Thank you, Kate. Welcome everyone. Really appreciate you joining us. I am very pleased today to be able to be joined by Aidan, who I believe is late in the evening where you are. So thanks for doing this. Aidan is the co-founder and CEO of Cohere. As some of you will know, as an intern at Google Brain under Geoffrey Hinton and Jeff Dean and others, you can tell us a bit about that history. He was involved in and a co-author of the paper 'Attention Is All You Need,' which kicked off the modern era of AI and formed the architecture of generative AI. That was back in 2017, way back when you were a teenager. Subsequent to that, in 2019, Aidan co-founded Cohere with Nick Frosst and Ivan Zhang. We were lucky enough to participate in helping them to launch the company and writing the first check. So we could talk a little bit about that history. Most recently, Cohere, it was announced, is going to combine with Aleph Alpha or acquire Aleph Alpha, a German LLM company, which will make Cohere a multi-continental business, both Canadian and German. But the business is actually global, and we'll talk a little bit about that and what that means in a world where sovereignty really matters. Cohere has grown incredibly quickly, revenues in the hundreds of millions of dollars, team has grown into the many hundreds, and we can talk about that. Aidan also has a Bachelor of Science in Computer Science from the University of Toronto, has a PhD from the University of Oxford, which he managed to do on the side while running Cohere. So for those of you who feel like life is overwhelming when you're in school, imagine doing that plus running a giant global company. He's also on the board of directors of Rivian, the electric car company, which is a very cool side job because he doesn't have enough else to do. Aidan spoke in our masterclass in 2022. So Aidan, thank you for coming back, and welcome.
Yeah, no, thank you for having me back. I think last time I told people to fire people much more, and I don't know what kind of impression I left, but thank you for inviting me back.
Yeah, no, it's okay. We've all forgotten that. And hopefully, you know, everyone's in a sunnier mood, but let's jump into it. Okay, you co-authored 'Attention Is All You Need' as an intern. Why did you — and you were working at Google — you were working at Google in the Bay Area, you were working at Google in Toronto. What made you leave and decide to start Cohere?
Yeah, I think it was a really unique period because we had gone from a world where machines that could write compelling language were totally out of reach — like we could barely get them to spell a sentence properly, and certainly none of it made any sense — to almost overnight one where there was pretty much entirely plausible human text that appeared in front of us. And I still remember getting this email from Łukasz Kaiser, who was my manager at Google. The subject was like, 'Aidan, look at this,' and the body was a Wikipedia page. By the way, sorry, I'm sick right now, so if I'm clearing my throat or coughing, I apologize. Yeah, the body of the email was like a Wikipedia page, and the title of the Wikipedia article was 'The Transformer.' And then it talked about this like Japanese punk rock band that had gotten together, done a few albums, broke up, the bassist got replaced, they got back together, did a tour. And then at the bottom of this email, after this Wikipedia article on this Japanese punk rock band called The Transformer, Łukasz had written, 'I just wrote the Transformer, and the machine wrote the rest.' And that was a really jaw-dropping moment. Like now it's so quaint, right? Like of course, you know, you could open up one of 20 different apps and prompt that, and obviously it will do it better, right? But back then it was like getting hit in the face with something that was out of your dreams, right? Something that was out of like a movie. Like you just — the night before you couldn't even — by the end of my career, if we had machines that could do that, it would have been an extraordinarily productive life in machine learning, you know, what a contribution to humanity. And it just showed up in front of us, and so it was obvious that this was going to be impactful. Like machines had sort of woken up and could talk back to us, and so that opened up a totally new interface to interacting with machinery, computers. I didn't know how to make money off of that. I didn't know how to build a business around that, but I had a pretty good idea of how to do the tech, how to do the modeling, right? Like I knew how these models were built. The trend that seemed to be working was scaling them up, and that's what I was doing my PhD on. And so then when I went home for my birthday, I came back home to Toronto, was like, you know, hanging out with friends, was mostly there to party. But then I was going to give a talk at TD in Layer 6, because that was like one of Toronto's primary ML groups that was in industry, and I wanted to give a talk on my research to see if they were interested in it, so that they might adopt some of this like efficient scaling of neural net techniques that I was developing. Presented that at Layer 6, met Tommy. Tommy told me, you know, you know who you have to speak to, Jordan. And the next morning, Jordan and I met. And Jordan was like, 'Do it. Let's figure it out.' And then later that day, I went to hang out with Nick and Ivan, who are now my co-founders, and I said, like, I had this crazy meeting with this guy, you know, Jordan Jacobs. Like, he thinks this might work. He thinks this technology might work. Should we do it? And you know, it took some convincing. It took a few months of work, I think, both on my part and Jordan's part, but eventually we were like, 'All right, let's quit Google and do this.'
And so, yeah. And just to put it in reference, like we're talking about, you know, almost eight years ago.
More than seven, seven and a half years ago.
At this point, it would be like seven and a half. Yeah. And so, and just so people understand the context, that's like three and a half years before ChatGPT. So it is early, right? There's no such thing as Anthropic. There is OpenAI, but they've not put a product out that anyone's really using at scale.
They were doing the robotic hand solving a Rubik's cube, and they were doing Dota. Like their two bets.
Exactly. The robotic hand became Covariant, which we invested in, the robotics company separately. Okay, so you guys decide to leave to do it, and it's true, Aidan was like all in early, and there required some convincing. I would say Nick was the most reticent. Nick's like, I have the most incredible job in the world. I worked for Geoff Hinton at Google Brain, and you know, what could be more amazing than this? So I remember some of those conversations. It was fun to try and convince him there was even brighter things ahead and bigger things ahead. In those conversations, there's a couple of things that were considerations that were — you made decisions that I think were very consequential. One was we're going to focus on enterprise, not on consumer. And a second was we're going to do this in Toronto, when you had been working in Mountain View. There's possibility of doing it in London. Talk us through the decision of those two things. Why focus on enterprise, and why do it in Toronto?
So enterprise, that was more because we didn't know the product to build. We started — like Jordan just said — well before there was any product-market fit. No one was buying large language models. No one was really buying models from a third party. They would build stuff internally inside their company, and it might be like rerankers for social media feeds, or like recommender systems basically. That was like the predominant actual industry deployment of these models. Or Google would build models for itself, for Translate, that type of thing. And so there was no market. It was just like a totally — and I remember some of these, like when we were starting to build this, like how do you price it? How do you charge for this thing? And like, you know, the unit of compute is a token, but no one's going to understand what a token is. Should we charge based on characters? Should we charge, you know, like words or something? Like have some completely different pricing model? There was no market, and so the initial bet was we don't know what product to build, and so we're going to build a platform so that when someone figures out the right product to build, we'll be the place that they go. And that pointed us towards companies, right, and serving B2B. It was — yeah, I think it was like the only — I think if we had gone very deep into consumer and we had made some like very hardcore product bet at that stage of the technology, we wouldn't be around right now. We wouldn't be having this conversation. I don't know if that's transferable advice. You know, I think a lot of people build in spaces where actually you should not build the most general thing, and that's not the fastest path to revenue, and that's not the best strategy. But if you're building on top of like an extremely new technology that has no users and no market, that's probably kind of what you're forced to do. And if you bet the whole farm on one specific product, you're going to die with high likelihood. Maybe you'll, you know, get lucky. On being like very global from the outset, I think that was inspired by advice from you again, Jordan, which we were like, you know, the default advice: go to the Valley. Like you're building a tech startup, do it in Silicon Valley. And the argument Jordan made was in Toronto, you're well networked and you will have a monopoly on talent. Like you'll be the only game in town, and people want to stay where they live. They don't want to have to uproot their lives and move. And so you will be very unique in that regard. So don't just start the company in the Bay Area. Start the company in the Bay Area, but also have an office in Toronto. And then that — well, we founded, and then COVID happened, and then we were just like, okay, we might as well go all in, and we'll just hire wherever the talent is. So we hired a guy in Singapore, we hired a guy in, you know, London. We hired all over the world. And that led to a very global orientation for the company. Like we were really good at multilingual and still hold like state-of-the-art in translation, very, very early. And so it paid dividends to the company at the technology level as well, not just the talent level. There were real costs. Like I hate remote working, and I have like the most remote company, and my direct leadership team is like in Miami, SF, Montreal, Toronto, London. I think that might be it. But like that's like just my reports are all over the place, and it's super, super hard. If I could like magically wave a wand and put us all in one city in one room, totally do that. It would be very good for productivity, potentially bad for our strategy, which has sort of fallen out of that global footprint and is serving the world. And as a product of us having so many investments in so many different places, we're sort of locals everywhere. Like I'm in London, my co-founders are in Toronto, you know, we have a HQ now in Berlin, offices in Seoul, Tokyo, Riyadh, like all over the place. We're sort of like local to the world. And for sovereign AI, that's an extremely good property to have. It's very, very salient because the country sees your — and the local players see your commitment to their ecosystem, their talent. And you're not showing up there just to extract money from them. You've been there for years without asking anything from them. And so I think it's a high-trust position, but it's a huge cost, huge price to pay.
Let's talk about sovereign AI a little bit. I mean, this was something that came up very early in the discussions about Cohere, but again, there was no market, and I don't think that anyone really understood what we were talking about when we talked about sovereign AI at that point. Now everybody does. Can you describe what it means? Like when people talk about sovereign AI, what are they talking about? How have you guys leaned into it, and what has the market opportunity become recently?
Yeah. I mean, what they mean is that no one can switch off their access to AI. They want unrestricted — they want control, fundamentally control over where it runs, control over who can see the data running through it. And then very importantly, they want operational continuity. Like in a crisis, it will not get shut off. And what they don't want is like the Mistral scenario where you have built a bunch of stuff on top of this model, and then one day you get rug-pulled and it's gone, and your infrastructure starts breaking down. I mean, it's just like it's very plainly and obviously unacceptable. Sovereignty isn't about like isolation. It's not like you have to do everything. It's not that every single country in the world is going to build its own model. I think that's a bit of a naive perspective given the amount of capital that is required to do that. But what it does mean is that you're going to have to diversify your supply chain. So you're not going to completely depend on any one party. You're going to have multiple countries, companies, etc., that you buy your AI from, and no one party can shut you off. That resilience through diversification is what people are really talking about.
And talk a little bit about what that has meant for you guys, particularly since Mistral was shut off, which feels like a long time ago, but it was only about three months ago, three and a half months ago. So did the world change after that? I mean, certainly it felt like there was an awareness of this issue that became ubiquitous, whereas before, some people maybe it was a theory or, you know, a concern in the background. And just to describe what happened: the US government shut off access for all non-Americans, and that means non-American companies, non-American governments, non-American employees of American companies even. And as a result, you had giant organizations around the world not getting access to something they knew that their competitors did have access to, and competitors whether you're a government or whether you're a bank or, you know, any other organization. So okay, so that happens. How has the world changed for Cohere since then?
I mean, yeah, it's like you say, it was a risk. It was something that was talked about and like by many parties sort of poo-pooed. Like they would never, you know, like we're such good friends with the States, like they would never do this to us. That's an unthinkable reality. But come on, we're in like Five Eyes together, you know, like we share everything.
Five Eyes being the intelligence sharing among Canada, UK, Australia, New Zealand, US.
Yeah. The most tightly coordinated set of allies of the United States being the intelligence sharing Five Eyes coordination.
Yeah. So there's like just so much deep integration, like this could never happen. And we were like still trying to make the argument. We're like, well, okay, you know, you thought free trade was a thing and it like vanished overnight. And ally and foe alike, same treatment. Like, you know, there's a — the risk is like a real risk. Like this seriously might happen, and you should prepare for it. But still folks were like, yeah, like maybe we can do some like little stuff, but not taking it seriously. And then Mistral hit, and then it was like, this is real. It not only could happen, it has happened. And we will plan for it to happen again. And so it was just like a big change in the world. It was sort of like a one-way door, like crossing the Rubicon. It was like we now exist in a different world where all these assumptions that we held true, we will never assume they hold true forever after. And in net, I think, you know, it's probably good for the world. Like it was a very fragile state that we were in. We had built up this system where we had a single point of failure. Everyone depended on only one other country to supply them all of their technology, and that's dangerous for democracy, right? Like, you know, systems engineering 101: don't build a single point of failure. And we used to not do that, right? Like there's multiple countries inside the G7 or like the democratic bloc that develop fighter jets, that develop missile tech, that develop energy supply systems, nuclear reactors, that have nukes, like bombs, right? Like we distributed this technology across the bloc of democracies. But then when you look back over the digital revolution, you look at like consumer devices, social media, major hyperscale clouds, you know, they're all from one country, like just 100%. And so there's like this line when digital tech started taking off where you just see this massive concentration risk emerge. And I really do believe that contributes to a lot of the weakening of democracy that we see in the world today, and the fragility that I think we all feel and see in the democratic bloc. So I'm personally like a big believer in probably one of the most important things we can do over the next decade is to ensure that multiple democracies develop the core capabilities, the core levers of economic, social, you know, national security technology. Otherwise we're just gonna — like we're not doing ourselves any favors. We're not doing the US any favors. We're not doing democracy any favors by persisting in this world where we engineer in fragility.
And so that kind of leads into how you've taken action. One of those ways is to combine with Aleph Alpha. Talk us through the strategic thinking behind that. You know, what does that mean for you now in Europe, and how you're perceived by customers, whether they're regulated industries, governments, etc.
Yes. So we're dual citizen essentially. So we're Canadian and we're German. We're European for all intents and purposes. Like we fit the definition of a European company under the EU AI Act. We have a headquarters in Berlin and one in Toronto. We have board representation from Germany, from Canada. We have board meetings in Germany and Canada. So at the governance level, at the operational level, etc., etc., etc., we are German-Canadian.
Or Canadian-German. I mean, obviously I don't have a German passport, so I'm a little bit biased, but as the CEO of Cohere, I'm not, right? Like I'm equally committed to the success of these two nations and protecting their interests and their sovereignty. And that opens up a massive market. The European market is immense. You know, it's fractured. It's like disorganized and a little bit like chaotic. And there are plenty of critiques of the EU. But it is an incredibly powerful and large market. And you can see it starting to unify, to act, to prioritize sovereignty. I mean, they literally have like a hot kinetic war on their eastern border. So they have to — like there's just — it's like marching towards them. And so they don't have a choice at this stage. They are going to have to defend themselves. People are already trying to impose their will and encroach and harass and attack them. And so you can see there's a real need for sovereignty, and it's a real privilege to be able to help them with that and to take that on. My family comes from Europe. I, you know, I'm the first one that was born in Canada. My mom's British. My dad is Spanish. Both of my families live in the UK and in Spain. My wife's sister lives in Berlin. Like protecting Europe is good for everyone. It's good for the States. It's good for Canada. It's good for anyone who believes in democracy and thinks that's a better way of life than despotic regimes and, you know, authoritarianism. So I really do think it's like such a crucial priority. Sorry, I forgot your question. I started rambling.
Oh no, you answered the question. Why combine with Aleph Alpha and what has it meant for you? So that was actually a great explanation.
See, by accident you got there. So the people that are here are people who are in AI thinking about becoming founders or on the journey of becoming founders. Just stepping back, kind of looking, putting your journey in perspective. What are — just give some pieces of advice about what people should be thinking about, how they should be approaching it. You know, how big should their ideas be? Should they be thinking about something small and very clear or something very big? Should they be, you know, massively ambitious or not? Should they be looking for co-founders? Like give us some of the nuggets of things that you've kind of learned along the way that maybe happened accidentally or maybe intentionally but have been beneficial, or things that you did wrong that you would have done differently. And you know, it's a grab bag. Just hit us with anything that you think of.
I got so lucky with my co-founders. Like I don't know how I got so lucky. But I've had lots of friends who have started companies and like the wrong co-founding unit is so painful. I don't know. I would recommend solo founding instead of founding with the wrong co-founding unit. So be very, very thoughtful about who you choose. The way that Nick and I went through it, we basically viewed it as a marriage. You know, like it's just like this is for life. Like our destinies are permanently intertwined. Our lives will be inseparable. We can't run away from each other. There's like, you know, it's — this is a permanent, irreversible commitment, and we are essentially becoming siblings. And we should approach it as siblings. Like we're going to fight, we're going to have disagreements, but you can't escape each other. Like you're in this for life. And you're going to have to work around the weaknesses of the other person, and like — but you still love each other, you know, and like you want the best for each other and the best for you both jointly. So I think that's how we've approached it, and that's what's given us the staying power and the stability in the company. So yeah, I don't know how to give advice around picking founders. I think I just got like super lucky.
The unit of three is interesting. So in startups, you know, as a founder, two of the companies I started — three, both AI companies — we were three in each case. Three is interesting. It's different than one. It's different than two. I mean, it's got its pros and cons. How have you found three, like having three of you, to be beneficial? Like what is unique about that configuration?
I don't have a counterexample, you know, because I don't know what it's like to be four or two or one or —
Well, here, I'll give you an example that I found in my experience. Doing a startup is hard. It's a roller coaster. You know, 10 a.m. you're going up the roller coaster. Yay. 2 p.m. you're upside down and vomiting out lunch. And it's just kind of like this all day, every day. The benefit of having co-founders is that you can share the experience with each other. You know, there's a shorthand and there's someone to vent to. Frankly, it's particularly when you're dividing up responsibilities. You're not doing the same things as each other, but you're familiar enough with what each other are doing. And there's this —
opportunity to share that experience and kind of vent and get it out. So you don't necessarily take it home with you to the same degree you otherwise might, and you're not dumping it on your spouse or your partner, you know, who's already suffering from being with someone who's a startup founder and is married to their startup as much as they are to them.
Having three is interesting because there's always someone different to talk to. You know, if your co-founder is having an even worse time than you and you don't want to go dump on them, well, maybe there's a second person you can go talk to who has a different perspective.
That's true. And if one is giving you a hard time and you're like really upset about it, you can go commiserate with the other sibling. You know, you're like, 'Nick's being so mean to me.' And they'll comfort you and then, you know, be like, 'Oh, but you know, you should hear him out,' blah, blah, blah. [laughter] It's a good number. But I think that what you just said is exactly the dynamic that is really important, which is if it's just two of you, you can get into a kind of standoff. And there's not a third person who can mediate. And if there's good intentions between the three of you, there's always a mediator. If people are in a disagreement, and the disagreement can be about anything. It can be about how to spend the money. It can be about who to hire. It can be about responsibilities. It can be about, you know, what to prioritize in your roadmap, in your product roadmap. It can be about anything. Who's going to get on the next plane, you know, after each of you have been on a plane a million times. So, having a third person to mediate can be super helpful.
Yeah, it's true. Like the opportunity for a devastating conflict between two people versus three is probably way higher. Like as soon as you get any conflict between the two, it's a crisis to the entire founding unit. Whereas if you have more than two, you can have these kind of bilateral conflicts and still see it through.
Yeah. Exactly. So I, as you know, I understand your advocacy for, you know, go start on your own rather than start...
Oh, I'm not advocating for that. I'm just saying rather than a bad co-founding group.
But I really do think actually the most powerful unit of founders is three. And if you can luck out like you guys did where you have this common understanding from the beginning that this is going to be siblings. It's going to be really tough some days, but we love each other in the end. And we're in it together for good. I think it's the most powerful configuration of founders. Okay. And there's lots of examples of, like by the way, with you guys that I've witnessed of that, you know, just third all the time. We fight all the time, but we always get through it.
Okay. We have, I think, a lot of questions in the chat. For those who are interested in asking questions, just post your question in the chat. We'll scan through. We're not going to get to every one of them. But I want to ask some of the questions that are here. And let me start with one that I saw. One question. I want to make sure I'm pronouncing the name correctly. Hopefully I'm getting it. VCO asks: Aiden, in the beginning, did you consider focusing on one industry vertical first or staying across industry part of the strategy?
Yeah, we definitely considered zooming in on sectors or verticals. It was at the beginning hard to pick one for the same reason it was hard to pick any product. It was like, where is this going to take off? No clue, no idea. So we stayed very, very general. We just viewed ourselves as, like what we said back then was just, we're building infrastructure, like we're building infrastructure for talking to computers. And we don't know what you're going to use this for. We're like a steel producer. Even today we have these debates about like what industry is the right one for us to start specializing the product into. And others have found good success going into like legal, like Harvey, or obviously code as an application domain has been probably the most successful for LLMs. I think at this stage of the technology and maybe like for the past three, four years you can totally specialize in a vertical, but for us it was too early to do that.
Another question from Jesse: Beyond Silicon Valley, where do you see the most significant AI talent hotspots globally?
I'm biased, but I would say Toronto and Montreal have like an insane concentration of talent, principally because of U of T, Waterloo, McGill, and Mila. They're just talent aggregators, like they're the training houses for a lot of the best ML talent. And then the other thing that I saw recently, the UK as well, like you have Oxford, Cambridge, UCL, Imperial. I would say those four that just pump out generations of incredible computer scientists led to DeepMind, you know, all this type of thing. A lot of DeepMind folks became profs or were profs in the UK. And so it just incubates a ton of ML talent. In Europe, I think Paris is super strong. Like École Polytechnique is like incredibly good. Yeah, those are probably like the top hubs outside of SF. New York is great, too.
