# AI will make money sooner than you think, says Cohere CEO Aidan Gomez

- **Show:** Decoder with Nilay Patel
- **Published:** 10 Jun 2024
- **Source:** https://www.podtrac.com/pts/redirect.mp3/pdst.fm/e/pscrb.fm/rss/p/mgln.ai/e/257/traffic.megaphone.fm/VMP2798911276.mp3?updated=1724755933
- **Transcript:** https://podcasts.opennash.com/shows/decoder-with-nilay-patel/ai-will-make-money-sooner-than-you-think-says-cohere-ceo-aidan-gomez-9ba5b0/
- **Provider:** OpenNash Podcasts (machine-generated)

## In brief

Cohere CEO Aidan Gomez argues that enterprise AI will become profitable sooner than critics expect. Cohere builds right-sized models for big companies rather than the biggest possible ones. Gomez says this keeps its costs far below consumer-focused rivals. He also argues that independence from any one cloud provider is a selling point for risk-averse, price-sensitive customers.

The central tension is that language models are probabilistic, while businesses built modern computing on predictable, deterministic software. Gomez says the answer is human oversight, guardrail systems, and deploying AI only where it is reliable enough. Host Nilay Patel presses on whether the industry is a bubble, whether hallucinations can reach zero, and whether LLMs can carry the expectations placed on them. Gomez stays optimistic.

## Key insights

1. **Enterprise AI fails at the scaling stage** A proof of concept with five users hides the cost of running huge models, Gomez says. Asking a finance office to approve the same system for 10 million users often makes the numbers fail, so Cohere builds smaller, faster models.
2. **Independence from clouds gives buyers leverage** Gomez says a large retailer may spend about $1 billion with one cloud provider and needs to be able to switch to get a good deal. Models that run on any cloud preserve that option, which is why he says Microsoft, Amazon and Google must also support Cohere.
3. **Focusing on RAG cuts Cohere's costs** Retrieval augmented generation (RAG) connects a model to a company's private data. By specializing in this common pattern instead of every use case, Gomez says Cohere spends a fraction of what rivals do on compute.
4. **Safety comes from systems, not one model** Gomez says reliability will come from larger systems where deterministic checks and watcher models monitor the main model and shut it down on known failure patterns. The model is one component of a sophisticated piece of software.
5. **Trust in AI is continuous, not binary** Gomez says there is no single moment when AI becomes trustworthy. Models already beat him at medical knowledge but should not diagnose or prescribe without a doctor, while lower-risk uses with a human in the loop are fine today.

## Questions this episode answers

### What is Cohere and how is it different from OpenAI?

Cohere builds AI language models and sells them to big companies, and Aidan Gomez says it is not trying to make a consumer chatbot like ChatGPT. Its pitch is a platform that helps enterprises add language interfaces to their products and make employees more productive.

### Why do businesses worry that AI is unpredictable?

Traditional software gives the same output for the same input, but language models work on probabilities and can occasionally say something absurd. Gomez says the first question from any customer's board is about risk, so Cohere pushes human review and guardrails.

### Does Cohere's CEO think AI is a bubble?

Gomez says no, because he uses language models constantly and sees real enterprise demand starting to move into production. He agrees that startups spending billions on ever-larger models without a clear business are on a precarious path.

### Can AI be trusted for medical advice yet?

Gomez says no: models should not prescribe drugs or diagnose people without human oversight. He thinks they will eventually match and then beat the average doctor, and says AI can already help by drafting doctors' notes.

### Will AI hallucinations ever go to zero?

Gomez did not promise zero. He says hallucination rates have dropped dramatically and that models can beat the human hallucination rate. Patel pushes back that computers have historically been trusted more than people, and Gomez replies that trust grows use case by use case.

## Facts

- **Aidan Gomez** [0:04:59]: "we were making good progress on translation, which is what we built the transformer for"
- **Aidan Gomez** [0:05:26]: "So I think it took about a year for the community to take notice."
- **Aidan Gomez** [0:07:18]: "There's no way seven years ago when we created the transformer, any of us thought we'd be here in seven years."
- **Aidan Gomez** [0:12:16]: "It was just trained on Wikipedia, and so the model could only write Wikipedia articles."
- **Aidan Gomez** [0:15:20]: "We build models and we make them available for enterprises."
- **Aidan Gomez** [0:31:17]: "we're about 350 at the moment, which is insane to me"
- **Aidan Gomez** [0:31:28]: "it was just Nick, Ivan and I. We were in like this tiny little. It was basically a closet"
- **Aidan Gomez** [0:32:10]: "Last year, it was sort of the year of the POC or the proof of concept."
- **Aidan Gomez** [0:33:42]: "We're talking like hundreds of billions, trillions of tunable parameters inside just a single one of these models."
- **Aidan Gomez** [0:35:57]: "We were both working for Jeff Hinton in Toronto, and he was the guy who kind of created neural networks."
- **Aidan Gomez** [0:36:29]: "the world needed Cohere and the ability to adopt this technology from an independent, an organization that wasn't tied to any one cloud, any one hyperscaler."
- **Aidan Gomez** [0:36:29]: "If you're like a CTO at a large retailer, you're probably spending half $1 billion, $1 billion, on one of the cloud providers for your compute."
- **Aidan Gomez** [0:39:24]: "In order to do what we do, you need to spend hundreds of millions of dollars a year."
- **Aidan Gomez** [0:41:23]: "we have the guy who invented RAG, Patrick Lewis, leading that effort at Cohere"
- **Aidan Gomez** [0:43:06]: "we might spend like twenty what some of our competitors spend on compute"
- **Aidan Gomez** [0:49:29]: "when we released Command R and Command R Plus, our our new models, they were first available on Azure."
- **Aidan Gomez** [0:49:58]: "fantastic companies like Databricks and Snowflake, which are independent"
- **Aidan Gomez** [0:59:22]: "doctors spend 40% of their time taking notes, and this is in between patient visits."
- **Aidan Gomez** [1:00:10]: "We bring that 40% down to 20%, like overnight, we have 25% more doctor hours."
- **Aidan Gomez** [1:01:50]: "When we're scraping code, we look at the licenses that are associated with that code."
- **Aidan Gomez** [1:04:20]: "you can spin up a million accounts, and you can create a a million fake people that project one idea and present a false consensus"
- **Aidan Gomez** [1:09:12]: "hallucination rates, you've you can track them over the years, they've just dropped dramatically, and they've gotten much better"

## Transcript

**[0:00:00] Ad read:** Support for the show comes from Comcast Business. With advances in AI, cybersecurity is getting more powerful, but so are attacks. In the past twelve months, Comcast Business detected 24.5 billion phishing attempts designed to exploit user behavior, steal credentials and compromise devices. AI is forcing a shift in how organizations defend themselves, demanding not only faster responses, but continuous intelligence. To survive in the current climate, leaders must build a security first culture to equip their team with the tools they need to stay protected.

**[0:00:32] Ad read:** With Comcast Business, connectivity, intelligence and response work as one orchestrated system. Get better equipped to handle today's cyber threat landscape with the 2026 Comcast Business Cybersecurity Threat Report. Read it today at Comcast Business dot com slash cybersecurity. Cyber confidence starts with Comcast Business. This holiday, switch to the simplicity plan from Verizon. It's one plan, one price, just $30 a line per month for you or the whole family. And on their best five G network, switching to Verizon and shopping their best deals makes the holidays simple.

**[0:01:09] Ad read:** Head to Verizon dot com or your nearest Verizon store today.

**[0:01:15] Nilay Patel:** Hello and welcome to Decoder. I'm Nilay Patel, editor in chief of The Verge, and Decoder is my show about big ideas and other problems. Today I'm talking with Aidan Gomez, the CEO and co-founder of Cohere. Notably, Aidan used to work at Google, where he was one of the authors of the paper called Attention Is All You Need that described transformers and really kicked off the LLM revolution in AI. Cohere is one of the busiest AI startups around right now, but its focus is a little different than many of the others.

**[0:01:42] Nilay Patel:** Unlike, say, OpenAI, it's not making consumer products at all. Instead, Cohere is focused on the enterprise market and making AI products for big companies. Aidan and I talked a lot about that difference and how it potentially gives Cohere a much clearer path to profitability than some of its competitors. Computing power is expensive, especially in AI. But you'll hear Aidan explain that the way Cohere is structured gives his company an advantage because it doesn't have to spend quite as much money to build its models.

**[0:02:10] Nilay Patel:** One interesting thing you'll also hear Aidan talk about is the benefit of competition in the enterprise space. A lot of the tech industry is very highly concentrated, with only a handful of options for various consumer services. Regular decoder listeners have heard us talk about this a lot before, especially in AI. If you want GPUs to power your AI models, you're probably buying something from Nvidia, ideally a big stack of Nvidia H one hundreds, if you can even get any.

**[0:02:34] Nilay Patel:** But Aidan points out that his enterprise customers are both risk averse and price sensitive. They want Cohere to be operating in a competitive landscape because then they can secure better deals instead of being locked into a single provider. So Cohere has had to be competitive from the beginning, which Aidan says has made the company thrive. Aidan and I also talked a lot about what AI can and can't do. We agreed that it's definitely not there yet.

**[0:02:57] Nilay Patel:** It's not ready, whatever you think the future might hold. And Aidan says, even if you're training AI on a limited, specific, deep set of data, like contract law, you still need a human in the loop. But he sees a time when AI will eventually surpass human knowledge, even in fields like medicine. You know, thing about me, you know, I am very skeptical of that idea. And then there's the really big tension you'll hear us get into all the way through this episode.

**[0:03:19] Nilay Patel:** Up until recently, computers have been deterministic. If you give computers a certain input, you usually know exactly what output you're going to get. It's predictable. There's a logic to it. But if we start talking to computers with human language and getting human language back, well, human language is pretty messy. And it makes the entire process of knowing what to put in and what exactly we're going to get out of our computers different than it's ever been before.

**[0:03:43] Nilay Patel:** And I really wanted to know if Aidan thinks that LLMs, as we have them today, can bear the weight of all of our expectations for AI, given that messiness. Okay, Aidan Gomez, CEO of Cohere, here we go. Aidan Gomez, you are the co-founder and CEO of Cohere. Welcome to Decoder. Thank you, thank you. Excited to be here. I'm excited to talk to you. It feels like Cohere has a very different approach to AI. You have a very different approach to AI.

**[0:04:18] Nilay Patel:** I want to talk about all of that. Uh, the competitive landscape. I'm dying to know if you think it's a bubble, but I want to start with a very big question. You are one of the eight co-authors on the paper that started this all. Attention is all you need. That's the paper that described transformers at Google, that's the T in GPTs. I always ask this question of people who have been on the journey, right? Like I think about music documentaries.

**[0:04:42] Nilay Patel:** There's the kids in the garage playing their instruments and then they're in the stadium. And no one ever talks about Act two. You were in the garage, right? You're writing this paper, you're developing transformers. When did you know this technology would be the basis for everything that's come in the modern AI boom?

**[0:04:59] Aidan Gomez:** I think it was not clear to me. Certainly while we were doing the work, it felt like a regular research project. You know, we were making good progress on translation, which is what we built the transformer for. But that was like a pretty well understood, well known problem. We already had Google Translate. We wanted to make it a little bit better. We improved the accuracy by a few few percent by creating this architecture. And I thought it was kind of it was done.

**[0:05:26] Aidan Gomez:** That was sort of the contribution, was that we improved translation a little bit. It happened later that we started to see the community pick up the architecture and start to apply it to way more stuff than we had ever contemplated when building it. So I think it took about a year for the community to take notice. First, it was published, it went into a academic conference, and then we just started to see this snowball effect where everyone started adapting it to new use cases.

**[0:06:03] Aidan Gomez:** It wasn't just for translation, it started being used for all of these other natural language processing or or natural language processing applications. And then we saw it applied towards language modeling and language representation. And that was really the spark where things started to change.

**[0:06:23] Nilay Patel:** One question I have in that process that you're describing, this is a very familiar kind of abstract process for any new technology product. People develop a new technology for a purpose. A lot of people get their hands on it, the purpose changes, the use cases expand beyond what the inventors ever thought of. And now the next version of the technology gets tailored to what the users are doing. Tell me about that, right? I want to talk about Cohere and the actual company you're building and enterprise use cases and all that.

**[0:06:52] Nilay Patel:** But that turn with transformers and LLMs and what people think they can do now. It feels like the gap is actually widening, right? What the technology can do and what people want it to do, because it's, you know, it's inspiring in its way to get a computer to talk to you. It feels like that gap is widening. And I'm just wondering, since you were there at the beginning and you saw that first turn, how did you feel about that first turn?

**[0:07:14] Nilay Patel:** And do you think we're getting beyond what the technology can do?

**[0:07:18] Aidan Gomez:** I like that description, the idea that the gap is widening because it's inspired so many people. I think the expectations are increasing dramatically. And I it's it's funny that it works that way. Yeah, the the technology, it's it's improved massively, and it's changed in terms of its utility dramatically. There's no way seven years ago when we created the transformer, any of us thought we'd be here in seven years. It it happened much, much faster than than anticipated.

**[0:07:51] Aidan Gomez:** But that being said, that just raises the bar in terms of what people expect. You know, it's a language model, and language is the the intellectual interface that we use. And so it's very easy to personify the technology. And you expect from the tech what you expect from a human. And I think that's reasonable. It's behaving in ways that are genuinely intelligent. And so all of us who are working on this technology project of of realizing language models and and bringing AI into reality, we're all pushing for the same thing.

**[0:08:28] Aidan Gomez:** And our expectations have raised. So I I like that characterization that the the bar for AI has risen. And I I think over the past seven years, there have been so many naysayers of AI. Oh, it's not gonna continue getting better. Oh, the methods that we're using, this architecture that we're using, it's not the right one, et cetera, et cetera. And and they would set bars saying, well, it can't do this. But then fast forward three months, the model can do that.

**[0:08:57] Aidan Gomez:** And they say, okay, well, it can do that, but it can't do. And so that that goalpost moving process is just kept going for like seven years. We've just kept beating expectations and surpassing what we thought was possible with the technology. That being said, there's a long way to go. Like you point out, I think there is still flaws to the technology. And and one of the things that I'm nervous about is because the technology is so similar to what it feels like to interact with a human, that people overestimate it or trust it more than they should and put it into deployment scenarios that it's not ready for.

**[0:09:36] Nilay Patel:** That actually brings me to one of my core questions that I think I'm going to start asking everybody who works in AI. You mentioned intelligence, you mentioned the capabilities. You said the word reasoning. I think. Do you think language is the same as intelligence here, or do you think they're evolving in the technology on different paths? That we're getting better and more capable of having computers use language, and then intelligence is increasing at a different rate, or maybe plateauing.

**[0:10:02] Aidan Gomez:** I I mean, I don't think that intelligence is the same thing as language. I think in order to understand language, you need a high degree of intelligence. There's a question as to whether these models understand language or whether they're sort of just parroting it back to us.

**[0:10:20] Nilay Patel:** This is the other very famous paper at Google, right? The stochastic parrots paper. It caused a lot of controversy, right? But the the claim in that paper is these things are just repeating words back at us, and there isn't some deeper intelligence there. And actually, by repeating things back at us, they will express the bias of the things they're trained on. That's what intelligence gets you over. That's the thing is you can learn a lot of things and your intelligence will help you transcend the things that you've learned.

**[0:10:47] Nilay Patel:** Again, you were there at the beginning. That's kind of why I want to start at that. Is like, is that how you see it, that the models can transcend their training, or they will always kind of be limited by that?

**[0:10:55] Aidan Gomez:** I, I would argue that humans do a lot of parroting, uh, and they have a lot of biases. And so,

**[0:11:00] Nilay Patel:** yeah, I agree.

**[0:11:01] Aidan Gomez:** uh, to a large extent, I think the intelligent systems that we do know exist humans, we do a lot of this. There's that saying we're the average of, like, you know, the ten books we read, uh, or the the ten people closest to us. And so we we model ourselves off of what we've seen in the world. At the same time, humans are genuinely creative. We do stuff that we've never seen before. We go beyond the training data, right?

**[0:11:28] Aidan Gomez:** And I think that's what people mean when they say intelligence is that you're able to discover new truths. And that's more than just parroting back what you've already seen. I think that these models don't just parrot back what they've seen. I think that they're able to extrapolate beyond what we've shown them. Recognize patterns in the data and apply that those patterns to new inputs that they've never seen before. I think definitively at this stage, we can say we're past the stochastic parrot hypothesis.

**[0:12:01] Nilay Patel:** Do you think that's? Uh? Is that an emergent behavior of these models that has surprised you? Is that something you thought about when you were working on transformers at the beginning? You said it's been a journey over seven years. When did that realization hit for you?

**[0:12:16] Aidan Gomez:** There were a few moments I very early on at Google, we started training language models with transformers. So we we just started playing around with it. And it wasn't the same sort of language model that you interact with today. It was just trained on Wikipedia, and so the model could only write Wikipedia articles. Um, but I I remember this, yeah, so it, you know, That might it was

**[0:12:39] Nilay Patel:** that might have been the most useful version of all of this.

**[0:12:42] Aidan Gomez:** yeah, maybe. It was a much simpler version of a language model, and it was a shock to see it. Because at that stage back then, computers could barely string a sentence together properly, like nothing they wrote made sense. There were spelling mistakes. It was just like a lot of noise. And then suddenly, one day, we we kind of woke up sampled from the model, and it was writing entire documents as fluently as a human. And that that just came as this like huge shock to me, like a moment of awe with the technology.

**[0:13:21] Aidan Gomez:** And that's just repeated again and again. I keep having these moments where, yeah, like you, you are nervous that maybe this thing is just a stochastic parrot. Maybe it'll never be able to reach the utility that we want it to reach because th there's some sort of fundamental bottleneck there. We can't make the thing smarter, we can't push it beyond a particular capability. Every time that we improve these models, it breaks through these thresholds. And so at this point, I I think that that breakthrough is going to continue.

**[0:13:58] Aidan Gomez:** And anything that we want these models to be able to do, given enough time, given enough resources, we'll be able to deliver that. Um, it's important to remember that, you know, we're not at that end state already. And so the technology, we shouldn't be letting these models prescribe drugs to people without human oversight. Like there are there are very like obvious applications where the tech's not ready. One day it might be right. Like at some point, you know, genuinely, you might have a model that has read all of humanity's knowledge about medicine.

**[0:14:30] Aidan Gomez:** And you're actually going to trust it more than you trust a human doctor who's only been able to, given the limited time that humans have, read a subset. So that I I view that as a very possible future, but today, the reality that exists. I really hope that no one is taking medical advice, uh, from these models and that it the human's still in the loop. So you have to be conscious of the limitations that exist, uh, with what's today.

**[0:14:55] Nilay Patel:** that's very much what I mean when I say the the gap is widening. And I I think that brings us to cohere. I wanted to start with what I think of as act two, because act two traditionally gets so little attention. But the I built a thing and then I turned it into a business and that was hard for seven years. I I feel like gets so little attention, but now it's easier to understand what you're trying to do at cohere.

**[0:15:18] Nilay Patel:** Cohere is very enterprise focused. Can you describe the company?

**[0:15:20] Aidan Gomez:** We build models and we make them available for enterprises. So we're not trying to do like a consumer like a a chat GPT competitor. What we're trying to build is a platform that lets enterprises adopt this technology. And we're really pushing on, I think, two fronts. The first is okay, we. We just got to the state where computers can understand language, they can speak to us now. And so that should mean that pretty much every computational system, every single product that we've built, we can refactor it to have that interface and to allow humans to interact with it through their language.

**[0:15:57] Aidan Gomez:** And so we we wanna help industry adopt this tech, implement language as an interface into all of their products. That's the first one. It's very external facing for these companies. The second one is internally facing, and it's productivity. So this this tech, I think it's becoming clear that we're entering into a new industrial revolution that's focused. Instead of taking physical labor off the backs of humanity, it's taking intellectual labor. These models are are smart, you know, they can do complicated work that requires reasoning, deep understanding, access to a lot of data and information, which is what a lot of humans do today in work.

**[0:16:37] Aidan Gomez:** And so we can take that labor and we can put it on these models and make organizations dramatically more productive. So th those are the two things that we're trying to accomplish.

**[0:16:47] Nilay Patel:** One of the things about using language to talk to computers and having computers speak to you in language is, I would say, famously, human language, uh, is prone to misunderstandings. Uh, most of history's great stories involve some deep misunderstanding in human language. It's non-deterministic in that way, right? It's re the way we use language is really fuzzy. Programming computers is historically very deterministic. It's very predictable. How do you think philosophically about bridging that gap? We're gonna sell you a product that makes the interface to your business a little fuzzier, a little messier, perhaps a little more prone to misunderstanding, but it'll be more comfortable, right?

**[0:17:25] Nilay Patel:** How do you think about that gap as you go into market with a product like this?

**[0:17:29] Aidan Gomez:** Yeah, the way that you program with this technology, it's non-deterministic, it's stochastic, it's probabilities. It like there's literally a chance that it could say everything right, like there's a there's some probability that it will say something completely absurd. I think our job as technology builders is to introduce good tools for controllability, so that that probability is one in many, many 1 trillion. And so you just in practice never observe it. That being said, I think that businesses are used to stochastic entities and conducting their business, uh, using that, cause we we are humans, right?

**[0:18:10] Aidan Gomez:** Like we are salespeople, and salespeople the most stochastic will say,

**[0:18:13] Nilay Patel:** of all.

**[0:18:14] Aidan Gomez:** Yeah, marketers, you know, like, we're very used to that. Like the world is robust to having that present, like we're robust to noise and error and mistakes. And what we do as humans when we're interacting with a salesperson, you know, hopefully you can trust every salesperson, right? Like, hopefully they they never, uh, mislead or overclaim, but in reality, they do mislead and overclaim sometimes. And so when you're being pitched to by a salesperson, you apply appropriate bounds around what they're saying.

**[0:18:45] Aidan Gomez:** I'm not gonna completely take whatever you say, uh, as the gospel. And so I I think that the world is actually super robust to having systems like these play a part. And while it might seem scary at first because it's like, oh, well, computer programs are completely deterministic. I know exactly what they're gonna output when I put in this input. That's actually unusual. That's weird in our world. It's super weird to have truly deterministic systems. It's a it's a new thing, and we're actually getting back to something that's much more natural.

**[0:19:18] Nilay Patel:** when I look at a jailbreak prompt for one of these chatbots, and you can see the sort of leading prompt, like you are a chatbot, don't say these things, make sure you answer in this way. Here's some stuff that's completely out of bounds for you. You know, those get leaked all the time, and they're, I find them fascinating to read. They're often very long. My first thought every time is this is an absolutely bananas way to program a computer, right?

**[0:19:42] Nilay Patel:** You're going to talk to it like a like a somewhat irresponsible teenager and say, like, this is your role and hopefully it follows it. And there maybe there's a one in 1 trillion chance it won't follow it and it'll say something crazy, but there's still a one in 1 trillion chance. Even after all of these instructions to a computer, it'll go completely off the rails. And I think the internet community delights in making these chatbots go off the rails.

**[0:20:04] Nilay Patel:** You're selling enterprise software, right? You're going into big companies and saying, here's our models. We can control them so that reduces that, so it reduces the possibility of chaos. But we want you to reinvent your business with these tools because they will make some things better, it will make your productivity higher, it will make your customers happier. Are you sensing a gap there? Like, that's the big cultural reset that I think about is that computers are deterministic.

**[0:20:33] Nilay Patel:** We've kind of built modernity around the very deterministic nature of computers. You, you know what outputs you'll get versus what inputs. And now you have to say to a bunch of businesses spend money, risk your business on a new way of thinking about computers, as you've described. It's a big change. Is that working? Are you seeing excitement around that? Are you seeing, uh, a pushback on it? What's the response?

**[0:20:55] Aidan Gomez:** That goes back to what I was saying about knowing where to deploy the technology and what it's ready for, what it's reliable enough for. And so, yeah, there there are places where we don't want to put this technology today because it's not robust enough. I'm lucky in that because Cohere is an enterprise company, we work really closely with our customers. It's not like we just throw it out there and hope that they succeed. We're we're very involved in the process and helping them think through where they deploy it and and sort of what change they're trying to drive.

**[0:21:28] Aidan Gomez:** There's no one who's giving access to their bank account to these models, right, to manage their money. I hope, uh, so there are places where, yeah, you want determinism, you want extremely high confidence guardrails. And so you're not gonna just put a model there and let it decide what it wants to do. In the vast majority of use cases and applications, it's actually about augmenting humans. So you have a a human employee who is trying to get some work done.

**[0:21:57] Aidan Gomez:** And they're gonna use this thing as a tool to basically make themselves faster, more effective, more efficient, more accurate, and it it's augmenting them. But they're still in the loop, so they're they're still checking that work. They're still making sure that the model is producing something that's sensible. And at the end of the day, they're they're accountable to the decisions that they make and and what they do with that tool as part of their job. I think what you're pointing to is what happens in those applications where a human is completely out of the loop.

**[0:22:27] Aidan Gomez:** And we're really offloading the full, the full job onto these models. That's a ways away. And I I think that you're right. We need to have much more trust and controllability and the ability to set up those guardrails so that they behave more deterministically. You pointed to the prompting of these models and how it's funny that the way you actually control them is by talking to them. Like, yeah, like you said, like a talking to a it's like a stern teenage

**[0:22:57] Nilay Patel:** lecture. It's it's that's.

**[0:22:58] Aidan Gomez:** thing, huh?

**[0:22:58] Nilay Patel:** It is crazy to me every time I look at one.

**[0:23:01] Aidan Gomez:** It's somewhat magical, right? The fact that you can actually control the behavior of these things effectively using that method, like that's pretty magical. But beyond that, beyond just prompting and talking to this thing, you can set up controls and guardrails outside of the model. You can have models watching this model and intervening and blocking it from doing certain actions in certain cases. And so I think what we need to start changing is our conception of is this one model?

**[0:23:34] Aidan Gomez:** It's like one AI, which we're just handing control over to. What if it messes up? What if everything goes wrong? In reality, it's going to be much larger systems that include observation systems that are deterministic and check for patterns of failure. If the model does this and this, you know it's gone off the rails and so shut it down, right? And that's a completely deterministic check. And then you'll have other models which can observe and sort of give feedback to the model, prevent it from taking actions.

**[0:24:03] Aidan Gomez:** If it's, you know, going astray, the programming paradigm or the technology paradigm. It started off as what you're describing, which is there's a model and you're gonna apply it to some use case, and so it's just the model and the use case. It's shifting towards bigger systems with much more complexity and components, and it's less like there's an AI that you're applying to go do work for you. And it's actually a sophisticated piece of software that you're deploying to go do work for you.

**[0:24:40] Nilay Patel:** We have to take a quick break. We'll be back in a minute.

**[0:24:52] Ad read:** Support for the show comes from Quo. You know, when you're juggling five things at once and your business phone rings and you just let it go. That call might have been your next best customer. That's why today's episode is brought to you by Quo, spelled Q U O, the business phone system built so you never miss a call. All your calls, texts and voicemails live in one place, so anyone on your team can pick up a conversation, see the full history and respond fast.

**[0:25:21] Ad read:** More than 90,000 businesses rely on Quo to stay reachable and look professional every day. Setup takes minutes, and you can keep your existing number and add teammates as you grow. Quo's optional built-in AI agent handles after hour calls, answers questions, and even books appointments, even when your team is offline. Money is on the line. Always say hello with Quo. Try Quo for free, plus get twenty off your first six months when you go to Quo dot com slash decoder, spelled Q U O dot com slash decoder.

**[0:25:58] Ad read:** Support for the show comes from Rippling. You just found out your sales team is at risk of missing quota. Don't panic, just ask Rippling AI. Since it's built on your real-time people and business data, Rippling AI can pull metrics from Rippling and Salesforce into a meeting-ready dashboard, which shows quota attainment, headcount plan, and monthly revenue to quota by region. In seconds, you'll see exactly what's behind your quota risk and fix it before it's missed. Question answered, action taken, crisis averted.

**[0:26:29] Ad read:** When you have critical business questions that need answers, don't just file a ticket and wait weeks for an outdated report. Describe what you need and have Rippling AI build it instantly from your live people and business data. Whether it's a dashboard with detailed charts or automated workflows with the right triggers, conditions, and approvals. Ready to rule your business? Head to rippling dot ai slash decoder. Get the only AI built to give you full visibility and take complex actions across your entire organization.

**[0:26:57] Ad read:** That's r i p p l i n g dot ai slash decoder. Sign up for exclusive access today. Support for the show comes from Granola. Have you ever left a meeting thinking that was productive? And then 10 min later, when it's time to take next steps, you're at a loss. Or worse, you've been heads down the entire meeting, furiously typing notes instead of actually listening, and now your brain is fried. Meetings are a mess. Granola fixes that.

**[0:27:29] Ad read:** Granola is an AI power notepad that makes your meeting notes useful. It capture what happens in your meetings and turns it into clean, structured notes with decisions and action items pulled out and made easy to find. And the best part, Granola works through your device's audio, which means it integrates seamlessly into the video conferencing tools you already use. No setup, no awkward bots. It's just your normal meeting with superpowers. You get to actually listen instead of frantically typing every word, and still walk away knowing exactly what was decided, who's doing what, and what comes next.

**[0:28:04] Ad read:** Once you try it on your first meeting, it's hard to go without. You could try it totally free. Just head to granola dot ai slash decoder. That's granola dot ai slash decoder to get your time back. Go to granola dot ai slash decoder and try it on your next call.

**[0:28:27] Nilay Patel:** Welcome back. I'm talking to Cohere CEO Aidan Gomez about a pretty foundational shift in the way people and computers interact. Up until now, we've been used to what you would call a deterministic model, where you give a computer a certain input and you know what kind of output you'll get. But generative AI is really changing how we interact with computers. Cohere right now has two models, Cohere Command, Cohere Embed. You're obviously working on those models, you're training them, developing them, applying them to customers.

**[0:28:54] Nilay Patel:** How much of the company's spending its time on this other thing you're describing, building the deterministic control systems, figuring out how to chain models together to provide more predictability?

**[0:29:04] Aidan Gomez:** I can speak to the enterprise world, and and their enterprises are super risk averse. They're always looking for opportunities, but they're extremely risk averse. And so that's the first thing that they're thinking about. Like, pretty much every initial conversation I have with a customer is about what you're asking. That's the first thing that comes to a person's mind is, can I use this system reliably? And so we need to show them, okay, well, let's look at the specific use case that you're pursuing.

**[0:29:31] Aidan Gomez:** Maybe it is assisting lawyers with contract drafting, which is something that we do with a company called Uh, Borderless. In that case, you need a human in the loop. There's no way that you're gonna be sending out contracts that are completely synthetically generated with no oversight. And so we come in and we we try to help guide and educate in terms of the types of systems that you can build for oversight, whether it's humans in the loop or more automated systems to help de-risk things.

**[0:30:01] Aidan Gomez:** But that that for enterprises at least, uh, with consumers, I think it's a little bit different. But for enterprises, the very first question we'll get from a board or a C-suite at the company, it's going to be related to risk and protecting against it.

**[0:30:16] Nilay Patel:** So apply that to Cohere and how you're developing your products. How is Cohere's structured? Is that reflected in how the company is structured?

**[0:30:24] Aidan Gomez:** I think so. I mean, like, we have safety teams internally that are focused on making our models more controllable, less biased. And at the same time, on the go to market side, that project, because this technology is new, it's an education campaign, right? Like it's getting people familiar with what this technology is. It's a paradigm shift in terms of how you build software and technology, like we were saying, it's stochastic, and so educating people about that.

**[0:30:54] Aidan Gomez:** We build stuff like the LLMU, which is like the LLM University, where we teach people what the pitfalls might be with the tech and how to how to protect against those. So for us, I think our structure is focused on helping the market get familiar with the technology and its limitations while they're adopting it.

**[0:31:15] Nilay Patel:** How many people are at Cohere?

**[0:31:17] Aidan Gomez:** It's always shocking to say, but we're about 350 at the moment, which is insane to me.

**[0:31:26] Nilay Patel:** It's only insane because you're the founder, right?

**[0:31:28] Aidan Gomez:** Yeah, yeah, well, I like, So y it feels like yesterday, it like, Yeah, it was just Nick, Ivan and I. We were in like this tiny little. It was basically a closet, like I I don't know how many square meters it was, but you know, single digits. We we had a company offsite a few weeks back, and it was hundreds of people, hundreds of people building this thing alongside you. And you do ask yourself, like, how did, how did we get here, you know, like, how did, how did all this happen?

**[0:31:58] Aidan Gomez:** So it's really, it's really fun.

**[0:32:00] Nilay Patel:** so of those 350 people, what's the split? How many are engineering? How many are sales? Enterprise companies need a lot of post-sales support. What's the split there?

**[0:32:10] Aidan Gomez:** Yeah, the overwhelming majority are engineers. Very recently, the go-to-market team has exploded, right? So I I think that market is just going into production now with this technology. It's starting to actually hit the hands of of employees, of of customers, of users. Last year, it was sort of the year of the POC or the proof of concept. So everyone became aware of the technology. We we've been working on this for nearly five years now, but it was only like 2023 when people, really, the general public, noticed it and started to use it and and fell in love with the technology.

**[0:32:50] Aidan Gomez:** And so that led to enterprises. You know, there are people too, they're hearing about this, they're using this, they they're thinking of how they can adopt the technology. So they got excited about it, and they spun up these tests, these POCs, to try and build a deeper understanding and familiarity with the tech. Those POCs, like the initial cohort of them, they're complete now, and people like the stuff that they've built. And so now it's a project of taking those pre-deployment tests and actually getting them into production in a scalable way.

**[0:33:24] Aidan Gomez:** That's the majority of our focus is scalability in production.

**[0:33:28] Nilay Patel:** Is that scalability in okay? We can add five more customers without a massive incremental cost. Is it scalability in compute? Is it scalability in how fast you're designing the solutions for people, or is it everything?

**[0:33:42] Aidan Gomez:** Yes, it's all of the above. I I would say one thing is, as a lot of people may have heard, the tech is expensive to build and expensive to run. We're talking like hundreds of billions, trillions of tunable parameters inside just a single one of these models. And so it requires a lot of memory to store these things, it requires tons of compute to run them. And so during the POC phase last year, there's like in a POC, you have like five users, and so scalability doesn't matter.

**[0:34:16] Aidan Gomez:** Like the cost, just it's kind of irrelevant. You just want to build a proof of what is possible. But then if you like what you've built and you you're gonna push this thing into production and you go to your finance office and you say, Yeah, okay, here's here's what it costs for five users. We'd like to put it in front of all 10 million. The numbers are they don't compute, right? Like it's not economically viable. And so for Cohere, we've been focused on not making the largest possible model, but instead making the model that market can actually consume and is actually useful for enterprises.

**[0:34:53] Aidan Gomez:** And so that that's doing what you say, which is focusing on compression, speed, scalability, uh, and ensuring that we can actually build a technology that market can consume. Because over the past few years, a lot of this stuff has been a research project without large-scale deployment. And so the concerns around scalability hadn't yet emerged. But we knew for enterprises, very cost-sensitive entities, very economically driven, right? If they can't make the numbers work in terms of return on investment.

**[0:35:27] Aidan Gomez:** I'm gonna spend this much to deploy this new product feature. If that doesn't work, they don't adopt it. It's very simple. And so we we've been focused on building a category of technology that is actually right-sized for market.

**[0:35:39] Nilay Patel:** You obviously started all this work at Google. Google has an infinite amount of resources. Google also has massive operational scale. Their ability to optimize and bring down the cost curve of new technologies like this is very high, given their infrastructure and their reach. What made you wanna go and do this on your own without their scale?

**[0:35:57] Aidan Gomez:** For Nick Ivan and I, uh, Nick was also at Google. We were both working for Jeff Hinton in Toronto, and he was the guy who kind of created neural networks. Like the the technology that underpins all of this, that underpins LLMs, it underpins pretty much every AI that you interact with on a daily basis. We loved it there, but I think what was missing was a product ambition and a velocity that we felt was necessary for us to to execute.

**[0:36:29] Aidan Gomez:** And so we had to, we had to start. Cohere Google was a great place to do research, and I think it has some of the smartest people in AI on the face of the planet. But for us, the world needed something new. Uh, the world needed Cohere and the ability to adopt this technology from an independent, an organization that wasn't tied to any one cloud, any one hyperscaler. Something that's very important to enterprises is optionality. If you're like a CTO at a large retailer, you're probably spending half $1 billion, $1 billion, on one of the cloud providers for your compute.

**[0:37:08] Aidan Gomez:** And in order to get a good deal, you need to be able to plausibly flip between compute providers. Otherwise, they're just gonna squeeze, you know, ad infinitum, and, you know, rip you off. So you need to be able to flip. And so you hate buying proprietary technology that's only available on one stack. You really want to preserve your optionality to flip between, and that's what Cohere allows for. So because we're independent, because we haven't gotten locked into one of these big clouds, we're able to offer that to market, which is super important to.

**[0:37:40] Nilay Patel:** Let me ask you the decoder question. We've talked a lot about the journey to get here, the challenges you need to solve. You're a founder, you've got 350 people now. How do you make decisions? What's your framework to make decisions?

**[0:37:51] Aidan Gomez:** What's my framework? I flip a coin.

**[0:37:54] Nilay Patel:** It's stochastic. You, you're allowed to say stochastic basically, I ask, now.

**[0:37:57] Aidan Gomez:** I ask the model, I ask the model, what? What do I say? I'm lucky in that I'm surrounded by people who are way smarter than me, like I'm just surrounded by them. Everyone at Cohere is better than me at the thing that they do. And so I have this luxury of just being able to go ask people for advice, whether it's the board of Cohere or the executive team of Cohere, or or the ICS, like the the people who are actually doing the real work.

**[0:38:27] Aidan Gomez:** I can just ask for advice and their takes, and I I can kind of be an aggregation point. And when there are ties, then it just comes down to me. And and usually it's just going with my intuition about what's right. Um, but fortunately, I don't have to make a lot of decisions, cause I have way smarter people that surround me.

**[0:38:47] Nilay Patel:** there are some big decisions you do have to make. You, you just, for example, announced two models in April Command R. There's one called re rank three. Models are costly to train, they're costly to develop. You got to rebuild your technology around the new models and its capabilities. Those are big calls. It feels like every AI company is sort of racing to develop the next generation of models. How are you thinking about that investment over time, right?

**[0:39:09] Nilay Patel:** You talked a lot about the cost of a proof of concept versus an operationalized thing. New models are the most expensive of them all. How are you thinking about those costs?

**[0:39:19] Aidan Gomez:** Yeah, it's really, really expensive, like, what could Give us a, give you're?

**[0:39:23] Nilay Patel:** us a number. Can you give us a number?

**[0:39:24] Aidan Gomez:** I don't know if I can give a specific number, but I can say, like, order of magnitude. In order to do what we do, you need to spend hundreds of millions of dollars a year. That's what it costs. And that's like, for us, we think that we're hyper capital efficient, extremely capital efficient. We're not trying to build models that are too big for market that are kind of superficial. We're trying to build stuff that market can actually consume.

**[0:39:47] Aidan Gomez:** And so because of that, it's cheaper for us and we can. We can focus our capital more. There are folks out there spending many, many billions of dollars a year to build their models. Yeah, I I think that's that's a huge consideration for us. We're lucky in that we're small, relatively speaking. Our strategy lends itself towards more capital efficiency and actually building the technology that market needs, as opposed to building prospective research projects. We we focus on actual tech that the market can consume.

**[0:40:22] Aidan Gomez:** But yeah, like you say, it's hugely expensive. And the way that we solve that is a raising money, right? Like getting the capital to actually pay for the work that we need to do. And then B, choosing to focus in our technology. So instead of trying to do everything, instead of trying to nail every single potential application of the technology, we focus on the patterns or use cases that we think are are going to be dominant or are dominant already in how people use it.

**[0:40:53] Aidan Gomez:** One example of that is, uh, RAG R A G or retrieval augmented generation. It's this idea that these models are trained on the internet, they have a lot of knowledge about, you know, public facts and that type of thing. But if you're an enterprise, you want it to know about you, you want it to know about your enterprise, your proprietary information. And so what RAG lets you do is sit your model down next to your private databases or stores of your knowledge and connect the two.

**[0:41:23] Aidan Gomez:** So that pattern, that's something that is ubiquitous. Anyone who's adopting this technology, they want it to have access to their internal information and and knowledge. And so we focused on getting extremely, extremely good at that pattern. And we're fortunate we have the guy who invented RAG, Patrick Lewis, leading that effort at Cohere. But because we're able to carve away a lot of the space of potential applications, it lets us be dramatically more efficient in what we want to do and what we want to build with these models.

**[0:41:53] Aidan Gomez:** And so that that'll continue into the future. But that's like I said, it's still a multi 100 million dollar a year project, so it's very, very capital intensive.

**[0:42:03] Nilay Patel:** I said I wanted to ask you if this was a bubble. I'll start with Cohere specifically, but then I want to talk about the industry in general. So it's multi $100 million a year just to run the company, to just run the compute. That's before you've paid a salary, and the AI salaries are pretty high. So that's another bunch of money you got to pay. And you got to, I don't know, pay for office space, you got to buy laptops, right?

**[0:42:24] Nilay Patel:** There's a whole bunch of stuff.

**[0:42:25] Aidan Gomez:** There's a lot of stuff, there's a lot of stuff,

**[0:42:27] Nilay Patel:** Uh, but just the compute is a hundred, hundreds of millions of dollars a year. That's the run rate on just the compute. Do you see a path to revenue that justifies that amount of just pure run rate in compute?

**[0:42:40] Aidan Gomez:** absolutely. But we wouldn't be building it if we didn't. I think Uh,

**[0:42:44] Nilay Patel:** your competitors are like, it'll come, like, th. There's like a there's a lot of wishful thinking. So starting with that question with you, because I, because you have started an enterprise business, I'm assuming you see a much clearer path. But in the industry, I see a lot of a lot of wishful thinking that it'll just arrive down the road.

**[0:43:02] Aidan Gomez:** I, I completely agree.

**[0:43:03] Nilay Patel:** What? So, what is Cohere's path specifically?

**[0:43:06] Aidan Gomez:** Like I said, we're dramatically more capital efficient. Like we, you know, we might spend like twenty what some of our competitors spend on compute. But what we build is very, very good at the stuff that market actually wants. And so we can chop off eighty of the expense and deliver something that is just as compelling to market. That's a core piece of our strategy of how we're gonna do this. And of course, if we didn't see a business that was many, many billions in revenue, like we, we just wouldn't be building this.

**[0:43:40] Nilay Patel:** What's the path to billions in revenue? What's the timeline?

**[0:43:44] Aidan Gomez:** The timeline, I I think billions of revenue, I don't know about specifically for Cohere, how much I can disclose. It's closer than you think. There's a lot of spend that's being activated in market. Certainly already, there's billions being spent on this technology in the enterprise market today. And a lot of that goes to the compute as opposed to the models, but there's a lot of spending happening in AI. And like I was saying, last year was very much a POC phase, and POC spend is, you know, 5%, three of what a production workload looks like.

**[0:44:23] Aidan Gomez:** But now those production workloads are coming online. This technology is hitting products that interact with tens, hundreds of millions of people, and so it's it's really becoming ubiquitous. It's close, it it's in a matter of a few years, and it's typical for like, a technology adoption cycle. Enterprises are slow, they tend to be slow to adopt, they're very sticky, right? Like once they've adopted something, it's there in perpetuity. But it takes them a while to build confidence and actually make the decision to commit and and adopt the technology.

**[0:44:59] Aidan Gomez:** So it's only been about a year and a half since people woke up to the tech, but in that year and a half, we're now starting to see real, serious adoption, serious production workloads.

**[0:45:13] Nilay Patel:** We have to take another quick break. When we come back, we'll dig into one of the biggest challenges for any enterprise company managing risk.

**[0:45:24] Ad read:** Black Friday is around the corner, your biggest chance to grow sales this year. Don't wait. Realize is Taboola's advertising platform for businesses, letting you launch campaigns on top premium websites beyond search and social. Add Realize to your media mix and scale this Q4. And as an exclusive offer for podcast listeners, new customers get $700 credit back. Just search Realize Podcast on Google or use the promo code Podcast twenty six. Don't miss out on this limited time offer.

**[0:45:56] Ad read:** Terms apply. Support for the show comes from CrowdStrike. AI is already fundamentally changing how companies build products and run their businesses. If you listen to the show, you probably already know this. But while AI drives innovation, every new model and agent also creates new cybersecurity vulnerabilities. That's why CrowdStrike built AI detection and response, or AI DR. AI DR helps organizations discover where AI is being used, monitors AI activity at runtime, and helps teams detect, investigate and respond to threats involving AI systems.

**[0:46:37] Ad read:** As companies adopt AI, security has to be built in from the start. More than seventy of the Fortune one hundred trust CrowdStrike to protect their businesses. Now, CrowdStrike is helping secure the AI era. You can learn more about AI detection and response at CrowdStrike dot com. Support for the show comes from Framer. If you want to create a website that looks great, feels handcrafted and comes together fast and seamlessly, then look no further than Framer. It's a complete website platform, not just a builder, so teams can launch and keep improving their sites in one place.

**[0:47:18] Ad read:** It helps creators, teams and businesses ship production-ready sites faster than ever while getting every detail right. You're able to prompt, inspect, edit, and publish in one place at a whole new pace. On Framer, agents and humans work in tandem. Agents bring speed and scale, while people bring taste, judgment and control. And that all takes place in the same workspace where the real site is designed, managed, reviewed, and published. Thousands of businesses, from early-stage startups to Fortune five hundreds, are choosing to build their websites in Framer, where changes take minutes instead of days.

**[0:47:58] Ad read:** Learn how you can get more out of your site from a Framer specialist or get started building for free today at framer dot com slash decoder for thirty off a Framer Pro annual plan. That's framer dot com slash decoder. Rules and restrictions may apply.

**[0:48:21] Nilay Patel:** Welcome back. I'm talking with Cohere CEO Aidan Gomez about how he convinces inherently risk-averse enterprise customers to sign up for AI services. When I think about enterprise technology that's very sticky, that will never go away, the first thing that comes to mind is Microsoft Office, which will never go away. That is the foundation of their enterprise strategy is Office three sixty five. Microsoft obviously a huge investment in OpenAI, they've got models of their own. Kevin Scott has been on the show, their CTO who's in charge of AI.

**[0:48:52] Nilay Patel:** They're the big competitor for you, right? They're the ones in market selling Azure to enterprise. They're a hyperscaler. They'll give you deals, they'll integrate it directly so you can talk to Excel. The pitch that I've heard many times from Microsoft folks is you've got people in the field who need to wait for an analyst to respond to them. But now they can just talk to the data directly and get the answer they need and be on their on their way.

**[0:49:14] Nilay Patel:** That's very compelling. I think it requires a lot of cultural change inside inside some of these enterprises to let those sorts of things happen. You're obviously the challenger, right? You're the startup. Microsoft is 300,000 people, you're 350 people. How are you winning business for Microsoft?

**[0:49:29] Aidan Gomez:** Well, I think they're a competitor in some respect, but they're also a partner and a channel for us. So when we released Command R and Command R Plus, our our new models, they were first available on Azure. And so I definitely view them as a partner in bringing this technology to enterprise. And I think that Microsoft views us as a partner as well. I think that they want to create an ecosystem powered by a bunch of different models.

**[0:49:58] Aidan Gomez:** I'm sure they'll have their own in there, they'll have OpenAI's, they'll have ours. And so it it'll be an an ecosystem, as opposed to just only proprietary Microsoft tech. And I think if you look at the story in databases, there you have, like, fantastic companies like Databricks and Snowflake, which are independent. That's not a subsidiary of, you know, Amazon or or Google or Microsoft. They're an independent company, and the reason they've done so well is because they have an incredible product vision.

**[0:50:30] Aidan Gomez:** So the product that they're building is genuinely the best option for for customers. But also the fact that they're independent is crucial to their success. For that reason that I was describing, where CTOs don't want to get locked into one proprietary software stack because it's such a pain and a a strategic risk to their ability to negotiate. So I think the same is going to be true. It's it's even more important with AI, where these models become an extension of your data.

**[0:50:59] Aidan Gomez:** They are the value of your data. The value of your data is that you'll be able to power an AI model that drives value for you. The the data in itself is not inherently valuable. I think the fact that we're independent, folks like Microsoft, Azure, AWS, GCP, they want us to exist and they have to support us. Because the market is going to reject them if they don't. The market is going to insist on being able to adopt independence that let them flip between clouds, and so they kind of have to support our models.

**[0:51:32] Aidan Gomez:** That's just what market wants. I don't feel like they're exclusively a competitor. I I view them as a partner to bring this technology to market.

**[0:51:43] Nilay Patel:** One thing that's interesting about this conversation, and one of the reasons I was excited to have it with you is because you are so focused on enterprise. There's a there's a certainty to what you're saying. Like, you've identified a bunch of customers with some needs, they've articulated their needs, they have money to spend. You can identify how much money it is, you can build your business around that money. You keep talking about the market. You can spend your budget on technology appropriately for the size of the money that's available in the market.

**[0:52:11] Nilay Patel:** When I ask if it's a bubble, what I'm really talking about is the consumer side, right? There's these big consumer AI companies that are building big consumer products and their idea is people will pay twenty bucks a month to talk to a model like this. And those companies are spending more money than on training than you are. They're spending more money per year on compute than you are. They are the leading edge companies. I'm talking about Google and OpenAI, obviously.

**[0:52:36] Nilay Patel:** But then there's a whole ecosystem of companies that are paying OpenAI and Google a margin to run on top of their models to go sell a consumer product at a lower rate. And that does not feel sustainable to me. Do you have that same worry about the rest of the industry? Because that's what's powering a lot of the attention and a lot of the interest and a lot of the inspiration, but it doesn't seem sustainable.

**[0:52:56] Aidan Gomez:** I mean, I think those folks who are building on top of OpenAI and and Google should be building on top of Cohere. So Like, we'll be a a better partner, I laid you know,

**[0:53:04] Nilay Patel:** that one out for you.

**[0:53:04] Aidan Gomez:** yeah, you're right to identify that the technology provider's focus, it might conflict with its users. And you might find yourself in situations where I don't want to name names, but let's say, you know, there's a consumer startup that's trying to build an AI application for the world. And it's building on top of one of my competitors, who is also building a consumer AI product. There's a conflict inherent there, and you might see one of my competitors sort of steal the ideas or rip off the ideas of that of that startup.

**[0:53:44] Aidan Gomez:** And that's why I think that Cohere needs to exist. You need to have folks who are like us, who are focused on building a platform to enable others to go create those applications and and are really invested in their success. Free of any conflicts or competitive nature, that's why I think we're a really good partner is because we're we're focused and we let our users succeed without trying to compete or or play in the same space.

**[0:54:12] Aidan Gomez:** We just build a platform that you can use to adopt the technology, that that's our whole business.

**[0:54:18] Nilay Patel:** Do you think it's a bubble when you look around the industry?

**[0:54:22] Aidan Gomez:** I don't, I really don't like I just in terms of myself. I don't know how much you use LLMs day to day. I use them constantly, like l- multiple times an hour type thing. How could it be a bubble?

**[0:54:35] Nilay Patel:** Well, I I think maybe the utility is there in some cases, the economics might not be there. That's that's how I would think about it being a bubble. Uh, I'll give you an example. You've talked a lot about the dangers of overhyping AI, even in this conversation, but you've talked about it publicly elsewhere. You've talked about you've got two ways to fund your compute. You can get customers and grow the business, or you can raise money.

**[0:54:56] Nilay Patel:** I look at how some of your competitors raise money, and it's by saying things like, we're gonna build AGI on the back of elephants, and saying things like, we actually need to pause development so we can catch up, uh, because we might destroy the world with this technology. That stuff to me seems pretty bubbly, right? Like, we need to raise a lot of money so we can continue training the next frontier model before we've built a business that can even support the compute of the existing model.

**[0:55:24] Nilay Patel:** But it doesn't seem like you're that worried about it. You think that's gonna even itself out?

**[0:55:28] Aidan Gomez:** I very much agree that is a precarious setup. I the reality is, for folks like Google and Microsoft, they can spend billions of dollars, they can spend tens of billions of dollars on this, and it's it's fine, right? The amount of capital that they have, it's like, it's genuinely okay. It doesn't really matter. It's a rounding error.

**[0:55:48] Nilay Patel:** that's interesting.

**[0:55:49] Aidan Gomez:** I think for for startups taking that strategy, like, you either need to become a subsidiary of one of those big guy, like the big tech companies that prints money, or do some very, very poor business building in order to to do that. That's not what Cohere is pursuing. And so I I think, like, I agree with you to a large extent. I think that that's a bad strategy. I I think that ours, the focus on actually delivering what market can consume and building the products and the technology that is right-sized or fit for our customers, that's what you need to do.

**[0:56:28] Aidan Gomez:** That's how you build a business, that's how all successful businesses were built. Yeah, we don't wanna get too far out in front of our skis, right? Like, we don't wanna be spending so much money that, like you say, it's hard to see a path towards profitability. Cohere's focus is very much on building a self-sustaining, independent business. And so we we're forced to actually think about this stuff and steer the company in in a direction that supports that.

**[0:56:58] Nilay Patel:** Yeah, you've called the idea that AI represents existential risk. Uh, I believe the the word you've used is absurd, uh, and you said it's a distraction. Why do you think it's absurd? And what? What do you think the real risks are?

**[0:57:10] Aidan Gomez:** I think the real risks are the ones that we spoke about, which is like over-eager deployment of the technology too early, people trusting it too much in scenarios where, frankly, they shouldn't. The doomsday scenarios or the terminator scenarios. Like, I'm super empathetic to the public's interest in those scenarios. Um, I'm interested in those scenarios because I've watched sci-fi and, like, I, you know, it always goes badly. So we we've been told those stories for decades and decades, so it's a very salient narrative.

**[0:57:42] Aidan Gomez:** It really captures the imagination. It's super exciting and fun to think about, but it's not reality, it's not our reality. As someone who's technical and quite close to the technology itself, I don't see us heading in a direction that supports the stories that are being told in the media. And like you say, often by companies that are that are building the tech. I really wish that our focus was on 2 things. One is like the risks that are here today, like over-eager deployment, deploying them into scenarios without human oversight, those sorts of discussions.

**[0:58:19] Aidan Gomez:** And when I talk to regulators, when I talk to folks in government, that's the stuff they actually care about. It's not doomsday scenarios. It's like, is this gonna hurt the general public if the financial industry adopts it in this way, or the medical industry adopts it in this way? They're quite practical and and actually grounded in the reality of the technology. Those stories are are very exciting, and so that's why they they dominate the the headlines.

**[0:58:48] Aidan Gomez:** I mean, the other thing that I I would really love to see a conversation on is the opportunity, like, like the positive side. We spend so much time on the negatives and fear and doom and gloom. I I really wish, like, someone was just talking about what we could do with the technology or what we wanna do. Because as much as it's important to steer away from the potential negative paths or bad applications, I I also wanna hear the public's opinion and public discourse about what are the opportunities, like, what good could we do?

**[0:59:22] Aidan Gomez:** I I think one example in medicine, like, apparently, doctors spend 40% of their time taking notes, and this is in between patient visits. You, you have your interaction with the patient, you then go off, you go to your computer and you say, so and so came in, they had this. I remember from a few weeks when they came in, it looked like this. We should check this the next time they come in. I prescribe this drug.

**[0:59:44] Aidan Gomez:** So they they spend a lot of time typing up these notes in between the interactions with patients 40% of their time, apparently. If we could attach passive listening mics that just go from patient meeting to patient meeting with them, transcribe the conversations and just like pre-populate that, so that instead of having to write this whole thing from scratch, they read through it and they say, No, I didn't say that, I said this and add that, and it becomes an editing process.

**[1:00:10] Aidan Gomez:** We bring that 40% down to 20%, like overnight, we have 25% more doctor hours. And I I think that's like, incredible. That's a that's a huge good for the world. We haven't paid to train doctors, we haven't added more doctors in school. We we have 25% more just by adopting technology. And so I I wanna find more ideas like that. What application should Cohere be prioritizing? What do we need to get good at? What, what should we solve to drive the good in the world that that we wanna see?

**[1:00:45] Aidan Gomez:** Th- there's no headlines about that. There's like, no one is talking about it, and I I really wish we were having that conversation.

**[1:00:53] Nilay Patel:** As somebody who writes headlines, I think one, there aren't enough examples of that yet to say it's real, which I think is something people are very skeptical of. Two, I hear that story and I think, oh boy, a bunch of private equity owners of urgent care clinics just realized they could shove 25% more patients into the doctor's schedule. And like, there's a real a real animosity towards that. What I hear from our audience, for example, is they feel like right now, the AI companies are taking a lot without giving enough in return, right?

**[1:01:22] Nilay Patel:** And that's a real challenge. Uh, that's been expressed mostly in the creative industries, right? We we see that generated, we see that anger kind of directed at the creative generative AI companies. You're obviously an enterprise, you don't feel it, but do you see that that you've trained a bunch of models, you should know where the data comes from, and then the people who made the original work that you're training on probably wanna get compensated for it.

**[1:01:44] Aidan Gomez:** Oh yeah, totally. I I'm very empathetic to that. I I think,

**[1:01:47] Nilay Patel:** Do you do you compensate? Where you train from?

**[1:01:50] Aidan Gomez:** yeah, we pay for data, but we pay a lot for data. So there's there's a bunch of different sources of data. There's stuff that we we scrape from the web, and when we do that, we try to abide by people's like. If they express, we don't want you to collect our data, we abide by that, right? We look at robots. Txt. When we're scraping code, we look at the licenses that are associated with that code. We filter out data where people have said, clearly, don't scrape this data, or, you know, don't use this code.

**[1:02:20] Aidan Gomez:** If someone emails us and says, Hey, I think that you scraped X, Y and Z, can you remove it? You know, we will, of course, remove that, and all future models won't include that data. So we we don't wanna be training on stuff that people don't want us training on, like full stop. So, like, I, I, I'm very, very empathetic to creators, and I I really want to support them, build tools to help make them more productive, right?

**[1:02:45] Aidan Gomez:** And and help them with their ideation and creative process. That's the impact that I wanna have, and I I really wanna respect their content.

**[1:02:54] Nilay Patel:** yeah. The other flip side of it is those same creators are watching the platforms they publish on get overrun with AI content, and they don't like it, right? So there's a little bit of a competitive aspect there. That's one of the dangers you've talked about, right? There's a straightforward misinformation dangers on social platforms that don't seem to be well mitigated yet. Do you have ideas on how you might mitigate just AI generated misinformation?

**[1:03:19] Aidan Gomez:** Yeah, I I like, that's one of the ones that scares me a lot is the democratic world is vulnerable to influence and manipulation. Just in general, it's like, take out AI, like, just democratic processes are very vulnerable to manipulation. If you're able to build yourself a platform where you can project ideas, I guess, you know, we started off the podcast saying this. People are the average of the last, you know, 50 posts that they've seen or whatever.

**[1:03:51] Aidan Gomez:** Like, you're, you're very influenced by what you perceive to be consensus. If you're like, I look out into the world on social media and everyone seems to agree X, then you're like, Okay, I guess X is right. Like, I trust the world, I trust consensus. Humans are we are pack animals, herd animals. Democracy is vulnerable, and it's something that needs to be very vigorously protected. And now you can ask the question about how does AI influence that?

**[1:04:20] Aidan Gomez:** What AI enables is much more scalable manipulation of public discourse. So you can spin up a million accounts, and you can create a a million fake people that project one idea and present a false consensus to to the people consuming that content. Now, that sounds really scary. I think that's terrifying. That's like, you know, this is a huge threat. I think it's actually very, very preventable. So, for instance, these social media platforms, they're they're the new town square, and in the old town square, you knew that the person standing on their soapbox was probably a voting citizen alongside you, and so you cared a lot about what they said.

**[1:05:10] Aidan Gomez:** In the digital town square, everyone is much more skeptical of the stuff that they see. It you don't just take it for granted. We also have methods of confirming humanity, like human verification on social media platforms is a thing, and we need to support it much more thoroughly so that people can see, is this account verified? Is it actually a person on the other side?

**[1:05:35] Nilay Patel:** what happens when the humans start using AI to generate lies at scale, right? Like me posting an AI generated image of a political event that didn't happen is just a, uh, I think just as damaging if people believe it, is thousands of robots doing it.

**[1:05:49] Aidan Gomez:** what I was talking about was when you can have a single entity creating many, many, many different voices saying the same thing to present consensus. You can stop that by just preventing fake accounts and just having with each account, there's a human verificate- verified ID behind it, so you know it's another person on the other side. Um, and that that stops that scaling of millions of fake accounts. On the other side, what you're describing is fake media.

**[1:06:18] Aidan Gomez:** I mean, there's already fake media, there's Photoshop. We we've had this tech for a while. I think it becomes easier to create fake media, and there's a notion of media verification. But you also you're gonna trust different sources differently. If it's your friend posting it who you know in the real world, you trust that a lot. If it's some random account, y- you don't necessarily believe everything that they claim. And if it's coming from a government agency, you're gonna trust it differently.

**[1:06:48] Aidan Gomez:** If it's coming from media, depending on the source, you're gonna trust it differently. And so we know how to assign appropriate levels of trust to different sources. So I I think it's it's definitely a concern, but it's one that is addressable. Humans are already very aware that other humans lie.

**[1:07:08] Nilay Patel:** yeah, I think you're being very optimistic about that. I look at Facebook and it's full of tens of thousands of pictures of Jesus made of spaghetti and people are happily liking those pictures, and there's nothing about that that seems like a person. I wanna ask you one last question. It's the one I've been thinking about the most, and it it come, it brings us back to where we start. We're putting a lot of weight on these models, business weight, cultural weight, inspirational weight, right?

**[1:07:32] Nilay Patel:** We we want our computers to to do these things, and the underlying technology is these LLMs. Can they take that weight? Can they withstand the burden of our expectations? That's the thing that is not yet clear to me. There's a reason Cohere is doing it in a targeted way, but then you just look broadly and there's a lot of weight being put on LLMs to get us to this next place in computing. You, you were there at the beginning.

**[1:07:58] Nilay Patel:** I'm wondering if you think the LLMs can actually take the weight that's being put on them, take the pressure that's being put on them.

**[1:08:04] Aidan Gomez:** I think we'll be perpetually dissatisfied with the technology, and I think that the technology will be continuously rising to the occasion. So I I don't think there's some sta- like, if you and I chat in two years, we're gonna be, like, disappointed that the models aren't inventing new materials fast enough to get us, you know, whatever, whatever. So I I think that we will be always disappointed and wanting more because that's just part of human nature.

**[1:08:38] Aidan Gomez:** But I think the technology will, at each stage, impress us and rise to the occasion and surpass our previous expectations of it. But there's no point at which people are gonna be like, we're done, like, we're good. It's Well, it it's

**[1:08:54] Nilay Patel:** it's not done. I I don't think the question, I'm I'm not asking if it's done. I'm saying, do you see, as the technology develops, that it can withstand the pressure of our expectations? That it has the the capability, or at least the potential capability, to actually build the things that people are expecting to build?

**[1:09:12] Aidan Gomez:** absolutely, I think it it will. There was a period of time where everyone was like, the models hallucinate, they make stuff up, they're never gonna be useful, we can't trust them, never gonna be useful. And now, hallucination rates, you've you can track them over the years, they've just dropped dramatically, and they've gotten much better. I think with each complaint or with each fundamental barrier, like all of us who are building this technology, we work on it and we improve the technology and it surpasses our expectations.

**[1:09:40] Aidan Gomez:** I expect that to continue. I see no reason why why it shouldn't.

**[1:09:43] Nilay Patel:** Do you see a point where hallucinations go to zero? That to me, that's when it unlocks, right? You can, you can start depending on it in real ways when it stops lying to you. Right now, the models across the board hallucinate in, uh, honestly, hilarious ways. But there's a a part to me anyway that says, I can't trust this yet. Is there a point where the hallucination rate goes to zero? Can you see that on the roadmap?

**[1:10:07] Nilay Patel:** Can you see some technical developments that might get us there?

**[1:10:11] Aidan Gomez:** like you and I have non-zero hallucination rates.

**[1:10:14] Nilay Patel:** Well, yeah, but no one trusts me to run anything. Like, Oh, well, as soon as I mean, I I sit here asking the questions, you're the CEO, like, fine, but I'm saying, computers, if you're gonna put them in the loop like this, you you want to get to zero.

**[1:10:28] Aidan Gomez:** no, like, I mean, humans, like, uh, they misremember stuff, they make stuff up, they they get facts wrong. If you're asking, like, whether we can beat the human hallucination rate, I think so, yeah, definitely. That's definitely an achievable goal, cause humans hallucinate a lot, like, you know, misremember stuff. I I think we can create something extremely, extremely useful for the world. Um,

**[1:10:52] Nilay Patel:** Useful or trustworthy? I I think that's what I'm getting at is there there's trust, and I think the the amount that you trust a person varies, sure, some people lie more than others. The amount that we have historically trusted computers has been on the order of, uh, a lot. And with some of this technology, that amount has dropped, which is really interesting. And I think my question is, is it on the roadmap to get to a place where you can fully trust a computer in a way that you cannot trust a person?

**[1:11:19] Nilay Patel:** We trust computers to fly F22s because a human being cannot operate an F22 without a computer, right? If you're like, the F22 control computer is gonna lie to you a little bit, like, we would not let that happen. It's weird that we have a new class of computers, we're like, well, trust it a little bit less.

**[1:11:36] Aidan Gomez:** yeah, yeah, like, uh, I mentioned before, I do not think that large language models should be like prescribing drugs for people or doing medicine. But I promise you, like, if you come to me, Aidan, with a set of symptoms and you ask me to diagnose you, you should trust Cohere's model more than me. Uh, they like, it knows way more about medicine than I do. Like, there there's just whatever I say is, like, gonna be much, much worse than the model.

**[1:12:05] Aidan Gomez:** A and that's already true, like, just today, like, in this exact moment. At the same time, neither me nor the model should be diagnosing people, but but it is more trustworthy, right? Like, you should genuinely trust that model more than this human with that use case. And and so I in reality, who you should be trusting is the actual doctor that's done, like, a decade of education on on all this stuff. So the bar is here, Aidan's here, the model is, you know, slightly above Aidan.

**[1:12:33] Aidan Gomez:** We will make it to that bar, I absolutely think. And and at that point, we can s like, put the stamp and say it is trustworthy. It's actually as accurate as the average doctor. One day, it'll be more accurate than the average doctor. We will get there with the technology. There there's no reason to believe why why we wouldn't, but it's continuous. It's not a binary between you can't trust the technology and you can. It's where can you trust it?

**[1:12:59] Aidan Gomez:** Cause right now in medicine, you can't, like, we shouldn't, like, really rely on humans, but in other places, you can, right? Like, when there's a human in the loop, it's actually just an aid, and it's like this augmentative tool that is really useful for making you more productive and and doing more, or having fun, or learning about the world. And so there's places where you totally can trust it effectively and deploy it effectively already today. That space of places that you can deploy this technology and and put your trust in it, it's only gonna grow.

**[1:13:31] Aidan Gomez:** And to your question about, you know, will the technology rise to the challenge of all the things that we want it to do? I really deeply believe it will.

**[1:13:42] Nilay Patel:** yeah, well, Aidan, that's a great place to end it. Thank you so much for spending all this time on Decoder. This was really great.

**[1:13:46] Aidan Gomez:** Yeah, thank you for having me.

**[1:13:51] Nilay Patel:** I'd like to thank Aidan for taking time to join Decoder and thank you for listening. I hope you enjoyed it. If you'd like to let us know what you thought about this episode or anything else, drop us a line. You can email us at decoder at the verge dot com. We really do read all the emails, or you can hit me up directly on Threads. I'm at at Reckless 1280. We also have a TikTok, for as long as there's a TikTok, it's at Decoder Pod.

**[1:14:07] Nilay Patel:** It's a lot of fun. If you like Decoder, please share it with your friends and subscribe wherever you get your podcasts. If you really like it, hit us with that five star review. Decoder is a production of The Verge, part of the Vox Media podcast network. Our producers are Kate Cox and Nick Statt. Our editor is Callie Wright. Our supervising producer is Liam James. The Decoder music is by Breakmaster. Cylinder. We'll see you next time.

**[1:14:32] Ad read:** As soon as autumn rolls around and you swap your iced lattes for pumpkin spice, that's your signal to start thinking about the holidays, the decor, the gifts, the shopping lists, but more importantly, the holiday treats, cheesecakes, cookies. Is there anything better? And when the recipe calls for cream cheese, there's only one option Philadelphia, the original cream cheese, made with fresh milk and real cream. This holiday season, whether you're going all out or keeping it simple, make it really, Philly good.

**[1:15:10] Ad read:** Find your favorite holiday recipe at creamcheese.