Show Notes 204: What if we got AI right? Why Ethics Matters More Than Ever

When Dr. Eleanor Drage walks into a room, you know you're about to have a conversation that challenges everything you think you know about AI. And that's exactly what happened when we sat down with one of the world's leading AI ethicists on the Cambridge Tech Podcast this week.

If you're building the next generation of AI companies, this episode is essential listening. Here's why.

The Problem with "Ethics Washing"

We've all seen it: companies slapping ethics statements on their websites and calling it a day. Dr. Drage calls this "ethics washing," and frankly, it's everywhere.

The real issue? Most organisations don't understand what they're actually trying to achieve. When a media company titled a panel "Balancing Bias," Dr. Drage asked the simple question: what does that even mean? The answer was revealing, because you can't balance bias against profits. You have to choose.

The Communication Gap Nobody's Talking About

Here's something that should worry every founder: the people building your AI systems and the people talking about AI ethics are speaking completely different languages.

Dr. Drage spent two years interviewing AI engineers about bias. Their answer? "It's the intercept between the x and y axis." Technically correct. Meaningless to the conversation happening in the media and boardrooms.

"These conversations we're having in the media about bias and debiasing are completely meaningless to the actual people making these technologies."

The HEAT Framework: A Practical Blueprint

So what does good practice actually look like? Dr. Drage and her team created HEAT—a framework that distils the EU AI Act into something founders can actually use.

Key principles:

• Human-centred design (talk to your users, not just your engineers)

• Ethical development at every stage

• Accountable processes

• Transparency throughout

• It's not linear. Ethics isn't a box to tick at the start—it's cyclical and constant.

The Stories We're Not Telling

Media portrayals of AI are stuck in a doom-utopia cycle, recycling the same tired imagery: the Terminator, digitised bodies, the Vitruvian Man. But these images tell us nothing about what AI actually is.

Instead, Dr. Drage advocates for the Better Images of AI resource—copyright-free alternatives that tell richer stories.

Cambridge's Competitive Advantage

What's fascinating is how Cambridge is doing this differently. Projects like "AI in the Street" bring academics and local government together to understand real-world impacts. The result? Innovation that's grounded in actual human needs.

The Takeaway

Whether you're a founder, investor, or engineer, this episode challenges you to think harder about what "doing AI right" actually means. It's not about grand gestures or impressive statements. It's about having difficult conversations, listening to your engineers, and building systems that reflect real human values.

Dr. Drage's new book, What if We Got AI Right?, explores these ideas in depth with practical examples from around the world—from pollution monitoring in the Niger Delta to sustainable cloud computing in Latin America.

Ready to challenge your assumptions? Hit subscribe and tune in.

AI Transcript

00:03

James Parton

Welcome to the Cambridge Tech Podcast, talking all things technology from the heart of the UK's tech capital. Here are your hosts, Faye Holland and James Parton.

00:28

Faye Holland

I'm Faye.

00:30

James Parton

And I'm James.

00:32

Faye Holland

So I'm delighted today that we have one of the world's leading AI ethicists on the podcast with us, Dr. Elena Drage. So, hi, Elena.

00:42

Eleanor Drage

Hi. I think maybe there's not many of us, so now we can claim to be the world's leading something or other, which is a real joy.

00:50

Faye Holland

It's brilliant. It says here on your long bio, very long bio, which is amazing. You're one of the 100 brilliant women on the AI ethics list. So I'm calling it out for you here.

01:02

Eleanor Drage

Oh, that's very coin. Thank you.

01:04

Faye Holland

The other thing, I mean, I'm not going to read the many different accolades that you have and the experiences you have under your belt because I think lots of people will know of you already. But the one bit that I absolutely love is that you describe yourself as a philosopher gone rogue.

01:22

Eleanor Drage

Well, I mean, I shouldn't have even been an academia and I'm barely a philosopher. I do what we call critical theory, which is a bit more modern, a bit more contemporary, less Aristotle, more justice. And so that's probably the rogueness of it. And particularly as we're talking in a Cambridge context where people are very specific at the university about what kind of disciplinary expertise counts and what doesn't. I'm pretty sure nothing I do would count as being kind of solid academic and any particular field. But that's what makes it so much more enjoyable, I think, than sitting in a room and not really having any human contact.

02:07

Eleanor Drage

What I do is very out there in the world talking with real people, asking them what effects of AI they feel, and trying to come up with cutting edge ways that are embedded in really good research at Cambridge to make AI better and safer for everybody.

02:24

Speaker 4

Excellent. I mean, it's a real pleasure. We certainly appreciate you taking the time as a seasoned podcaster yourself. We'll be looking for some tips from you after the interview. I think one of your previous books, the Good Robot, will be familiar to our longer term listeners because we covered that back in episode 80. Today, though, we're very excited to find out more about your fourth book. What if We Got AI Right?

02:44

Faye Holland

How did this book come about? I like the title for a start off, what if We Got AI Right? Because it's initially actually starting the conversation a different way because there's a lot of doom and gloom and a lot of it's going to go catastrophically wrong, which is also a word used in the title. So why did you start from that space?

03:07

Eleanor Drage

Well, my mum isn't going to read something that is really depressing for a start. And I'd written all these other books that she probably couldn't get past two pages. And I really wanted to put something out there that was practical and useful and connected to the world, do a lot of critique. I mean, I just talked about being a critical theorist, whatever that means. And of course a lot of the book is very much embedded in the realities and the stupidity of bad science and silly ideologies that make their way into AI. Then I wanted to do something more because I'm quite a practical person. So I do always ask, well, what do we do now? What now?

03:49

Eleanor Drage

And I was reading this morning a nice quote by a guy called Manus Ray, who's a professor in cultural studies at the center for Studies and Social Sciences in Calcutta. And he said, instead of being a routinized supplier of mere critique, today's intellectual needs to provide a more informed basis for practical choice and imagination. And that's really what I want to do. And I don't want to just cut things down, I do want to rebuild them.

04:20

Eleanor Drage

And I'm so inspired by the people I talk to all the time on my podcast, the Good Robot, but also at conferences where I see people creating technologies in low resource environments with off the shelf computer parts and solar panels and using them to collect really important data in the Niger Delta, for example, on air pollution and technologies that people understand and local people can put together and take apart and hack. And I think I'd be remiss and a bit immoral, in fact, if I wasn't focusing on these incredible efforts that get basically no media attention. I'm really inspirational and can help everybody have a good understanding of what actually it means to get AI right, and not even just AI, but all technology. What is good technology?

05:13

Eleanor Drage

And how are people going about creating really good approaches to medical technologies that understand that race isn't a biological reality, it's something that's socially constructed and has been embedded in the medical sciences for a long time, or technology that can be used all around the planet that hopefully we'll talk about to redistribute wealth and are just, I think, worth waking up for and worth spending every day taking steps to say, right, we can make a good go of this.

05:50

Speaker 4

Yeah, that's really interesting. I think so much of this is around good narrative and storytelling and what's really interesting is this piece around how the media is approaching AI and it tends to, I think you described this as this doom and utopia cycle where it's either extremely great or extremely negative. Of course, I guess you are arguing that's not helpful. Of course there's a challenge though to kind of quickly and simply translate these really complex technology and ethical concepts into something that's suitable for the mainstream. And we do live in this kind of bite sized information world these days. So what's your take on that? What's your advice to media? How can they constructively move the conversation forward?

06:28

Eleanor Drage

Well, one of the things that me and my team at Cambridge, at the center for Future Intelligence, that I've looked at for a long time, and I'm a newbie, relatively speaking, to this amazing team who's been exploring the narratives, the stories that we tell around AI for ages, like Dr. Stephen Cave and Counter D', Hal, is that in the media we are only seeing the terminator and we are only seeing datafied bodies that are blue and see through and turned into numbers. And that doesn't tell us anything at all about AI. In fact, I think we're being descaled by these images and these portrayals of AI. And why do we see them a lot? Well, they're free, they don't have copyright, they're very easy to use, especially from by media companies that don't have a lot of money to commission things.

07:21

Eleanor Drage

But then even the BBC have used images to describe Yuval Harari's work that are just appalling. Like there's a one image of the orthogonic view of evolution which is ape to man, the ape in different stages of development rising up to man. And it's got like Hamlet and Shakespeare and a, they're all male. And then at the end you've got some sort of transhuman thing. And I know that they're trying to be tongue in cheek, but that view of evolution has been widely debunked. We didn't develop in a straight line. There is no straight line to evolution. Evolution is wildly complicated and it doesn't account for this sort of view of human progress, doesn't account for all the complexities of illness and queer lives that don't depend necessarily on reproduction or disability.

08:21

Eleanor Drage

And they're very grounded in eugenicist ideas about human progress that I think are enormously detrimental. And so when we see images of the Vitruvian man who Da Vinci is perfect, very beautiful man that none of us look like, and of course there is no Vitruvian woman hanging above the desk at DeepMind and in some newspaper which describes the future of AI. What we get a sense of is how completely insignificant we are on an individual level compared to these amazing profound images of human transformation in the future. And I find that really isolating. And the other thing is there's a great repository of images if you're wondering where to go to source really great images of our AI.

09:09

Eleanor Drage

And one of them is called Better Images of AI and it's put together by Tanya Duarte and others and one image from it that I really like. They're all by the way, copyright free and artists, as long as you put the artist's name and credit them so they're free to use, it's great. Is a block of silicon, a silicon monocrystal, which is what we use. It's made out of quartz. And so you go into the quartz mines in North Carolina, which by the way are hugely susceptible to climate change. And then from there you can make these silicon blocks. And those are used for microchips. And the silicon blocks are so beautiful, if you've never seen one, look it up. They're sort of glistening, shiny, kind of fool's gold. And I love the materiality of the, of technology.

09:57

Eleanor Drage

So I really like the technical things. I love the, the tangible. Not just some ephemeral idea of what the Terminator might be in the future. But these images are beautiful and they tell us a story about the environment, about the creation of technology, about the labor that goes into it, about geopolitical tensions and kind of the rush to own these resources. And that gives us a really good understanding, I think, of what AI is and what it's made from.

10:28

Faye Holland

So I absolutely love that. And I was writing down where can we get better images? What images do you want to see? So thank you for giving listeners something really tangible they can do. And it kind of makes me think of the amount of when you see people using stock image of it's all males on panels and there's no doubt, it's almost like you need to police it a little bit. You need to go, don't use this image. Use, go to this resource. So thank you for sharing that. And I think that was a good flavor in terms of what people, our listeners can read more of in the book in terms of the doom Utopia cycle.

11:01

Faye Holland

The other thing you talk about quite a lot that I want to dig into with you is around Ethics, Washington and what you actually mean by that. What it looks like and can we spot it? What do we do about it if we do spot it?

11:16

Eleanor Drage

Well, I think we've had a lot of conversation about the kinds of ethics washing you see from, I don't know, supermarkets, putting pride flags on sandwiches and calling it a day, or making some vague gesture towards AI ethics in a statement on a website. And a while ago I interviewed Alex Hanna, who was working at Google and had just quit, and they were doing this systematic takedown of the different ethical imperatives that they had at the time that they just weren't acting on. These were just words on a website effectively. And unfortunately we're now seeing that across the globe.

11:56

Eleanor Drage

I'm doing a study with Professor Muhammad Ghali and my amazing research assistant Farah, and about AI ethics frameworks in the Middle east and how they just copy paste Silicon Valley ethics frameworks, which is really unhelpful because of course the Middle east has very significant cultures and very diverse ones. And if we're all copying and pasting not just ethical imperatives from the west, but from Silicon Valley where they have no ethics, really, that's move fast, break things, culture exacerbated by Trump in his immorality, then we're creating systems that are not grounded in culture or context or anything. So that we've seen, I think, and people do know about that. But what I was more interested in is companies who are trying to make an effort but just don't know what they're doing and end up sort of doing nothing.

12:49

Eleanor Drage

And a while ago I was giving a talk at a big media company and the title they'd chosen for the panel that I was on was called Balancing Bias. And I thought that was interesting because you can't balance bias. And I asked the audience, these smart people at the company, what does balancing bias look like? And of course they said, well, no, I don't think you can, because what are you balancing against profits, really? We have to ask ourselves and have companies ask themselves, how far am I willing to go in pursuit of justice to turn the table, really? Because ethics does not come by our hand holding pleasantries that you think of when you hear the words balancing bias. It's not a yoga pose, it's not a balanced diet. It's really difficult.

13:46

Eleanor Drage

And ask anyone who works in a charity or in a feminist organization, they'll tell you, my God, anything pro justice is really hard and involves a lot of difficult conversations and disagreements. And so those are the kinds of questions and the kind of heightened pitch at which I think we should be having these conversations. And certainly the employees of this company were ready to have a meaningful conversation about what it meant to do good while using AI in their media environment. And they felt a bit let down by the wishy washy balancing nonsense that the C suite was, I think, more comfortable with. And the last thing I'll say on this is that I interviewed AI engineers at a big tech company for two years about what they thought bias was.

14:39

Eleanor Drage

And they said very sensible things that make sense in an engineering context, like it's the intercept between the x and y axis. And how do you know what bias you're looking for? Because there's lots of different kinds of bias. And of course, that is what you'd expect a mathematician or esthetistician or an engineer to say. So these conversations we're having in the media about bias and debiasing are completely meaningless to the actual people making these technologies. And I think we really need to have a sit down and listen to and respect the expertise of these people and say, what language around AI and harm is meaningful to you? And how can we integrate a really good approach to AI across the company in a way that's meaningful to everybody?

15:31

Speaker 4

The next question, I guess, is a continuation of that train of thought. As you'd expect for a tech podcast in Cambridge, we're talking to founders, investors every week that are building the next AI generation of companies. Right. What does good practice actually look like at that practitioner level? Because in the absence of any frameworks, how do they know they're making those right calls? Or are there examples of best practice that you might be able to share where they could take inspiration?

15:58

Eleanor Drage

Yeah, absolutely. So with my team and a team of AI engineers at a great company called Amagama in Italy that's now been bought by Accenture, we created a framework for AI development called HEAT. HEAT H E A T T and I'll just check heat.accenture.com now. And it's amazing because it takes the European Union's AI act and it distills it into an easy framework anyone can understand, and then it defines it through cutting edge approaches to pro justice AI ethics, basically in ways that are really simple and meaningful to people. So if you're an engineer, you can see, right, this is what I should do at each stage of AI development. And we have the kind of seven different stages and we try and make it so it's not just linear, but we show that maintenance and ethics is something that is constant and cyclical.

16:59

Eleanor Drage

And you have to come keep coming back to and ask new questions and different kinds of questions depending on what stage you're at. So I'm really proud of that. And the work of my team who are experts in participatory design, for example, who say, like, look, if you're making a HR AI tool for recruiters to hire different kinds of applicants, then you need to talk to recruiters and you need to talk to applicants. You can't just sit there in a vacuum and decide this is a great way to do it because you'll end up with real nonsense. Like some tools I debunked a few years ago as part of a big study that was on the BBC that claim to be able to understand your personality by looking at your face.

17:46

Eleanor Drage

And they claim that this is debiasing because it reduces humans to neutral data points and neutral data points don't have race or gender. And this was so ridiculous an idea. And with a team of second year computer scientists at Cambridge, we recreated a similar tool. And then I got these second years to sit around and talk to me about what they thought. It was really sweet because I thought that engineers would, I don't know, not really have a politics or an opinion about this kind of stuff and that I'd have to explain a lot about what race and gender were and why it's not something that can be eradicated from the candidate profile. And why then when you went back into a company, you want to be treated well, right?

18:34

Eleanor Drage

And they said to me like, no, this is awful because if I enter a company through one of these tools, what about childcare? What about being treated well? What about being promoted in a way that's equal at the same time as everybody else. This doesn't guarantee any kind of positive relationship to a company that's going to employ me. And I don't want a recruiting software that's colorblind, that's going to parrot some liberal ideology of sameness. I want to be seen and loved for who I am and really appreciated as a candidate. So, like, it's always a pleasure to work with engineers at different stage of their careers and hear their voice.

19:15

Eleanor Drage

And I know that there was some real problems with engineer well being in Silicon Valley a while ago, because being asked to develop systems they didn't agree with and not really having a language to explain what they thought was wrong. And so that's why I'm so encouraged by watching engineering teams, for example, in Latin America, develop small scale, environmentally acceptable cloud computing technologies that don't rely on Amazon Web Services are embedded in particular communities. Or look at data center creators who are interested and invested in modular data centers that are not enormous hyperscalers. They're small, they are very well planned. So the energy isn't coming from fossil fuels available last minute, but from renewables. And then the heat is going back into a sports center or a school or a sauna is being well used. And Wikimedia foundation are really excellent at creating that.

20:16

Eleanor Drage

So there's lots of really cool ways that people are right now developing great technology.

20:21

Faye Holland

I think that's a great example to pull out there. And obviously the book is full of other examples as well, isn't it? So, one thing I did want to also ask you was around. It's a little bit around Cambridge, so we've talked a bit about AI, the tech side. Now I need to talk a little bit about Cambridge and I appreciate your remit and AI and what you're doing is much broader than Cambridge, but I think it's a really good example because it's a city that is both a big driver of AI development and also home to leverhulme and other serious thinking about the risks associated with AI. So do you see a tension between those things happening, the real innovation space, and then aligning governance and policy in those types of areas?

21:12

Eleanor Drage

Well, I feel really lucky to be in Cambridge and I've spoken at Cambridge Tech Week and seen firsthand how the frugal AI lab, for example, at Judge Business School, is working with different kinds of innovators to make sure these amazing approaches to AI that are frugal, by which I mean low resource, low energy, can be embedded in different kinds of technologies. So that's really co. And then my colleague Maya Ganesh and Louise Hickman did this incredible project with the local government in Cambridge, pointing out AI in the street. And essentially they went for a walk with the Cambridge Council.

21:55

Eleanor Drage

And Louise, who has lots of different kinds of access needs, but she's in a wheelchair, was pointing out to them, the people from Cambridge Council, like I that, look, I'll get up this bend and there's this thing watching me and I can't get round here. And I just. I mean, poor Cambridge counsel. But this is this amazing tension between, like, these really feisty academics who are really good at understanding what AI innovation is and how it's made available to everyday publics and how that affects people's privacy and access needs in the street, for example, with local communities. And they're super invested in working with the council, improving the plight of everyday people with Access needs and just connecting on a bigger level.

22:49

Eleanor Drage

And I think actually because AI is so interdisciplinary, a lot of the AI academics I've met at Cambridge worked a lot with the city and with other innovators across the city and they're just much more chatty and approachable and interpersonal and willing to participate, I guess, with the world. So, yes, you see that tension and that's really important because it does make change. But I think it's really productive. And there's no place quite like Cambridge.

23:26

Faye Holland

Absolutely. Elena, thank you. I know we've only got a short amount of time with you today because you're on the proper circuit here promoting the book as well. What I will say is that the comment you made earlier on about you want something that your mum's going to be able to pick up, I think the whole point of the book is, and you've said it on the podcast here, wishy washy balancing in boardrooms, bad science and silly ideology. You really do get into that level of detail within the book. So thank you very much. And where do people go to get a copy of the book for themselves?

23:58

Eleanor Drage

It's a great question. I need to get out so I can see, but it seems to be in most bookshops. I got sent a picture of my friend in Switzerland holding one. So anyway, you get your books and then online in the usual places. I think it's only like 14.99 or something, which to me seems fine. So, yeah, anywhere online. And then also I should mention my audiobook which I spent three days recording myself. It's not AI, it's me with loads of ibuprofen in the recording studio and some throat spray. And I think that's really nice actually, because some of the book is quite gossipy. It's quite like who I don't like and don't think is great in the AI sphere. And you can hear my own voice views about these different people. So that was fun.

24:46

Faye Holland

If you do well, when you do get out, and I suggest you do get out and about as well, go and do a Tim Minshall because he's been posting all over the place where he's reorganizing the bookshelves to make sure his is at the front. So I expect to see you in a bookshop doing the same.

25:01

Eleanor Drage

I do that all the time. I constantly taking out my books from the feminist section and putting them in like big reads or like bestseller. We all do it. It's that like, trope of the guy in American fiction, the movie is taking his books out of the like Black Life bookshelf and putting it into some more widely read section of a prominent table.

25:24

Speaker 4

Nothing wrong with a bit of growth hacking. That's well, thanks so much Elena for joining us today. It's been a real pleasure. Tune in next week when we'll be revisiting the first of our two part episode with Cambridge legend David Clevely.

25:42

James Parton

Today's show was produced by Joe Donaghy of Cambridge TV and supported by our media partner, Business Weekly. The Cambridge Tech Podcast is available on all major podcast platforms and on cambridgetechpodcast.com if you've enjoyed this podcast, please give it a five star review. It'll really help others discover the show. Technology moves fast and now so can you. Discover Polestar, the all electric performance brand vehicles redefining how we drive. Precision engineering, minimalist design and software that evolves with you. Experience it at Holden Group in Norwich and Bury St. Edmunds or let us bring the test drive to your door. Proud sponsors of the Cambridge Tech Podcast Polestar electric performance redefined discoveroldengroup.co.uk Polestar.

Transcribed by https://fireflies.ai/

To listen and subscribe, search for ‘Cambridge Tech Podcast’ on your favourite podcasting platform or visit cambridgetechpodcast.com.