn't, it has no chance to do well.
Sarah: Yeah, well, baptised with oil. You're coming out with some absolute gems today, I love it.
Stephanie Antonian: Yeah, it's like that's why it's tough because people want to blame the tech leaders and they want to be like they're evil or they're whatever. And like they're actually quite nice people.
Sarah: Yeah, yeah, I mean, it's good to hear you say that because... For people who don't work in the industry or don't really spend their day to day learning about what's going on in the industry, which is having massive impacts on XYZ, people, it's like you can attribute this to lots of different things. People will look for the easy answer and the easy answer is, it must be the leaders of the tech companies who are the bad ones.
Stephanie Antonian: Yeah. Yeah.
Sarah: But it's good to hear that your rationale for why it's not necessarily them. And it's actually kind of a
Stephanie Antonian: Yeah.
Sarah: part of a much bigger systematic problem.
Stephanie Antonian: Yeah, like no one's that powerful by themselves. And the problem is, yeah, it's like, in life, you always think you're doing good. Like in the moment, you always think you're the good guy doing good and helping people. And then later you look back and think, shit, I was a bit of a dick actually. Like
Sarah: Ha
Stephanie Antonian: I should have messed that up. you know, and you see loads of people leaving the tech companies, but also people who are still in there who were like, really grappling with those things, because they genuinely do care. But
Sarah: Yeah.
Stephanie Antonian: like, you think you're doing good at the time. And, and then there's all these unintended consequences. And it's overwhelming, or you know, you do think you can change it or what have you. But it's like one mind can't hold like, the whole systems of how the world works. And Yeah, that's what I think is almost like even scarier is that it's not like there's this evil group of people being like, I'm gonna dominate the world. It's like people being like, wow, this is so cool. And I think it could be used for this and this and this. Wouldn't it be great if this and then the rest happens.
Sarah: Yeah, yeah, A story is old of time. But I want to talk about how sort of... Prediction shapes perception. So obviously this podcast is called Perception. And I like to speak to people from different areas around, you know, asking how people, organizations, and increasingly the AI systems decide what matters. So if machine learning systems infer what is probable from existing patterns, what happens when that starts to look like the most repeated or common account? And that's the truth that we get.
Stephanie Antonian: Room.
Sarah: like it seems it kind of blows my mind when we think of it in this way but what are your thoughts on that?
Stephanie Antonian: Well, I mean, I think we're sort of living through that. I mean, you just need to go on LinkedIn to think, what is going on? Yeah, like, can't, I can't live like this. Like, I can't, I can't, I can't read this stuff anymore. And I think it all converges to the same. And, you know, you see it happening with the media and with the news and it's like, everyone's sort of saying the same thing. You don't really get that much diversity because it's not just the tech. that's moving to it, but like the incentive structures are putting everything together. So even for example, like it is mad that I'm the person who wrote the essay on rationality and machine learning failing the principles of logic, like I had never read it before, couldn't find it anywhere. It is absolutely mental that that that I had the opportunity to write that because there's so many philosophers as people whose jobs it is to be philosophers in AI. And there's academics and thought in AI, but it's all funded by big techs. Of course you can't say that. Like, of course you can't say stuff like this. Like this is not what people want to read or want to hear in the intent structure. So the good thing is, is that I think it makes people just turn off. I think we just get so bored that people turn off and they look for more creative things. And so I actually think it's a bit of a good news story because it means new people have chances to say interesting things. And the ideas could come from anywhere. And people are looking more and more for different stuff. like, know, reconnected because I handed you a paper essay. You know, that that that's because like the reason why you read it is because there must be a party that's fed up of all this love and all the bullshit that's being said. So
Sarah: yeah. Yep.
Stephanie Antonian: when it converges, it's really it's, you know, a bit it's scary and it's depressing, but then on the flip side, we just all start looking for something else. We just go the other way, it swings on a pendulum. We're like, okay, I want something interesting. So I'm to have to go to another source and I'm going to have to find more original thinkers and maybe I do like artistic. And so I think it's a wonderful time for creatives now if you can hold your nerve, because it just means people will look for something else. And that's where it on the business side, that's where it's quite difficult because the more you embed these these tools, the more you're aligning to convergence, which means the less you the less competitive advantage you have actually in reality.
Sarah: Yeah, yeah, and you just become reliant on them. you handed me a piece of paper that was in itself interesting to me. And what else? So it's, yeah, I'm bored of the slop. That's absolutely one thing. But I'm also, you know, as a new business owner myself, I have never had to do what I'm having to do. All of the things that you have to do when you're starting a business, you have to be content creator, you have to do PR, have to do,
Stephanie Antonian: Yeah. you
Sarah: you have to sell your services, to everything as you know, but I guess one thing that I've sort of found very overwhelming, it's not just the slop, it's the, you know, sitting behind a screen, which is a very small screen, and the world seems infinite, and it's overwhelming, because actually
Stephanie Antonian: you you
Sarah: everyone's a creator these days, you don't actually know where to look, or it's just a sea of overwhelm, I think, online. So to have something physical, and like I said, I didn't read it straight away,
Stephanie Antonian: Okay.
Sarah: because I was, you I had loads of stuff going on, but I put it on my desk and I looked at it and I went, yeah, that's, I need to remember to read that. So it's having that physical, and it was creative and I really liked the way you thought and you positioned it. So it's kind of the, takes me back to human, I met you in person, we didn't speak for very long, but I remembered you. I had this physical
Stephanie Antonian: Okay. Okay.
Sarah: piece of content on my desk, which no, I've not had that for, I don't remember how long, and it was a different point of view and it was clearly not AI slot. So there were three things for me that made it really powerful. And I really wanna go back to, I just wanna go back to the humans. So I think conferences are very important.
Stephanie Antonian: love the human aspect too and what I'm noticing is just more and more communities building like in in person and more analog stuff and I see more and more people spending less and less time online and that's why I think it's like we don't need to be scared of this tech and anxious this tech because it kills itself. has its own. It has its own ending. Yeah, like, okay, amazing. You can make an app in the five minutes, but like, I can't have a million apps. I'm only interested in tech. So, you know, there's there is it, it's got its own endpoints and it destroys itself by itself. So focusing more on what's more fun and positive like building connections with people, building community, like doing things in real life. that's the way it's going. I mean, that's what I see. I don't even think this is a revolution. I think it's more of a renaissance because the number of communities that are coming together is really incredible.
Sarah: Yeah. I also see that. I've been saying that for quite a while whenever I'm speaking to anyone about it. I'm seeing a real shift back to human and human connection. I think, you know, we had the pandemic, it took a while for I think people to get back to that mentality of getting back to face to face. And
Stephanie Antonian: Yeah.
Sarah: I think people are just so fed up. Personally, I've been sat in a room for years now and on my own and I don't like that anymore. I don't want to do it. I was also wanting to ask you, so as the AI systems increasingly mediate the discovery and comparison and recommendation, what do you think credibility means for a business or a person? What does a machine need to establish before that interpretation can influence a human decision?
Stephanie Antonian: Just that it's correct, that the answer is correct.
Sarah: sound's so simple.
Stephanie Antonian: Yeah, does it give you the answer that you want? Like, because if it does, then it's great. That's fantastic. But like, if it's wrong, like three out of five times, four out of five times, one out of five times, like you need to know that because then you need to adjust that if it if it's correct, it's fantastic.
Sarah: Yeah. The question is, how do we ensure it's correct?
Stephanie Antonian: you just got to test how many times that you just do a test on it. I'm like, it's funny because people are like, we got to explain how the model works. We got to do all these difficult things. It's like, let me tell you, just is it just look at the endpoint and is it useful? Is it not useful? Like, you know, you can search on chat GPT, like ask all the things about your company 20 times and then look at how many times it was correct. Like, you can run small experiments to see how much you can trust it. And it's really interesting because there's so much talk about trust. But really, it's coercion because to be trustworthy, you don't have to talk about trust, you don't have to go and be like, I'm gonna build everybody's
Sarah: You
Stephanie Antonian: trust, you just need to do what you say. Like you just need to be like, like that's, that's it. That's all it takes to build trust. Like, but when you have these huge marketing teams on like trust and safety and, It's really coercion because you basically say I've not built something that's correct, but I want you to trust me anyway. It's like, well, just do it properly. Do what you say, follow through and that's it. People will trust it. I'm not saying
Sarah: Yeah.
Stephanie Antonian: you shouldn't trust a machine. There's lots of things that I trust that I don't understand. When I go to the doctors and I get prescribed medicine, I do not understand how the medicine works or whether... What ingredients in the medicine but I take it, you know because I trust them like or
Sarah: Yeah, yeah.
Stephanie Antonian: the building I'm not like always this up to spec. Is it gonna fall on my head? I'm like, okay, don't understand this stuff Trust that somebody does and that you know, it's fine. So I'm like you it's fine to trust machines as long as that crack
Sarah: got one more question for you before we go. And I just want to know what you think a genuinely responsible relationship between that prediction and judgment would look like. So like where should machines help us see patterns and where do you think humans still need to remain accountable? Like goes back into that thing about trust, doesn't it? Like how much should we trust machines versus not?
Stephanie Antonian: Well, So there's lots of use cases where looking at patterns at scale is really helpful. Like, you know, in medicine, it is helpful. And I'm not saying that AI won't come up with great or machine learning won't come up with new drug discovery or cures. I'm just saying, okay, tell me now once you've done it, I don't need to hear about it every day. I'm happy for you. That's not what I'm going to be working on because I don't have the technical skills for that. But so let me know when you've done it. But there's lots of different, there's loads of use cases where looking at patterns is really helpful. Personally, I don't think judgments about people should be made by machines, even really by people. I'm trying to let that stuff go. But that's where I think it gets really difficult because whenever it starts to become judgments about people, it's like it's only ever used on superficial layers again, because of the way the technology works. So like it, it will look, are you a woman? Are you Middle Eastern? Like, are you educated? Are you wealthy? Like it will look at superficial categories and then make its determination of yeah, you could, you should get alone. Like you're a white man who's rich, you get alone. Like, you know, you're a woman who's foreign, like you do not get a loan. And, that's using historical data, obviously, but the issue is like, it just keeps us entrenched and judging people at these really surface level factors. And that's what I think takes us away from intelligence. So at beginning, I said that love and rationality are actually linked. And what had, you know, before all this like, got me started on it is just We used to be really aligned on the philosophy of the love of mankind. Like we used to, like this, central principle of everything was this unstated, or actually sometimes stated view that to love mankind was intelligence. And that's what a lot of the Greek philosophers also talk about alongside rationality and like all of this only works when you value other people as yourself. And like even Adam Smith of capitalism, he's like, spoke a lot about how the invisible hand is the love of mankind, not necessarily only greed, but he's a Christian theologian as well. He talks a lot about love in his other works. But that's all got gotten, been forgotten, because machine learning cannot treat humans as equals. Because as
Sarah: Yeah.
Stephanie Antonian: soon as you bring judgment in it has to determine who's more worthy than others about other things. And so I think there's lots of amazing use cases for machine learning on identifying patterns. And where the red line should be is on making any judgment about humans. Like if we genuinely want to move to intelligence, enlightenment or something like that, I think we have to have the principles that, you know, all humans are equal. And so it becomes too difficult for it to make the judgments. But the problem with that
Sarah: Wow, yeah.
Stephanie Antonian: is that those are most of the uses that they're selling.
Sarah: Hmm, yep, yeah, well yeah. For now, we'll see, we'll see, because there might be a shift change at some point when people start to realise that,
Stephanie Antonian: Yeah, and I think, for example, like in recruitment, there's so many AI tools for recruitment. And now it's like, you know, AI read your CV, it decides who's going. But the reason why it's not like, it's not very useful is that most people don't really know what makes a good hire. So like, you give it all this date, when you give all the data to an AI system, it's going to look at like, what universities people go to, did they go to university? like where they work before, what are the brand names, like what's the nationality of the name, is it like, it's going to look at all these really superficial things and then it's going to say, yeah, they can have an interview. But none of those things are really what makes somebody a good hire. Like,
Sarah: Yeah. Yeah.
Stephanie Antonian: like often you just say, do you know what? I really like It's like something or somebody tells you a story like. met with this one girl and you know she was doing an evening degree and was a carer for her parents and I was like anything you need from me I will be it like I'm like I can read between the lines understand so much about you for what you're doing but all of that goes so the people aren't the best but because the core issue is we don't actually know what makes good hire like can't put it into words
Sarah: Yeah, yeah, yeah. Yeah. I mean, that's a great example of a use case that everyone will understand because everyone's been through a hiring process and recruitment process. Fascinating.
Stephanie Antonian: Yeah, yeah, which is also why they've just turned nepotism because I'm like, it's just so much stuff is coming to the top that then you're like, okay, well, my friend's child's reached out. So
Sarah: Yeah, yeah, yeah, yeah.
Stephanie Antonian: not good. Yeah, as someone who really depended on meritocracy, like,
Sarah: Yeah, yeah, yeah.
Stephanie Antonian: not good, that is not what you want to see. Yeah.
Sarah: No, I feel like that's a whole other podcast. That's like opening a can of worms. But I am going to wrap it up and I thank you so much I can't believe time has gone that quickly. I've really been engaged and listening to everything you're saying and I think it's fascinating and I'm looking forward to continuing the conversation. But is there anything else you just wanted to share before we wrap up or do you want to?
Stephanie Antonian: I
Sarah: Go back to your day.
Stephanie Antonian: you so much for having me on. was really fun talking to you. Thanks. Bye.
Sarah: Great, all right, well thank you so much and I'll see you soon. Bye.
