Disruptors

Speaker 1 [00:00:01] Hi, it’s John here. Welcome back. This is our seventh season on Disruptors, and we have some exciting insights and big ideas that we think are going to disrupt the market this year. So let’s jump right in. When you think about prediction. What comes to mind? Or what about disruption or innovation or maybe automation? If you answered eye to any of these questions, you’re not alone. This truly was the summer of AI, and there’s a tech wave out there that’s surging. And a lot of us are both excited and worried about what it will mean for our jobs, for the economy, for our communities and, yes, the planet. And while these tech advances have immense potential, we all need to think harder about how they’ll be applied and where the advantages will be. There’s a new think tank at Toronto Metropolitan University called Dais, and they put out a really important study this summer that found only 3.7% of Canadian firms had deployed AI in their business in any capacity — 3.7%. When compared to other advanced economies, Canadian business may be at the back of the pack, but this adoption rate is pretty uneven. It’s been more rapid among big firms, those with more than 100 employees, where 20% said they’re already using AI compared to only 3% of the smallest firms. Adoption has also been leaving some equity seeking groups behind, businesses owned by women, for instance, or indigenous peoples, and people living with disabilities are far less likely than other businesses to currently use AI. So what can Canada do to better innovate and fill these gaps? We know that Canada is a leader in AI science. We have many of the world’s best AI scientists and several of the top universities when it comes to research, we’re a podium nation. But when it comes to application, we’re going to have to step it up and do a lot more to take advantage of this tech revolution. How can we put A.I. to use, whether it’s in business or in health care or education? How can we ensure that those uses are advancing Canada’s prosperity? And in true Canadian fashion, how can we do all this in a way that is fair and ethical? This is Disruptors. An RBC podcast. I’m John Stackhouse. We’re in the early stages of what seems to be a generative AI revolution, and it’s been remarkable to watch its rise and how quickly its evolving. At scale, AI can be integrated into pretty much any organization, and as it’s value to our daily lives and the economy grows. It’s also the topic du jour among regulators as they race to realize its impact, both good and bad. To help us make sense of it all, I’m joined today by two pioneers in Canada’s AI sector. Our first guest, Nick Frost, is the co-founder of Cohere, a Canadian startup that provides natural language processing models to help companies improve human machine interaction. As you’re about to hear, Nick is an optimist. He envisions a world where humans can rely on AI largely like personal assistants, to make life easier. Interacting with it as much as we do with our cell phones. Hey, Nick. Welcome to Disruptors.

Speaker 2 [00:03:30] Hey, thanks so much for having me, John.

Speaker 1 [00:03:31] It’s great to have you on the podcast. I’m so excited to learn more about Cohere because it’s such a fascinating and impressive Canadian company that’s now being noticed around the world. But before we get into the Cohere story, let’s start with some basic definitions of AI. I find it’s a term that everyone loves to say. It’s getting buzz everywhere, but not a lot of people can define it. Can you just give us a quick and dirty definition of AI and also generative AI?

Speaker 2 [00:04:00] Yeah, for sure. So in general, when people talk about AI in a historical context, what they really mean is just computers doing things they didn’t expect computers could do. So in today’s moment, when someone talks about AI, what they’re almost certainly referencing is neural networks, which are a form of machine learning. And specifically, if they’re talking about generative AI, they’re talking about the ability for neural networks to create various forms of media that they thought only people could make. That’s writing sensible text, answering text questions, doing any kind of text based intellectual task. There’s image generative AI increasing, there’s video and audio, but all of those are powered by advances in neural networks and various forms of machine learning.

Speaker 1 [00:04:44] That’s a great, succinct definition. Tell us a bit about Cohere. Maybe start with the origin story. What was the launch of the company?

Speaker 2 [00:04:53] So the launch of the company came out of a realization that Aiden, our CEO, Ivan, and I had several years ago. So Aidan was a coauthor of a paper called Attention is All You Need, which was a machine learning paper which introduced a new type of neural network specifically for language. The realization was that, hey, this tech is really powerful. It can do really great things, but it’s very difficult to create a large language model. It’s even more difficult to make sure that it’s deployed in the right environments, that it actually solves problems for people. And we were at Google at the time and realized that, you know, this incredible technology was not going to be available to the broad public and to every company unless there was a company like ours that set out to make this stuff ready for enterprise and ready to solve business problems.

Speaker 1 [00:05:46] Of course, a lot of people hear AI and they immediately think of the machines taking over that this is going to eliminate jobs, whereas you’re integrating it in things that are already going on and just making them more efficient, more productive, and presumably helping people do more. How do you think through that challenge of technology, both as a disruptor and enabler?

Speaker 2 [00:06:06] I think whenever you have a technology as impactful as this, it will do both those things. I think it’s really important to think about the consequences of your creations as a technologist. These days there’s a lot of talk around existential risks posed by artificial intelligence. I think a lot of that is is kind of misaligned with where the technology is today. And I think our time is better spent thinking on the more immediate things that will be impacted. So that is things like job retraining programs to make sure people know how to be augmented by this technology as opposed to replaced by it means coming up with good policy to make sure that these things are deployed in a way that benefits the entire Canadian populace.

Speaker 1 [00:06:49] Can you give us a couple of examples of how limbs are transforming in a positive way, jobs, companies, activities out there?

Speaker 2 [00:06:58] Yeah, absolutely. One of the things that at Cohere we’ve worked on a lot is something called retrieval augmented generation. So this is where instead of just having an LLM write you a paragraph, you have it look through a whole bunch of documents, write a paragraph based on the information in those documents and cite its sources. This is a real breakthrough because it allows you to like, you know, actually trust the things that are coming out of this LLM because it’s telling you where it got this information. That I think will massively impact people’s ability to do research and synthesize large documents. So, for example, like writing a summary of an article, that’s a really great use case of LLMs. Or answering a question based on like a whole bunch of documents, other things like, you know, predicting how the stock market is going to go tomorrow. The input is not text, the output is not text. And the information that you’d need couldn’t possibly be given to the LLM anyways because there’s things that influence that that are not written down anywhere.

Speaker 1 [00:07:56] That’s a great description. I often think of those moments that we all encounter every day when someone says, let me get back to you. I don’t know. And whether it’s a call centre rep or could be a lawyer or it could be a doctor who says, like, I’ve got to go look something up. Well, yes, saves the time. And then all the energy that goes into that.

Speaker 2 [00:08:15] Absolutely.

Speaker 1 [00:08:16] You talk about Cohere’s global reach. Tell us a bit about how Canada looks to you in terms of applications compared to other markets that you’re in.

Speaker 2 [00:08:27] I would say that Canada is always Canadian. We’re always a little more hesitant to deploy new technology, and that is often to our fate. Like that’s often a good thing for Canada. That often means that things get rolled out and they go slightly better having watched the mistakes of others. But I do find that when talking to people in other parts of the world, they are more willing to like jump on a new technology and deploy it and make slight make mistakes and course correct as they’re going.

Speaker 1 [00:08:55] What sectors do you think need to lean more into this than others?

Speaker 2 [00:08:59] I would love to see all knowledge, work and white collar work lean into this heavier. I think there’s a real opportunity for this to empower people and free up time for us to do the things that we’re really good at and allow LLMs to do the things that they’re really good at. Any work whose input is text and output as text, I would love for them to be working with this.

Speaker 1 [00:09:18] Great way of framing a text in, text out. In technology, there’s often this saying that the second mouse gets the cheese. We know what happens to the first mouse going into the mouse trap. And that’s true of some technologies. But this may be different. There’s also a bit of a frenzy out there, as you as you know, around AI. Some may see it as a bubble. Others may just see it as the growth curve that’s playing out. But I was struck visiting a number of companies in Silicon Valley in the late spring, how many were repositioning themselves as AI firms, but they were effectively just enterprise software firms. How does that affect your thinking as you build out the company? When you see all these large companies well capitalized and have repositioning themselves as AI companies and here you’re a scrappy startup going up against them. Is that a challenge for you as you think about your growth?

Speaker 2 [00:10:14] Yeah, I think undeniably there’s a lot of hype. People are really, really excited about generative AI. A lot of that excitement is warranted based on the impact of this technology. It does really cool stuff. It does stuff we didn’t think computers could do. It continues to surprise me, a person who’s been working on this for many, many years now. But some of the hype is hype. And some of the companies out there who are just selling a traditional SaaS offering that is useful and adding value, recently feel the need to shoehorn generative AI in so that they can, you know, capture a new cycle or capture some investment. We could go back and create a history of the past several decades and name which technology that was happening to in any given year. This year it’s it’s generative AI. I think from our perspective, it’s cool to see people excited, but we’re focused on how this technology delivers value. We’re focused on trying to build something that is useful to people and not just capitalizing on the hype.

Speaker 1 [00:11:14] You mentioned that even you get surprised by some of the progress, what in the last year has surprised you most in terms of what generative AI has been able to do?

Speaker 2 [00:11:24] One of the things that surprised me recently was that our model’s ability to make citations. So it was a few, a year and a half ago or something, and we were first playing around with this retrieval augmented generation. So you give it a document, ask a question, get the answer from the information in that document. And that worked quite well, but it works way better than I thought it did. And now not only can I tell you, hey, this is the answer, it can tell you, and I got it from this paragraph or I got it from this document. And that’s a very complicated problem.

Speaker 1 [00:11:53] Nick, we’ve only scratched the surface of AI’s potential in this conversation. I wonder, as you look to the future and say, think out, five years, where do you think we’ll be?

Speaker 2 [00:12:05] Yeah, I think we will be in a world in which your primary interaction with a computer will be based on language. I also think you won’t think about it very much in the same way that you don’t think about the touch screen and you don’t think about the graphic user interface. I think we’re going to move to that. So I think you will open up a computer, there’ll be a chat box or a microphone for you to speak into. You’ll ask your computer to do things. It will do it for you, and that will be your daily interaction.

Speaker 1 [00:12:37] That sounds very positive. Nick, this has been such a great conversation. We could go on and on, but mindful of your time and we want to thank you for being on Disruptors.

Speaker 2 [00:12:47] Thank you for having me. Really enjoyed it.

Speaker 1 [00:12:50] When we come back, we’ll meet someone who took a career pivot from entertainment law to artificial intelligence and is now at the global forefront.

Speaker 1 [00:13:07] Welcome back. Today, we’re talking about AI and Canada’s standing as a podium nation in AI research and advancement. Our next guest, Jordan Jacobs, is at the forefront of applying AI to pretty much every sector. Jordan is co-founder and managing partner at Radical Ventures in Toronto. It’s now considered the world’s largest AI venture capital firm. He also co-founded the Vector Institute, a world leading AI centre that empowers researchers, businesses and governments to develop and adopt AI responsibly. Jordan also helped author Canada’s first national AI strategy. Jordan, welcome to Disruptors.

Speaker 3 [00:13:46] Thanks for having me.

Speaker 1 [00:13:47] Oh, it’s great to have you in the conversation. I want to start with a bit of your own background because you’ve had a fascinating career. Tell us what lured you into AI?

Speaker 3 [00:13:55] Yeah, I started my career, actually, securities learning financings for tech and media and moved into a group that we built at the firm. I was at doing tech in the entertainment and sports. After a number of years, I left and set up my own firm. But really to do more entrepreneurial stuff that led to building a media company. I made a TV series with Elton John and Elvis Costello called Spectacle that ran around the world. And in the course of that, we partnered with a charity, Bono’s charity, Product Red. And the guy who was running that charity, and I, became very good friends. And we had this idea for basically a new type of social network that would be focused on cultural content. And I’d been reading about deep learning, and I was, you know, thankfully naive enough to not understand it hadn’t really worked in the wild, but it was largely being led out of Toronto by Geoffrey Hinton. One thing leads to another, decide to sell the law practice I had in the media company and go focus on this. My partner quit Product Red where he was the president, and we meet our third partner, Tomi Poutanen. And Tomi had studied with Geoff in the 90s before going and running search in the Valley at into me and Yahoo! And then became the ranking engine of Bing. The big breakthrough in AI wasn’t until 2012 when Geoff and two of his students won a Stanford competition that proved that deep learning was better than other approaches to image understanding. That was basically the moment that caused this boom that’s happened over the last decade. So we started off in our and our first hire was out of Geoff Hinton’s lab, and when we hired him, a lot of friends in tech said, Oh, you do not know what you’re doing. You have no product, you have no data. You do not hire the machine learning Ph.D. first. And our answer was, well, we’re going to architect a system that’s going to produce the signal for him to use. So that was really the genesis of my departure from law and going and doing an AI start up before modern AI worked.

Speaker 1 [00:15:47] Jordan, tell us a bit more about the Radical story. What inspired you to launch what you’ve now described as the world’s biggest AI VC firm?

Speaker 3 [00:15:55] Well, we had built and the AI Company, we had been spun out into a second company called Layer 6. We were ramping up in our sales and during the fundraise, a few things happened. One is we kept having successively higher offers for acquisition from a big tech company. And on the fifth offer, you know we had just again said to them go away without thinking about it. And they came back with a sixth offer. And I turned to my partner Tomi and said, you know Tomi we should talk about why we’re saying no instead of just saying no. So let’s think this through. And we came to a few conclusions. One is we really did believe then that AI would change everything. Second, the pan-Canadian strategy and the Vector Institute had been launched and we’d seen that it was winning talent and it was changing the conditions on the ground in Canada. Third, we realized there was no other VC in the Western world that was focused on AI. And then lastly, something really important happened which ties back to Cohere, which is we read a pre read of the Transformer paper. It Was written in 2017 by a group at Google Brain that included Aidan Gomez who’s now the CEO of Cohere. That paper was basically designing a new architecture for neural nets and we thought, this technology is going to get adopted quickly inside Google. It will probably take another five years to get adopted beyond Google. And then from that moment, it will be a ten year replacement cycle of all the software in the world. So we sold the company at the beginning of 2018, told the buyer, We’re going to go do this. When I left about six months later to launch Radical, we’ve been investing as angels for 70 years. At that point we had a little version of Radical, we were running on the side with our own money and some other individuals, but the first institutional fund was May of 2019, the US $325 million fund. We deployed into 27 companies in the first fund. Raised the new fund. It’s a $550 million fund. We’re now at about a billion US AUM. Our performance has been great, teams amazing. So, it’s been a fun ride. And then of course ChatGPT happens. ChatGPT came out exactly five years and two weeks after we had that prediction that it would take five years to get beyond Google. And what we’re seeing now is this adoption curve being straight up and into that ten year cycle of replacing all the software is happening probably at a faster cadence than we expected.

Speaker 1 [00:18:21] Some people look at that acceleration as a bit of a hype cycle. AI has been through cycles going back to the 70s and 80s. What makes this time different?

Speaker 3 [00:18:31] If you measure it in very short term, there’s too much hype. Medium and long term, I think it’s actually under hyped. People don’t understand how impactful this will be on our lives, on our health, on climate, on basically everything that you can think of that we do out there as humans and all these things we could never do before. So the reason it’s different is it works. What’s really interesting about the technology is all the things that people are not yet paying attention to. Designing molecules for material science to cure diseases. It can be creating new materials that we couldn’t create before or would take years to create that you can create now in a couple of hours. For energy, for aviation, all kinds of other applications. If you are an industrial company, you’re going to have AI that is in your factory assembly line monitoring the equipment and predicting where there’s going to be a breakdown before it happens. It’s going to supplement all of your back office functions for every business, whether it’s accounting software, HR software. So I would say every business is going to be touched by this because even if they’re not deploying it into the core of what their business is, their first thing, be using it in the back office functions because it’s just going to be part of the software suite. Anyone who’s got data, I think can deploy immediately, AI, or certainly over the next few years as the solutions get developed for their business.

Speaker 1 [00:19:49] Where there’s data, there should be AI. That’s a great, great message. Jordan, I want to take you back to 2018, because you said at the time that Canada was uniquely positioned to become a world leader in AI. Do you still believe that?

Speaker 3 [00:20:03] Yes. First of all, a lot of the stuff was invented here or by Canadians. So the research pedigree is elite in the world. And I think that first strategy that was called the pan-Canadian AI strategy, the government adopted in 2017 that we helped author, I think it has worked incredibly well in retaining talent and bringing in new talent. It’s not well recognized in Canada, but when you travel outside Canada, everyone says that strategy worked and that there’s now, I think, 40 plus countries that have emulated it, U.K., France, U.S., China. And I think that is the foundation for the ecosystem. The opportunity to then commercialize and build big companies around it is something that we have to do, it’s happening, but we have to continue to double down and do a better job of.

Speaker 1 [00:20:51] How do we do that at a high level? How do we have that mindset shift in business or in public sector organizations, hospitals, schools and other groups that could benefit from AI?

Speaker 3 [00:21:02] The first thing was we needed to have the practitioners, the researchers, the students you know, the profs, in proximity to businesses so that when businesses realize they need this stuff, they can go hire from down the street instead of having to recruit Google Brain or DeepMind across a continent or an ocean, because that’s really hard. So getting people who are local was important. So that, I would say, is largely solved as long as we continue to support those institutes and the development of that talent. Beyond that, Canadians are historically conservatives in adopting new technology. We’re laggards in spending on research and innovation inside corporations. The problem withAI is it’s very different. A.I. is basically learning software. And so when you deploy it, it starts to learn from your data as a company or your customer experiences, whatever way you’re deploying it. And it’s getting better and better and better. So if you are a company and your competitor is deployed AI and you haven’t, they’re not just pulling away, you know, in the number of months that they are ahead of you. It’s actually a curve and they’re pulling away faster and faster and faster. So how do we get people to adopt faster? You have to show them the return on investment for them doing it and what the benefits are to them in terms of their competition or avoiding disruption. But ideally, you want to show them how it makes them more money or saves them money. What we tend to find is that Canadian companies don’t like to buy until there’s a stamp of approval from the US. We have to change that mindset. We have to get Canadian companies willing to step out a little bit sooner. One party that could do that as a reference customer is the government. In truth, in technology, the US government is very involved and very often is the first buyer, whether it’s through DARPA or just becoming a customer and deploying things at scale. In Canada, we’re terrible at it. So I think the Canadian government being better at buying from Canadian companies. I’m not suggesting they buy stuff that is worse quality. I’m just saying they give it equal shake. Second, I think the government should develop incentives for Canadian companies to buy Canadian technology. So those are just two very quick things that I think can be done. There are others, too. Building capacity in compute. In order to build AI companies now you need access to massive compute. There’s much more demand then supply. The government could step in, for example, and be a bulk buyer on behalf of Canadian businesses. Government doesn’t have to spend in incremental dollars. It’s just fronting and being a guarantor of payments. So having the government as a guarantor of that in a bulk buyer I think gets better pricing and also guaranteed access for a country of Canada’s size. I think there’s there’s an advantage in having a bulk buying power for business and research.

Speaker 1 [00:23:58] Jordan, before we move to close, I want to talk about one more application, which is climate, and that’s going to be a special focus of disruptors this season, including the intersections of AI and climate. I wonder if you can share some insights on where you think AI can play a more constructive role in helping us take on the climate crisis with more urgency.

Speaker 3 [00:24:20] Yeah, I think it has to be a very important part of both addressing existing issues and also help improving things going forward. So we’ve invested in a couple of different satellite companies, one called Nuance Space. The first platform they’re building is for fire prediction and detection, early detection. Basically, they’ll put up a constellation of satellites around the world that’ll be able to detect forest fires before they get out of control. We invested in another company called Climate AI that predicts the weather and climate up to ten years in advance. So their first customers were in the food space, seed growers right through end users of the supply chain who can redistribute their food because they understand it’s going to be drier or wetter in a particular place. We are right in the middle of closing a deal right now with a company that is using AI to determine what existing materials can better do carbon capture at scale. And so that’s an interesting one because it has the chance to actually reverse the effects of climate change instead of just addressing the implications of climate change. So I think we’re going to see more and more of those kinds of companies, the ones that are trying to fix the problem, not just address the problems that already exist.

Speaker 1 [00:25:39] And what a great opportunity for Canada, given the strengths here in both AI and clean tech. So lots more to come on that. Jordan, last questio if we’re lucky enough to have you back on the podcast a year from now, what do you hope will have changed?

Speaker 3 [00:25:55] In Canada itself, I’d love to see faster adoption by companies and by governments of our local world leading AI solutions. There are some companies here that can become absolutely huge global champions. I’d really love to come back and be able to tell you we’re well on our way to a much faster adoption curve by Canadian businesses and government of this technology. I think it would be good all around for the economy. We should be one of the key pillars on earth when people think about where is AI built, based and growing. Canada should be among a couple of places in the world that are thought of that way. And we’re there, but we have to really double down on it.

Speaker 1 [00:26:33] What a great call to action. Passionate but focused. Jordan, thank you so much for being on Disruptors.

Speaker 3 [00:26:39] My pleasure. Thanks for having me.

Speaker 1 [00:26:43] There’s no doubt that this is the age of AI, and when Canada looks into the AI mirror, it’s critical that we do so through a human lens that is responsible and sustainable. The technology has incredible potential, as we’ve heard, to revolutionize health care delivery and diagnosis, transform supply chains and strengthen the pace of learning in schools and of course, improve the productivity in pretty much everything we do. AI can also help us tackle the bigger systemic challenges that we face as a country and as a world. Climate change, lagging economic growth, health care inequality, and so much more. The question isn’t can we do it? It’s, can we do it responsibly? And can we apply it fast enough to bring meaningful change when the world needs it? This season promises to be our best yet. We’ll be speaking with incredible innovators and disruptors with big and bold ideas who are already making waves across Canada and around the world. So be sure to follow us wherever you get your podcasts. And while you’re at it, why not leave us a review? We’d love to hear your thoughts. Until next time. I’m John Stackhouse and this is Disruptors, an RBC podcast. Talk to you soon.

Speaker 3 [00:28:01] Disruptors and RBC Podcast is created by the RBC Thought Leadership Group and does not constitute a recommendation for any organization, product or service. For more Disruptors content, visit RBC.com/disruptors and leave us a five star rating, if you like our show.

This article is intended as general information only and is not to be relied upon as constituting legal, financial or other professional advice. The reader is solely liable for any use of the information contained in this document and Royal Bank of Canada (“RBC”) nor any of its affiliates nor any of their respective directors, officers, employees or agents shall be held responsible for any direct or indirect damages arising from the use of this document by the reader. A professional advisor should be consulted regarding your specific situation. Information presented is believed to be factual and up-to-date but we do not guarantee its accuracy and it should not be regarded as a complete analysis of the subjects discussed. All expressions of opinion reflect the judgment of the authors as of the date of publication and are subject to change. No endorsement of any third parties or their advice, opinions, information, products or services is expressly given or implied by Royal Bank of Canada or any of its affiliates. This document may contain forward-looking statements within the meaning of certain securities laws, which are subject to RBC’s caution regarding forward- looking statements. ESG (including climate) metrics, data and other information contained on this website are or may be based on assumptions, estimates and judgements. For cautionary statements relating to the information on this website, refer to the “Caution regarding forward-looking statements” and the “Important notice regarding this document” sections in our latest climate report or sustainability report, available at: https://www.rbc.com/community-social- impact/reporting-performance/index.html. Except as required by law, none of RBC nor any of its affiliates undertake to update any information in this document.