English transcript:
Grant McDonald:
Today on banking, on AI. We sit down with Kirsti Racine, VP and AI technology lead here at TD, who walks us through her very unique career path as a pioneer in AI and the importance of women leading the field.
Kirsti, thank you so much for being here today. I'm really, really excited about the conversation we're going to have.
You have so much experience in this world, and I want to dive into it in terms of AI tech in general. So, thank you so much for joining us.
Kristi Racine:
No problem. Glad to be here.
Grant McDonald:
I want to start with where you started. You must have a journey that you could tell stories about, you know, breakfast, lunch, dinner over and over and over again. I'd love to get a sense of where it sort of began for you. And we don't have to go all the way to present day. But I really want to get a sense of how you started.
And I know from what I've read is you had quite a unique beginning in terms of what you were trying to accomplish and the, the pushback, I suppose, that you may have faced along the way.
Kristi Racine:
Yeah. Well, my journey started actually ridiculously early in life. My father was a professor of neurosciences at McMaster University. And he was very interested in what AI could do for him. And so as my first summer job, as a 16-year-old, I actually built a neural network for my dad. That tried to predict the onset of epileptic seizures in people and modeled the brain activity that occurred right before, seizure would occur, to see if there was some way that we could one, predict and to prevent.
So I fell in love with AI at that point in time because I realized that a 16-year-old for two months wrote an application that could have a seismic impact on the world. And I just wanted to do that for the rest of my life. So when I went to university, I continued in AI. And this was during the cold winter of AI. I think that's what they call it. Nobody was doing it. I was repeatedly told I would never have a career in artificial intelligence. I should move into a different discipline. But I persevered. And when I went to, university, I went to Simon Fraser University. And we did, we were working with Shaw and Rogers Cable, and we're working on the fact that at that point in time, they were they were deploying out new game systems.
And now it's super simple, right? You get a PlayStation, you hook it up to your TV, away you go. When those systems first came out, you had to have a PhD in electricity to be able to figure out how to put which wire where to get the game to actually play. So the number of calls into the contact centre spectacularly overblown.
I think their calls increased by 300% over a six-month period. It was insanity. So they were trying to figure out what's the best way to arm their agents with information to answer the customer's queries, when you know, the customer has got a TV that's 40 years old. So, you're going to have to buy all sorts of new adapters and all sorts of new cables to be able to wire in this awesome, new gaming system so you could play Legends of Zelda.
We used something called case-based reasoning, which was, which was, a field in artificial intelligence at the time, which was pretty popular, called expert systems. Obviously, it's since been debunked and nobody uses it anymore. But it's based on the idea that the single best predictor of future behavior is previous behavior.
Grant McDonald:
Oh, interesting.
Kristi Racine:
Right. So, if you can build up a case base, and knowledge base of all of the previous things that have happened and how you solve them, then you have a really good chance at being able to solve the next question that comes up into the queue. So we built a, we built a, an application called Case Advisor for contact centre agents to do a deep AI search on unstructured data.
Grant McDonald:
And this is at the time, like, this doesn't really exist. No, but you're just you're, you're creating this.
Kristi Racine:
Absolutely. At the time, the only thing that existed were keyword searches. So you absolutely had to hit key words that were in the content to be able to return the correct result. So we did something that was very novel at the time, but is everywhere now called semantic search, where you, you could search by meaning as opposed to having the explicit keyword, that, that you needed to be able to pull back the correct case.
So it was a ton of fun. And I kind of wish I had stuck with it because, you know, when you see, ChatGPT come in and shake the world seismically and think that could have been me.
Grant McDonald:
So this is revolutionary at its time. Walk me through what's going through your mind, and the mind of your team, as you're trying to, I assume, explain to people who don't know what you're talking about, what you're trying to accomplish. How do you even start that conversation?
Kristi Racine:
It was very difficult. I always made a joke that there were six people in the world that understood the research that we were doing, and unfortunately, there were only six people who cared about the research that we were doing. And it was it was tough to explain that this thing that no one trusted at the time, and nobody believed that it could do anything except for very, very twee, predictions, can have such a seismic impact and could provide such material customer benefit, colleague benefit, to the organization. We had a sponsor, Abaco Epic Scholarships. And they got it. And they were they were so in. So they funded all of our research and they funded the platform that we built. And we did end up selling that to Rogers and Shaw Cable.
We didn't make a lot of money on it, but we did end up selling it. So I felt very proud of myself for making money for the first time in my life, as opposed to the starving student route that I was going down. But, it, it's very difficult. And it historically was always very difficult to explain AI to people, which is why ChatGPT was so amazing for me, because that democratized AI.
It's not AI is doing something, and you don't really understand what it's doing or why it's doing it. You can now materially see how the AI is making the decisions and returning the results to you. As opposed to going, why on earth is Amazon telling me to buy this vacuum cleaner? Like what about my previous purchasing has, has made Amazon select this particular vacuum cleaner?
You can go through the steps and the content and that the AI was iterating over to understand how they got to the actual result that they got to. And 1 billion people use ChatGPT in 11 months. One billion people. Yeah. So every, every single one of those people can now feel, touch, understand AI and understand what it brings to the table and what it can bring to bear for them.
Which makes it really exciting for the dinosaurs like me that have been in this, area for a really long time. And no one understood what we were talking about.
Grant McDonald:
You know, let's, let's, let's dive into that a little bit deeper in terms of your time working in tech and AI. In today's like point of view and the actual stats behind it, you know, a 2023 study came out and showed only like 30% of the AI workforce is women. Again, 2023 more recently, it may have changed over the past couple of years, but that's the stat that we have. So, let's go back to when you first pushed through and were leading and pioneering in this world. It must have looked even more different then.
Kristi Racine:
Absolutely.
Grant McDonald:
So I'm slowly talk me through some of those challenges where you're not only explaining to people that may not understand the general sense of what you're talking about, but it's also coming from someone who there may not it may not be used to working with someone like yourself.
Kristi Racine:
Yeah, I remember one time my boss told me not to sit next to him with a pad of paper and a pen because people would think I was his secretary. You know, what? Okay. When I when I started university and in my class. And at Simon Fraser, I was the only female in that class. In my undergraduate degree in computer science.
Out of a class of 84, there was three females. So I, I'm kind of used to now being the only woman at the table. But I'm terrified of that 30% stat in AI because, I hate the stat in STEM as well. And I'm a big advocate of, of we have to get into the high schools and we have to get to girls early.
And we have to do a much better job of marketing what, what an IT career looks like. Because there's this sense that it's this isolated thing and it's really hands on keyboard, and you're doing a lot of things all by yourself. And it's very independent. And it's not. It's very collaborative.
There's a ton of communication that you need to do. There's a ton of stakeholders that you need to partner with. And I think we do a very poor job of sort of emphasizing the camaraderie that, that we have to build to be able to build any IT application in this world. The reason that it terrifies me, and AI adds yet another piece to that puzzle is if you think about model development, a model is going to reflect the people who built it.
And I'll give you an example of what I mean. We went to go see this this firm in Toronto, and they do a lot of content creation. And we went into this room and they had, you know, all of these amazing AI generated images on the walls, like walls were surrounded by AI-generated images. Every single person in those images was a white man.
Grant McDonald:
They're, you know, and they're pulling based on how they've been trained is what I'm assuming.
Kristi Racine:
You got it. So not only do we only have 30% of, you know, the workforce in AI is female. But unfortunately, if you look at the usage of AI as well that also is a very, very, unfortunate statistic. I think Harvard Business Review just released a study and 27% of the folks that have downloaded GPT to their phones are female. 27%.
If you think about how these models are trained and how they learn, a lot of the way that they learn is how we use it, i.e. how we prompt it. But only 27% of those prompts are coming from females. So the bias is, I believe, getting compounded within the models. And it's something that, it's, it's very alarming.
So we have to do a really, really good job of getting to kids early and making AI education mandatory for everyone.
Grant McDonald:
That's a really interesting perspective on that. Mandatory training on it because the reality is just like taking English or math or science, that you have to take it to a certain degree to understand it. I'd love to focus on that sort of the barriers that you've already described. What other barriers are you seeing for women in tech?
But then let's take it back a little bit to what you just discussed in terms of students and girls in tech. How did we get around that? Is it just getting there earlier? Is it a combination of different workforce policy, like what goes through your mind when you're trying to solve for that? Not an easy solve.
Kristi Racine:
Yeah, it's definitely not an easy solve. If I if I could have solved it, I would have solved it, you know, many years ago. But, I do believe that mandatory education early is imperative. And I honestly also believe that we have to do something to instill confidence in women in general. I'll give you an example.
There's a number of studies, out there that show that women will self-select out of job postings, for example, you know, and, and we do it. We put out these job postings and it's a wish list. Here's 20 skills that my ideal candidate has.
Grant McDonald:
We hope you have it, right? We're going to get a couple.
Kristi Racine:
And in general, males will look at that and go, oh, I've done three of those things before I'm in, I'll apply. But women will be like, oh, I only have 19 out of 20, right.
Grant McDonald:
Right. They will full stop.
Kristi Racine:
They'll select out. Yeah. So we've actually changed the way that we write job postings to, to make sure that it's more inclusive and that we get more women actually applying. But it is a confidence game. I was doing a presentation for, for women in tech earlier this year, and there was a fellow from Google there, a male ally.
And he raised a really interesting point. He said, I have never had a bad interview with a woman. And first I was like, yeah, you go, girls. Right? Like awesome. And then I thought about it and I was like, actually, that's bad. Because that means no woman took a chance and went in and applied for a job that they really weren't qualified at.
The only women that he is, he is talking to, are women who have either already been doing the job or are exceptionally well qualified to do the job.
Grant McDonald:
They have the 20 out of 20. That's who you'll hear, right?
Kristi Racine:
And those are the only people he's hearing from. And I don't know how we instill that sense of confidence, but I think education very, very early on is super important. I think making sure that we drive and build AI applications that are not gender specific is also very critically important because you want to build something that females are going to want to use, right?
And you want them to feel safe about using them. So that that those are things that I think about a lot. I do, occasional, lectures at, at various universities. I also think that's too late, but I want to get in the door, just to show people that one, you can have personality in and have a job in IT.
I think that's important, to, you know, it's not a lonely career. It, it's a career just like any other where you make friends, you work with people. You do not spend your time in a basement or a garage alone on a keyboard. And I think that's really important to get that message out there.
Grant McDonald:
You know, you're, you're such a unique example of, like, what you were doing at 16. I don't think a lot of people were doing that at 16. No. And so looking at that story, I'd love you to, you know, walk me through after that point, mostly because I've heard stories of this really unique background that you have where you're trying to make, like, punch cards at a certain things and you're trying to like, I think you did a lot of fooling software into doing things that it didn't know it could do.
Walk me through a little bit about that. And then where we are sort of present day and where we're kind of going with it and how women can be at the forefront of it.
Kristi Racine:
Yeah, I, I okay, so my mother has a degree in computer science, and as I said, my dad's a neuroscience professor. So, I would agree that I did not have sort of the traditional upbringing, that, that most other people have. You know, back in the day where computers that were very, very new and nobody had them, we had like five in our house, you know, and my, my dad's like, hey, you want to take this computer apart with me? And I'm like, awesome.
And I recognize that that's, you know, sort of a unique experience. And you used to have to really know how a computer worked to get it to do the things you wanted it to do.
I always make the joke that my master's thesis was 6 million lines of code, and it can now be done in six lines of code and TensorFlow. To run the model to, to to run the model training, took 16 hours. I think it would take about six milliseconds right now. The computing power has changed.
And the, the thing that generative AI brought to the table is now you can describe things in natural language and actually generate code. And there's a lot of schools of thought about vibe coding and, you know, is it good, is it bad, can it take over? And, no, I, I believe software engineering is still a very, very critical discipline, but it certainly allows you to get started a lot easier than, than going and writing 6 million lines of code to do a very, very simple thing.
So, I think that's another thing that we should be promoting is there's an easy access path now. Before you had to be deep into math, you had to be deep into science. You had to have a very methodical brain. Now you don't, and you can be very expressive and build AI applications or IT applications in general that do magnificent things.
Grant McDonald:
And so people who are thinking of maybe getting into it, maybe they already work at TD, maybe they're curious about what it looks like in the world of a giant bank. Mentorship is such a huge part of the culture here. And so I always, I'm curious for individuals like yourself who have such a wealth of knowledge, not only of the industry itself, but just how we got here.
And then also just experience working in the world that we currently live in. What's your approach to, you know, mentorship and helping women along the way if they're interested in this field or just curious?
Kristi Racine:
Well, I do a lot of work. I have a lot of meetings. I have a lot of coffee chats. I don't think I've ever had this many coffee chats in my entire career. All of a sudden, all of a sudden, AI is cool again, and people want to have coffee chats with me.
There are a number of women in technology mentoring circles at TD and I think there's about 350 women that are in that circle. So, we will be going out to them with very specific education. Hey, I'm here to answer your questions. Hey, I'm here to do anything with you.
My first recommendation is to most women who come and talk to me is when you see a job that you think you want, I recommend you apply for that job because you get 0% of the jobs you did not apply for. Because I've had so many women come up to me afterwards and they're like, I saw that job posting, I was really I was super interested. And I'm like, oh, did you apply? No. You're kind of handcuffing me here. Because applying for, you know, and this has happened to me a number of times in my career is I've had people come in and apply for a job, and they weren't the right fit for that particular job.
But you click with them, or you think, oh my goodness, you know what? You would be fantastic at this other job over here. Can we talk about that opportunity? Or you tell Bobby that, oh, I just interviewed this fantastic woman. I think she'd be perfect on your team. I'm going to introduce you guys. And extending that network is so, so critical, right?
Getting that breadth of experiences and that breadth of people who know who you are and know who your brand is, is very important. So get yourself out there, reach out to people. I have two women who work directly for me. They also have a lot of coffee chats, and they also do a lot of talking on, on various panels.
But we're here and we're willing to, to have a chat with you and talk to you and, and help you out with your journey. Another thing that we're trying to work on, for everyone at TD, not just women, but to make it mandatory is AI education. Here are the basics I don't care if you're in tech or not.
You need to understand this because it is going to inform almost every single job in the future. And the people that learn how to work with AI, I believe, are going to be more successful than the people who refuse to work with AI.
Grant McDonald:
Those are very good insights. And I you know, that is the reality that we're in. So excellent advice on the mentorship side. Really good. Just for the overall mindset of PD colleagues and knowing like you should, you know, try to understand some of this stuff. I always love to get a sense of, from individuals like yourself. People ask you all the time what you do, what does the future look like?
I know that's not a question that's easy to answer, but sort of at a, you know, at a dinner party when people are saying, tell me something interesting about AI that I would not even have the slightest clue about. What's one story they always like go into where you're like, I know the exact moment where their mind I can see their mind being blown as I tell this story, or what the potential for AI could be.
I'd love just to get a little walk through, but we'll keep it like realistic in terms of you without looking at a crystal ball.
Kristi Racine:
It was funny when I when I was working at IBM, I was a, I was an architect and we used to do these like ten-year strategies. And you know, then then we're like, okay, the world of IT is moving really quickly. So we did five-year strategies. Now I'm reluctant to even do a five-minute strategy.
This world is moving so quickly. I don't know if I have one thing that I say to people that that will blow their minds, but, there's one example. Like, my friend's 83 -year-old mother is using ChatGPT to create her grocery lists, to design her exercise routines, to recommend the next book that she reads. To create her schedule for the week to ensure that she's not too busy, but that she has an interaction with a person every single day.
So she, she just continually piles more and more things on this, and she's literally got ChatGPT running her life. That's my friend's 83-year-old mother. If she can figure out a way to make AI this comfortable for her, then surely the rest of us can do so as well, right? And it's amazing. She's like, I'm eating so much healthier now, and my grocery bill has gone down because I ask AI you know, what's the best value for the dollar?
And it tells me, well, you shouldn't buy that 99-cent can. You should buy this 69-cent can over here. Yeah. She's literally now planning her route around going to the grocery store to get the best deal on tomato sauce. But, you know, she's getting exercise while doing so, so kudos to her. Right. And then my nine-year-old nephew is interested in coding.
I think I told you this story already, but I was like, oh, this will be perfect. I'm going to use, I'm going to use DeepSeek because I've never used it before. Just to, to generate a game for him. And I was like, I grew up playing Tetris. Like, that was the coolest game ever.
So we're, we're going to I'm going to show him how to code Tetris. So I put in literally one prompt into DeepSeek and it generated the whole game. And then it goes, hey, would you like me to install this for you? And I'm like, yes. And it's like, hey, would you like me to run this for you?
And I'm like, sure? So we had an awesome game of Tetris, but my nephew learned nothing. But I, you know, again, that's something that required millions and millions of lines of code. And now you can generate that game for yourself, using natural language.
Grant McDonald:
You've seen like, such a rollercoaster of information and changing and just there's just so many things coming all at once. I really like that story of 83, nine-year-olds. It doesn't matter the age. And it goes back to your point in terms of saying, now we can actually see what this is like. It makes sense to us.
So a grocery list makes sense. I love that story. There's a lot of unknowns for the future. And there are a lot of really interesting things happening now. But what really excites you, I guess now, and potentially for the future, when it comes to this world that you're helping lead.
Kristi Racine:
Like I'm super excited that TD's embraced AI so holistically. We have every single line of business pursuing AI use cases. And then future, future, I'm really excited to see what happens with physical. And we're going to get there. Meta is coming out with these glasses that will auto translate for you. So you can go to Italy and have a conversation where one person is talking in Italian and you're responding in English. I think the, the world of physical is, is where we're going next, and there's a ton of research and it's very promising. And it'll be fascinating to see what happens.
Grant McDonald:
Kirsti, thank you so much for sitting down with us today, giving all the insights that you have, walking us through your very unique career path. It's been an absolute pleasure.
Kristi Racine:
Thank you very much for having me.
Want to learn more about TD podcast?