AI Special 5#: Sara El-Hanfy, Head of Artificial Intelligence and Machine Learning at Innovate UK : Achieving Ethical AI

0:00The risk of extinction, I think, is small where we are right now. It may be a risk that comes down the line, but I think there are other risks, you know, with some of these technologies currently.

0:14So we have finally reached the crescendo of this incredible thought provoking series. Carenza and I have had an absolute ball doing this five part collective of amazing conversations with some of the UK's top thought leaders within this space. So we had to end on a bang. Sarah L. Hanfi is the Head of Artificial Intelligence and Machine Learning at Innovate UK, also an advisory board member for UK Parliament and an advocate for diversity in STEM and digital inclusion. In this episode, we discuss the fundamental societal and business consideration in relation to the evolution of AI technology. We discuss things like ethics, AI in education, and also UK government policy and where the UK sits on the global stage.

1:04As somebody who's very well-versed in this matter, Sara gives us her advice or gives her advice to businesses looking to incorporate AI into their processes and product sets. This is such a fitting way to end the series. Sara is fantastic. Carenza's on top form. I think you will love it. It's Sarah Elhanthi.

1:34Sarah, thank you so much for coming on. Carenza and I have been so excited about this. Well, thank you for having me. It's so nice to meet you, Sarah, at long last. It's great to be here. Really looking forward to it. Absolutely. So much to talk about, so much to unpack. You have got the honour of being the final guest in the AI series. The grand finale. Yeah, the grand finale. No pressure. Let's save the best for last. So yeah, but thank you so much for doing this. We've got, like I said, we've got so many things we want to talk about, but we always start with this big question. What are you most excited about in relation to the evolution of AI technology?

2:09Oh, and that is a really good question. I think, I mean, there's so much to be excited about. I think at the minute, there's a lot of the talk about the risk and the safety of AI, rightly so but I think my background sort of is a non-linear path into tech I guess I've always had kind of equal interest in the arts as much as tech so I think for me the things that really excite me are more around the applications of the technologies and the amazing things that we're already seeing in health and the applications towards things like climate change and so it's those kind of applications which I find really exciting however the generative AI space is very difficult not to find that fascinating because I think even people like myself who I think are quite close to their technology, I think have been a little bit surprised at just the speed of the step change that we've seen over the last few months.

3:00Do you think that's because of computing processing power or do you think it's because people are beginning to get more bold and creative with the potential of what you can actually do with the technology? I think a lot of it is underpinned by the compute power that we've got nowadays. I mean, that's obviously led to this kind of current wave of AI. But I think that that kind of industrial research that's focusing on it is definitely, you know, leading to kind of people trying new things and trying new techniques. I think once you have one kind of lab concentration or something, it's encouraging more people to sort of crowd into that space.

3:39And that, I think it's almost like competition, isn't it? And that research from multiple angles that kind of raises the bar and encourages new discovery. Exactly. And I just think it's almost like the democratization of technology through a different means, isn't it? So many people now can have these tools at their fingertips before they were probably unaware that AI was in their lives through so many of the technology they interfaced with, but now they can actually have it and they can control it. And that is incredibly exciting. And I think it's really opened the conversation a bit more because like you say, I think there was a lot of AI that was being used day to day anyway.

4:13And I think society at large probably didn't really understand that or appreciate that. And I think one of the really positive things about things like ChatGPT is actually it's really raised awareness of people of kind of what these systems can do, but also, you know, open that conversation about kind of risk and harm and how we want them to be deployed. So I think that's a real positive that actually more people kind of grasp what AI is and then are starting to have conversations about that. I mean, I know in my friendship group constantly having chats about you can't escape it it's constantly the topic of conversation yeah for sure I mean Carenza and I were talking before that we kind of had a bit of a feeling in our gut when we first saw chat gbt that wow this is incredible it's going to be amazing and I kind of remember that feeling I remember where I was when I first saw it and interacted with it because I just it just felt like AI had finally come to the masses you know It finally sort of, the hype was real.

5:08Did you have a similar thoughts when you first saw the technology? Yeah, I think it was difficult not to be impressed by ChatGPT, but also the Dali, the image generation as well. I think, yeah, I think it definitely, like I said, was kind of like, I think it took everyone a little bit by surprise that we'd kind of taken that step up as quickly as we had. For sure. So for the listeners who may not know much about yourself and Innovate UK, can you give us a little bit of a potted history of Sara and maybe a little bit about Innovate UK and the purpose of the organization as well. Yeah, of course.

5:39So I'm head of AI and machine learning at Innovate UK, which is the UK's innovation agency. And I'll talk a little bit more about who we are and what we do in a moment. I'm also on the advisory board of the all party parliamentary group, AI. I'm an RSA fellow, and I'm also a mentor. So at the minute, I currently mentor in support of the 10 ,000 Black Interns Programme. Innovate UK is part of a larger organization called UK Research and innovation. So Innovate UK has been going for a long time, as have the research councils, but we've all been brought together under this umbrella organisation, which strives to create knowledge with impact.

6:14So we're funded by DSIT, the Department for Science, Innovation and Technology, to the tune of about£7 billion a year, and we operate across the whole of the UK. But while the research councils fund different research elements, so from arts and humanities to engineering and physical sciences innovate uk is very much focused on supporting business-led innovation so helping to drive productivity and economic growth supporting businesses to develop and realize the potential of new ideas including those that come from the research base but but not exclusively so we do that in a number of ways so probably the bread and butter way that most people know innovate uk is through our investments so traditionally we've provided grant funding to organizations, but increasingly we've diversified that as well through innovation loans or investor partnerships.

7:00So bringing our funding together with investors to sort of crowd that cash and knowledge into organizations at an earlier stage than they might otherwise get investment. But we also want to create that vibrant and successful innovation ecosystem. So that's through involving organizations and providing the business support that they need from parts of our organisation like Innovate UK Edge and also connecting them with the wider ecosystem through our Innovate UK KTN or Knowledge Transfer Network. And then finally, we sort of see ourselves as kind of inspiring. So basically creating that shared vision that the ecosystem and innovators in the UK can get behind producing strategies and roadmaps for priority sectors and technologies.

7:40So what's your day-to-day life like, Sarah? Walk us through your average working weeks. What sort of thing lands on your inbox? Oh, it's fascinating, actually. I think when I took the job, I was really excited that I'd come from the private sector, so very much working on specific projects or advising specific clients. And I was always quite interested when I took the job at that kind of macro level, being able to see what was going on across the whole of the AI ecosystem. So I sit in a team where I look after AI in a non-sector capacity. So I'm not focused on one specific sector, which is challenging because one day you can be talking about AI and healthcare, and then you can be talking about generative AI and then a completely different sector.

8:22So sometimes it's a bit confusing about where are we doing, what are we doing, what are we talking about? But I think that's what keeps it interesting and keeps it exciting. So if we're running competitions, there can be a lot of the kind of the day-to-day admin. So doing the business cases, getting the competitions through our sort of audit trail systems, dealing with project change requests for live projects that we have that we've given funding to, through to working with government departments on policy. So attending roundtables, reviewing documents, and also going out and meeting the community.

8:56So a lot of what we do is actually speaking to businesses and understanding what their needs are so that we can, A, when we have those conversations with policymakers, we can be helping to inform. But also when we're designing our competitions, we're designing something that's fit for perfect and actually going to have an impact and benefit the communities that we're trying to benefit. bit. I see us very much as that conduit of doing stuff for the businesses in this space. And we can have opinions about what we think is needed, but we're never going to actually design what we need unless we're speaking to them and using their knowledge to inform that process.

9:32It's probably a good moment to talk about Bridge AI, isn't it? You've just launched it. It must be so exciting. Very exciting. It's been a long time in the works. So to actually get there, it's been great. I think because Innovate UK has always been focused on funding innovation. and that kind of, I guess, identifying, supporting and accelerating AI innovation in the UK. But when the recent government innovation strategy came out, we had more of a remit as Innovate UK to focus on adoption. And I think for me, that feels the area in the UK that actually we've got a world-class research base. We've got lots of really thriving startup ecosystem, but actually the pace and scale of adoption could be faster.

10:13So the Bridge AI program is really there to do just that, to stimulate that adoption of AI into a number of key target sectors. So we're targeting agriculture, construction, transport and logistics, and parts of the creative industries. And we've got£100 million to both support the creation of end user-driven products and services, but also because it's an adoption and diffusion program, to really focus on building sector capabilities, So upskilling and reskilling people, bringing cohorts of adopters and developers together. So to try and address some of those sort of information asymmetries that might exist between different parties and really kind of address it as quite a holistic, not a one size, you know, there's no one size fits all when it comes to adoption and diffusion.

10:56So, yeah, it's a really exciting time for us to be launching the programme. And then we've got a website that we've just launched, bridgeai.net, which has a lot of information about either what's already live and also some of the interventions that are going to come on down the line. So if any of our listeners are in those sectors, how do they progress potentially working with you and benefiting from the programme? So I think the first thing is really to go to the bridgeai.net portal and sign up for more information, because that means you'll be kept in touch with as competitions are going live or as different support mechanisms are coming online, you'll be kept informed of that.

11:30But you'll be able to kind of apply through the portal and actually get triaging. So if you're not, you know, particularly as an adopter, I think if you're not sure what you need or you see a lot of the interventions and it's a bit confusing, we will be kind of triaging and helping people to kind of navigate the different funding options and the different support options to identify what's right for them. But also there'll be a lot of content, so either disseminated from some of the activities. So if we're identifying specific standards that need to be created or there's obviously generative AI being a really hot topic, we'll be starting to talk about some of those topics as well.

12:03So people can get a really good overview, really good understanding of ethics as well, of everything that encompasses adopting AI and be able to choose from whether that's they just need to consume a little bit of content or actually they need proper business support from one of the partners that we've got on the program. So it's a real spectrum depending on where you are in your journey. So, Sarah, you said that adoption could be faster. What kind of thing is holding this back then? So from the research we did, I think there's a number of different factors. I think there's a big skills gap that we have in the UK.

12:35Well, I mean, not just the UK, to be honest, it's a global problem, but it's both technical skills. So I think some of the organizations that might adopt AI don't necessarily have the technical skills in-house, but also, I guess, other skills that sit more in that kind of commercial or leadership space. I think there's a lot of uncertainty about how you calculate the ROI of these solutions. And so unless you can really kind of justify why you're bringing that in, particularly if you're kind of, you have low margins, that's a huge risk for you to take. Then I think there's a lot, you know, there's so much activity in the UK that actually then it can be quite hard to navigate where to go.

13:15So I think that was kind of what I was talking about, about the kind of silos of activity, because there's a lot going on in the skill space. But again, that's quite difficult to navigate. So we're hoping to kind of by bringing a program like this together, we can kind of collate all of that activity and help curate and signpost people to what's the most relevant thing for them at this time. I mean, other barriers, I think the ethics is a big concern for businesses because, you know, no one wants to do the wrong thing, but it's very difficult for businesses to know how to necessarily navigate that risk.

13:48So I think certainly some of the training pieces that we have there and some of the tooling that we're looking to develop will help address some of that challenge. And then, I mean, perhaps more for developers, but I think the cost, you know, there's certainly cost barriers around developing these tools because it's not just a standard digital technology. AI technology has a longer development pipeline. So therefore, you know, the cost barriers can be bigger for those developers. It's funny. We were talking recently about new jobs that would be created as a result of AI, mass AI adoption and innovation around AI.

14:22And one of the roles, I'm pretty sure it was on one of the podcasts we did, Carenza, was an AI ethicist or an ethics advisor, an ethics consultant within a corporate organization around AI and language models and things like that. Because we need to be careful of things like bias and just basically slanting outcomes towards unfavorable results. How big is the ethics conversation in your world, Sarah? Are you talking about that all the time? Certainly. So I think UK research and innovation, although we've been investing in sort of, I guess, the responsible AI banner, which ultimately includes ethics for a long time and are continuing to do that.

15:02And I think there's research, I think, that is needed in certain areas. So, you know, how you actually have trustworthy autonomous systems. But I think also there's that kind of practical advice for businesses of, you know, frameworks that can be iterated through. So there's always going to be a trade off with these sorts of things. ethics is a very difficult that there's not going to be agreement from everyone in the world over what's the ethical way of deploying these technologies but i think it's ensuring for businesses it's ensuring that you thought about something in the broadest context and particularly i think it's same wendy hall always says you know if it's not diverse it's not ethical and you know a lot of the teams that are developing these solutions because we you know we know we don't have huge numbers of women working in the AI or the tech space, a lot of the teams aren't hugely diverse.

15:48So you only know what you know. So I think it's anything that like frameworks and practical tools that can help people really think outside of that box and look at problems from a different angle to just know that you kind of iterated through a series of questions that helps you think about it in a slightly broader way. And ultimately, for me, it's also then about being transparent about what you've done and how you've done it and then people can make an informed choice I think that that for me is kind of the the key to this kind of practical approach at the moment is kind of just yeah iterating through a series of questions and thinking more broadly about topics.

16:23When you speak to businesses and they confide in you about their worries whether it's about how to apply thinking around ethics or whether it's about how to actually apply automation in their businesses what are the sorts of worries and concerns that they raise with you when you're going to a lot of events, you're going to a lot of roundtables, you're convening a lot of people. Just interested to see the kind of themes that are bubbling up when they talk to you. The awareness has certainly raised, because I remember when I was doing computer science, there was no ethics module or anything like that, which I think that situation is changing a lot.

16:57So I think there's either kind of just not feeling like they really even know where to start, certainly on maybe more the developer side, or it's I think it's that risk of what am I opening myself up for what am I going to be liable for if I'm implementing these technologies and I don't know how to implement them ethically what does that mean for me and what does that mean for my business and I think that's it's that uncertainty that leads to sort of inaction even if you think actually this could be quite revolutionary for my productivity or it just almost becomes quite paralyzing for businesses I just wanted to ask you about your government advisory stuff and the question I've got for you is how does the UK fail on the global stage in terms of policy and regulation around AI?

17:38The UK has a really long history in AI. And as I said before, we've got, you know, really strong academic heritage. We've got, you know, great innovators in this country. And I think we, for a number of years, the government has been making AI a priority, and I think that's only increasing. So I joined Innovate UK back in 2017, probably at the start of this current wave of interest and investment. So the Paul Pesenti report had just been published, then the sector deal kind of followed up, so a billion pounds to, I guess, take forward some of the recommendations from that sector deal. But we've since had the national AI strategy, which I think is, even though that was pre this kind of generative AI sort of boom that we're seeing now, I think that was a really important document, a 10-year plan to build on our existing strengths and really make the UK a global AI superpower.

18:29And I think it was written in such a way that it actually had that sort of broad framing of the challenges. So even though we've now had this kind of, you know, real generative AI explosion, I think that document is still really valid and really has that good balance between how do we unlock potential and not risk losing the innovation that this technology presents, but alongside kind of balancing with the risks. So I think even though we're seeing more conversation about risk and more conversation about AI safety, for me, I think that was always in the government policy, that clear balance between the two things.

19:04So I think the positioning is, we're positioning ourselves very strongly. There's lots of big investments coming. So obviously there's Bridge AI, 100 million, but there's other research council funding, other, the 100 million power for the child support. So I think actually not only are we seeing the government say they're supporting AI we're actually seeing the money I guess to kind of back that up and position the UK strongly. So if I was to be a cynic then in relation to AI the evolution of AI technology and you must have heard this one you know about people's concerns around AI replacing human workers and causing chaos you know with people being laid off left right and centre and we could be in a position where robots are taking over an enormous amount of the workforce.

19:47And that's obviously going to create societal unrest, potential problems across every sector. Well, what do you say to that, Sarah? What's your position on this point? Any kind of transformative technology or things like the Industrial Revolution. I mean, there could be roles that are replaced. I think for the majority of people, it's probably more likely to be tasks within their roles rather than like entire jobs, although, you know, they may. But I think on the flip side, there are going to be plenty of new jobs that get created. So we already talked about the AI emphasis, you know, prompt engineers.

20:21This time last year, I don't think anyone was employed as a prompt engineer. And now we've got prompt engineers. I looked at a course the other day to become a qualified prompt engineer. It takes 18 months, by the way. I'd love to do that. It'd be a great thing to tell your mates down the pub, wouldn't it, you know? After 18 months, there'll be another three or four more new technologies springing up and taking over. So I mean, yeah, with a different job. The curriculum will be entirely out of date. Yeah, that's true. Yeah. So I guess while I think there is that risk of, you know, displacement of certain tasks and some roles, I think for me, there's always two things, especially that these are technologies that are designed and deployed by humans.

21:01So actually business owners, society at large need to be talking about actually what do we want to see how do we want the future of work to look like? I think that's a really important conversation. But also I think it's really crucial that we have a culture of support so that people can access training, they can access retraining opportunities so that they can take advantage of the roles that get created. That for me is kind of one of the really important facets of BridgeAI is the fact that we've got that kind of upskilling and retraining element so people can, A, they can understand what skill support is available across the whole ecosystem, But also, you know, we've got some leadership components.

21:39We've got some access to training through some of our strategic delivery partners, the Alan Turing Institute, the Hartree Centre Digital Catapult. So actually, when these tools do get adopted, they can be absorbed by the organisations because people understand them, they can work alongside them. So that, yeah, that for me is kind of, I think we need much more focus on education. And a lot of that, by the way that probably falls to employers to actually help people transition into the new roles that get created once these technologies have been adopted. But if we can stick with the theme of education for a minute and teleport you back, Sarah, aged eight at school, just talk us through you being that kind of bright, curious eight-year-old.

22:21and what led you ultimately to become a data scientist and into the world of AI? How did you plot your course? And what happened is just be very interesting, I think, for our listeners just to understand what inspired a little girl to end up moving into this incredible world of technology. I mean, I'd love to say it was planned, but I don't think in all honesty I can say it was a planned part. I've always enjoyed all my subjects at school and I've always had lots of interests. I guess I'm kind of quite a generalist. I just like learning and I like learning new things. I like meeting new people.

22:54So I think when I was young, I wanted to be everything. I wanted to be a vet at one point, which made sense, performer. I wanted to go into PR. At one point, I wanted to be a mechanic, which I don't think I've really ever understood why, because I'm not a petrolhead and I never have been. So I don't know where that one came from. But so it wanted to be so many things. So I think when I was kind of going through the path of your GCSEs and your A-levels, I just picked on things I enjoyed. That was my kind of main strategy. There was never a kind of a career path. And actually, I took more art subjects.

23:26So my A-levels were history, English literature and information technology, which I guess I don't think they really do that anymore. So I'm showing my age, but it was more the essay kind of base side of computing, I guess. I wanted to do a history degree. But when I was doing my A-levels, I just suddenly felt that actually computer science would be quite fun. And the more I kind of sort of started delving into the prospectuses and, you know, what the course content was, the more I got quite excited about it. So I started to explore whether it would be possible for me to do. Obviously, I hadn't set myself up for the greatest success rate because a lot of universities want you to have done maths A level to do or at that point in time, you had to have had a maths A level to do computer science.

24:11But there were some reputable universities that didn't think that. They felt that they could teach you everything you needed within the course. So I ended up at Swansea University doing computer science. And certainly at that time, it was one of the top computer science courses. I think it still is, isn't it, Swansea? You've still got an amazing reputation. I haven't checked the latest, so I didn't want to. Great university, by the sea, lovely. I must have bumped into you in JC's or Divas, Sarah. Sarah knows what I'm talking about now. If you ever went to any gigs at JCs or Divas, then it was probably me.

24:46Yeah, we must have, I must have come. It's not the biggest university in the world, is it? Although it's a lot bigger than it was when we were there. But just for context, we established in the sort of preamble that we were in university in Swansea at the same time. And nearly, Sarah, you nearly did history. So you might even have been in the same class as me, but you missed out, eh? I mean, I think it just goes to show that I think it's a relevant point for AI nowadays, because actually now we're in a point where particularly with AI technologies there's a real focus and emphasis on that kind of multidisciplinary and interdisciplinary and actually you know the government's funded a lot of conversion courses and masters for people that are coming from different disciplines so I guess I think I hope my story can be slightly inspirational to people who think they would like to get into AI and they maybe haven't followed a traditional career path that actually you don't need to necessarily and there are ways to still get involved and actually we don't need everyone to be data scientists we actually also need people from different disciplines who can bring you know other skills to the table i think it's beyond a lot of the other digital areas it is really an area that i think really values and and needs that kind of you know philosophy and arts and there's lots of areas that can can add value to the ai um conversation yeah you're one of the few people i've ever met who is art and tech which my background is similarly squiggly.