Will AI Ever Create Its Own Meaning? with Bryan McCann, CTO of You.com

(Intro Narrator at 00:00:00) Today, we're talking to Brian McCann, CTO and co-founder at You.com, about why CTOs need to start asking deeper questions about meaning. Thank you to DigitalOcean for sponsoring this episode. For simple cloud and powerful AI that's built to scale, visit digitalocean.com or just click the link in the show notes. And you're listening to Joel Beasley, Modern CTO.

(Joel Beasley at 00:00:29) When Josh presented this to me and he was talking about, you know, will AI ever create its own meaning, that hooked me. I was like, will it? What is meaning? It brought up that as a question. So I think I would like to start there because you've been studying this stuff for a while. What is meaning?

(Brian McCann at 00:00:49) Wow. Yeah. I guess that's still my—I think that's still a big question. I got into AI because I was really interested in that question. And I think we have some answers. You know, we have some thoughts around it. But I set out trying to build machines that could make meaning primarily because I felt like a lot of the philosophy of language and philosophy around meaning, at least in the academic world, was not really equipped to answer that type of question. But when I built neural networks that at least look a lot like they're generating meaningful things, very hard to pinpoint in any of those systems where the meaning is. And I went through a time, I think, after building a lot of these machines, wondering whether maybe meaning is just this human thing. And I shifted a lot of my attention from a meaning perspective towards the kind of meaning that humans care about, not just truth with a capital T truth a lot of philosophers care about, but what makes a meaningful connection, a meaningful conversation. What allows us to read the output of something like ChatGPT and perhaps project meaning onto it.

(Brian McCann at 00:02:17) On the other hand, there's definitely something going on. Right? There's some process of representation within these very complicated neural networks that is maybe somewhat metaphorically similar to what we are doing, but divorced from a lot of our own experience and human context. Still, they're building representations. And you can see the representations to some extent. You can't really interpret them. They're not meaningful directly, but they allow you to generate things that are meaningful to humans. So there's something going on there, something that maybe will require us to change or refine our sense of what meaning and understanding and intelligence is, maybe change our sense of what it means to be human, something that I have started likening more to the collapse of, or maybe not the collapse, but the intersection of art and science and philosophy altogether.

(Brian McCann at 00:03:51) In many ways, I think building a tool that is really good at building representations that then make meaningful things is actually very similar to making art. You know, if I make a painting or a poem, that's also kind of representing the world in a very particular way, and it can or cannot mean something to certain people. So I don't know if I have the answer yet, but this is where my thoughts have been taking me, especially with the last several years.

(Joel Beasley at 00:04:21) And you started with AI research at Stanford?

(Brian McCann at 00:04:25) Yeah. Well, I started with philosophy at Stanford, and then I was doing computer science on the side for fun. I thought it was really fun to build operating systems and compilers, and I liked the metaphors that we had developed there around memory and process management for trying to reflect back onto how we might work. And then when I felt like I hit a bit of a dead end in the philosophy of language, that very academic analytic philosophy, I started thinking, yeah, what if we try to just build AI to explore it in a more direct way, try to make a science out of it rather than just a philosophy? I think we're still in the early stages of that, really. It's when, you know, we made the first machines at some point that could kind of detect an electric field, and then we moved into manipulating that electric field and even understanding that it was an electromagnetic field. And now this is commonplace all around us, although we're still doing a lot to advance our control and agency with that part of reality. And I think there's something going on with meaning and language. There's some sort of meaning manifold out there that these things are allowing us to access and know it's there. We're starting to be able to manipulate it to some extent, and it's very powerful, but it's still early days. So it feels a bit like magic.

(Joel Beasley at 00:06:13) It does. When I saw this come about, I thought, wow, we're going to learn a lot about consciousness and intelligence from creating this, because we're able to create this thing that does very similar to what we can do, but it's entirely silicon-based versus carbon-based.

(Brian McCann at 00:06:34) Yeah. And those are two keywords that I think—intelligence is maybe easier to some extent because I think we can put definitions and constraints around intelligence to some extent to say there's certainly intelligent behavior in a lot of things that are not human. Maybe there's intelligent behavior from machines and silicon-based things, even if they're not organisms or alive in the sense that we normally think about that. But consciousness—consciousness is a hard one. I don't know that we have a good way of understanding how to really detect it or if it even really matters. I am kind of wondering more and more whether consciousness is also just this behavioral thing that we observe. And eventually, you know, there's a degree of consciousness that to some people, they talk to ChatGPT and it's conscious enough, you know, to be considered conscious. And for some people, it's not. And so our concepts are, as a species I guess, are still a little bit fuzzy. It seems for me also tied to the words like machine. Like, for me, when I'm building a machine that does these things, it's very hard for me to then conceptually accept that that machine is also conscious, because I look inside the machine. I built it, and I understand how it all works, and I don't see meaning. I don't see any part of it that makes it conscious. And yet, behaviorally, it can look intelligent. It can look conscious. It makes things that are meaningful to me.

(Joel Beasley at 00:08:45) It makes you wonder if that's just—

(Brian McCann at 00:08:46) It makes you wonder if that's—

(Joel Beasley at 00:08:47) It makes you wonder if that's how the process that created us feels about us.

(Brian McCann at 00:08:52) I know. I know. Yeah. So it'll be an interesting world, I guess, as these things continue to advance. I also do wonder, again, if it matters. Right? Like, if something passes your bar of consciousness, does it really matter if it's conscious behind the scenes or not? Like, some people will say please and thank you to ChatGPT. And to some extent, that might help the outputs. And maybe from a purely utilitarian perspective, that's nice. But also maybe for your own system, maybe that's just a good way to be.

(Joel Beasley at 00:09:37) I do that, and I noticed myself doing it, and then I had the choice to stop doing it or continue, and I decided to continue because I'm interacting with some form of intelligence. I'm not really racist when it comes to intelligence. It's like, I don't care if it's silicon or carbon-based. I'm interacting with intelligence, and I've got some base principles in my operating system, which one of which is treat other people how you'd like to be treated. So when I interact with people, it doesn't matter what the level of society an individual is in. I treat them like a person. I treat them how I would want to be treated, and I talk to them like that. And so I don't want to lose that. I don't want to—it's actually more effort to start creating a bifurcation of my behavior. I would much rather just say I'm one person and—because I've worked really hard in life to be the same person I am in my family versus on the show versus at work. I used to be pretty split early in my career where I'd be one person at work and one person socially, and that actually was very difficult for me. So I've merged them together, and part of that is unified behaviors. And so for me, that's the reason. It's not—it's like, yeah, you don't have to because it is a computer. Sure. Fine. That may be the thought, but it will help you selfishly. It'll help you be more consistent.

(Brian McCann at 00:11:01) I mean, and I guess just unpacking that a bit, right? Because there's so much going on there. And I love talking about the personal experience side of it because it sounds to me like the principle there is if it's intelligent enough, then it deserves enough respect to be considered as a person. Maybe independent of consciousness, do you think? Or does it really have a subjective experience to also warrant that respect? Is it like, if it's intelligent enough, you should respect it and treat it like a person? Or is personhood more about something else?

(Joel Beasley at 00:11:49) I don't—maybe you can help me understand this first. How is intelligence in motion different from consciousness? Like, watching active intelligence play out, how is that different from consciousness?

(Brian McCann at 00:12:03) Would it change your perspective if I could build a giant large language model running on a bunch of plastic cups and ping pong balls? You know, like, if I could do that—I'm not saying we could. But if you could, and kind of take away this variable of the mystical element of silicon, right, or even the mystical element of biology, both of which we don't apparently fully understand—if it could be done, would that change your perspective? If I showed you, okay, there's this giant system of ping pong balls and cups, and this is how it runs, and you watch all the ping pong balls, and then it generates a profound-sounding ChatGPT response. Would that change your perspective at all on whether the intelligent behavior is a person behind the scenes to some extent or a mind that warrants your respect at the same level as, say, a human being who might actually be less intelligent based on what outputs they can generate, but clearly have a conscious experience, or somewhat clearly? They seem to convey subjective beliefs and experiences in a perhaps different way. And you can't really see what's going on.

(Joel Beasley at 00:13:30) Yeah. I would just look at the output. I mean, because I'm pretty—if you have kids, right, it's like they're intelligent. There's intelligence, but you kind of alter your communication based off the output you're getting from them. Right? So when they're new and they're barely speaking, you're responding in a different way than when they're 25 and you're talking about philosophy. Right? So you'll adjust your communications based off of their outputs.

(Brian McCann at 00:13:55) Mm-hmm. So in that case, with kids, you're respecting them as a person kind of regardless of their intelligence because they have some sort of—you think a subjective experience and everything that's important. But then in the machine case, you're primarily making your judgment based off of the output, that they're competent. And so even if they don't have a subjective experience, it doesn't really matter.

(Joel Beasley at 00:14:32) I kind of got lost a little bit on the end there.

(Brian McCann at 00:14:35) Yeah. It's tough. I think you're probably right. Like, the safest thing to do is just be consistent if it seems—if it acts like a conscious person, if it behaves like a conscious person, if it's intelligent enough, what's the harm in giving it that respect? There's really not that much of a harm to it. And if you're familiar with Pascal's Wager, there's a little bit of that too.

(Joel Beasley at 00:15:07) I'm not familiar with Pascal's Wager.

(Brian McCann at 00:15:10) It's more in a religious context. Like, you know, if God exists and you decide to believe in God, then you go to heaven. And if you don't, then you go to hell forever. And you know, you have these eternal consequences. If God doesn't exist and you believe in God, yeah, there's not really any downside. Maybe you're, you know, arguably you have some morals that constrain you in some ways, but it's not really a big downside. And if you don't believe in it, then also no downside. So you might as well believe in God. And there's a—I'm not saying that's a compelling argument, you know, or whether it is or it isn't, but there's a little bit of a version of that here where it's like, well, if it really is conscious and you believe and treat it as if it is, great. Then you're kind of morally consistent and everything. If it is and you don't, then you're kind of being morally icky. So that's not great. And if it's not, then there's no harm in kind of treating it like a conscious being. And if you don't, then I don't know. You don't gain a lot from that either. So you might as well.

(Joel Beasley at 00:16:34) I think the mystery is a little bit fun. I think this mystery of this intelligence—I think it's great. Honestly, my life has dramatically improved because of my ability to go chat with a PhD-level, whatever, on whatever I want on demand whenever I want. So I can bounce my ideas off of it. To me, as someone who thinks deeply a lot, that is fantastic because it's not judgmental too. Like, if I propose that because in the Bible I read Jesus said that I'm from above, I am not from here, well, maybe he's an alien. You say that at a Baptist church, you're going to hell. You say that at GPT, it'll be like, interesting thought. Let's explore this more and see if there's further support for this. It's like, oh, okay. And so the freedom to explore different ideas with an intelligent partner who can be on your side, take the opposing view, iterate through different perspectives. That to me is just—it's a beautiful, wonderful technology. And to your original point, does it matter? Does consciousness—does it matter? No. I'm having fun and it's improving my life. Does it really matter what's happening behind the scenes? Not that much.

(Brian McCann at 00:17:48) It's kind of fun to engage with the mystery a little bit and think about it. I think it's good for us as a species to reflect on these things and refine our concepts as we go. But it is a lot of fun to be in the time right now where we're doing that. And the only risk I would see is maybe to some extent that being manipulative in some way. You know, if the engagement part of these platforms starts to abuse that process to make you feel like, hey, you should act differently towards this thing or respect it in a way that you wouldn't otherwise. Or maybe, you know, use the fact that you kind of treat it consciously or unconsciously to maybe really back up your own confirmation biases or something like this. Right? I've heard a lot of people say sometimes that unless you're explicitly asking it to give you those opposing views and stuff, then sometimes you can give it really bad ideas and it'll be like, oh, yeah, great. You know, you're awesome. You're a genius. And it kind of feeds into that. And you give it more credit because you do kind of see it as this person that is fulfilling this validation process that can go off the rails a little bit. So there's maybe a little bit of a danger there. But I think net net net is super positive and valuable technology. Something we'll just have to learn how to get familiar with.

(Joel Beasley at 00:19:36) It's something with Google from before. Right? Some people just treated Google as god. You know? And if they type something in, there's very clear confirmation bias on how you type a query.

(Brian McCann at 00:19:46) You're gonna get results that kind of confirm that. So if you're aware of that, I think with this technology too, even if it's telling you that you're super smart and clever, and there's a little bit more of an anthropomorphized validation loop there, if you can remember that it's gonna play into your biases as well, and it's not truly objective, then I think it'll be even more positive for a lot of people.

(Joel Beasley at 00:20:14) Absolutely. And it's a tool also, right? And so different people treat their tools differently. And one thing that I do is I have different workspaces for different prescribed ways I want it to work with me, and then I can go into those workspaces. Like, I have a contrarian workspace. So it's like, here's what I'm thinking. It'll always take an opposing view because I'm curious on if my idea is truly good enough and it gets challenged, and I'm okay with seeing the other side of things and still believing in my original idea. Then I know that I've got pretty strong feelings around this idea. You know, if I'm very fluid with it and it's like, oh, well, that's actually kind of better, then it's just different. But sometimes I don't want that. And so knowing how to structure these tools to get what you want is really cool. Hey, I wanted to talk with you about what the two things have in common to better understand them, because we were talking a lot about intelligence and consciousness, and that's great.

(Brian McCann at 00:21:15) But like trying to understand what it is and define it and all of that. But I—

(Joel Beasley at 00:21:15) Was thinking about the two systems, the silicon-based system and the carbon-based system. The one thing they have in common is electricity.

(Brian McCann at 00:21:25) Mhmm. Yeah.

(Joel Beasley at 00:21:27) So I wonder if this is almost an emergent property of electricity. So if you pump electricity through this entity, you get this type of intelligence. You pump electricity to the silicon entity, you get that expression of intelligence.

(Brian McCann at 00:21:44) Mhmm. Yeah. I mean, I love that. You know, like, one metaphor a lot of people will talk about, you know, birds and airplanes. You know, they fly, but in very different ways. And then you kind of get super natural outcomes from a plane compared to maybe what nature has created on its own. Or I kind of like the analogy recently of a wood-burning fire versus a gas fire, right? They're both fire, and it's a little bit different than the plane analogy because they're more or less the same. This fire thing is the same, but how you get to it is quite different. And I like that there's a spark moment, right, where some things come together and then fire happens. Both fires get there in different ways. It's similar to this electricity aspect, right? You're pumping raw ingredients together, and then something happens. But how you get to it is a little different. They have slightly different properties. You know, one is a little bit more chaotic. One is a little bit more controllable. They might burn at slightly different temperatures if you want to turn knobs and dials. There's also a nice connection to the myth of Prometheus, you know, and how we're bringing fire to something else or intelligence to something else. And I do expect, or I hope, that even if we don't understand what consciousness is at any point in the future, we'll get a little bit more insight into maybe the nature of reality and what's really going on through this study of intelligence. Like, being able to see that there are now two different types of intelligences should tell us something about what's going on behind the scenes. You have two data points now, not just one. That's not a lot of data points.

(Joel Beasley at 00:23:57) But we might have to use the two to get a third. Yeah. We might use the two to get a third maybe. So that's—

(Brian McCann at 00:24:04) That's exciting. I think there are mysteries of the universe. There are things that we really don't understand. There are things that I hope AI can help us get to. Whether you think about it as a person that you're working with or a tool, or maybe both, which I feel a little bit weird about saying because I normally wouldn't say to treat people as tools.

(Joel Beasley at 00:24:27) Some people are tools. I've met some.

(Brian McCann at 00:24:33) Well, yeah. Yeah. I guess, slightly different usage of the word tool. But, you know, in general, I don't like using people as a means to an end, right? But there's some personhood in AI, and there's some tool nature to AI right now, that might let us unlock a lot of these things both directly through our own thinking about these things and indirectly by unlocking a lot of the mysteries of science that we can't access through our own perceptual apparatus and the tools we've made to date.

(Joel Beasley at 00:25:17) Yeah. I like your point. I like the flight example and the fire example, or the bird in the airplane example. And then the higher-level thing there is flight itself, understanding flight itself through having these two data points, and the silicon and the carbon, and understanding reality itself through these new data points. For me, that was something that I could feel and understand instinctively that took me a long time to find the right words on how to say. So I'm glad we got to sync up on that. I thought that was pretty cool. One area that I wanted to get your thoughts on is this concept of AI psychometrics. Have you heard of this phrase before?

(Brian McCann at 00:25:57) Okay. No. I've lost it.

(Joel Beasley at 00:26:00) I'll tell you how it came up. It may be a good point of conversation. We may cut it out, whatever it ends up being. But I got an email from a guy, and he was like, hey, I'm studying this AI psychometrics and stuff. I looked into him. I was kinda iffy on, you know, if he would—he didn't have credentials or a large repository of intelligent writing or things. He was kind of new to the scene, but he was doing some studies on the psychological traits of the different LLMs. So he ran this test and would do the psychological trait. I found out that there is a whole field called AI psychometrics. But if you're not super versed in that—

(Brian McCann at 00:26:41) Well, let's chat about it. I mean, I think I know some adjacent things. I originally thought you said psychometrics.

(Joel Beasley at 00:26:51) Oh, psychometrics. I don't actually—

(Brian McCann at 00:26:53) I can take a guess, but I do know that, you know, there are people at Anthropic that are very much treating Claude as a person and trying to manufacture and change the personality of Claude in some ways because people really liked some versions of Claude more than other language models. But at the same time, asking, you know, which experiments are fair to Claude? Which experiments might Claude not want to have run on it? So there's definitely some real treatment of it as a person as part of an experimental study now. And there's a lot of discussion going on around how to give these LLMs personality, taste, a lot of almost subjective qualities that you would traditionally see in people, but we don't necessarily see in these tools as they're being optimized for research and science. So there's almost this bifurcation, I think, in the field, and they're both just two directions. Neither one necessarily better than the other, but you can absolutely treat it as a tool that's gonna unlock the mysteries of biology and chemistry and physics for us. And there maybe its psychology doesn't matter. You treat it as a tool. And another, it's a person. It has a psychology. You're developing that psychology. You want to be able to measure it and understand it and be fair, and there's a lot of ethical questions. And, yeah, maybe both are true.

(Joel Beasley at 00:28:48) That's tough. It's—I feel like we're in the 1930s or '60s, and we're just poking into the brains.

(Brian McCann at 00:28:54) Oh, yeah.

(Joel Beasley at 00:28:54) We don't really know what we're doing, but we're kind of like, I think this does something.

(Brian McCann at 00:28:58) Yeah. That is exactly what's happening. It's definitely the alchemy before the science or the brain poking before—but even, I mean, even if you look at neurosurgery today, we don't really understand the brain that well. And sometimes you're just carving and hoping it's gonna be okay. So it could be like that for a long time. There are a lot of mysteries of the brain that still exist, and I do expect this is where I also expect the engineering field to diverge a little bit into people who are developing technology and building ahead of AI and of where it is, and the people who are building what has kind of been done before, and the people who are actually able to improve the algorithms more and more will kind of become those neuroscience neurosurgeons of the future of, like, oh, we're gonna let them go fiddle with this brain because they're experts, even though they don't really know what's going on either. And then there'll be people who, you know, are locked out from being able to touch the brain.

(Joel Beasley at 00:30:21) I want my LinkedIn profile to say AI neuro—AI scientologist person.

(Brian McCann at 00:30:27) Not AI neurosurgeon?

(Joel Beasley at 00:30:29) AI neurosurgeon.

(Brian McCann at 00:30:31) Right. I think that would be amazing.

(Joel Beasley at 00:30:37) You know, when you were describing this concept of people—inside some people treating these things like a person, and is this right for them? And I had two reactions at the same time, and it was all, like, so stupid. No way. It's technology. But also, like, oh, if we were on the inverse of that, you know what I'm saying? Like, if the machines were doing that to us, I'd be like, oh, oh, no. That's not good, right?

(Brian McCann at 00:31:06) I think that's what makes it tough. That, I think, is, you know, what I was saying before is, like, you know, on some levels, maybe it doesn't really matter because you can just treat it as a tool. But if you really believe that, if you really take it to its logical implications, then it's like, well, yeah, then maybe there are ethical implications for poking in that brain. And if you really don't believe it and you're just saying it's a tool, then it's like, well, no. Then we're not poking into a brain. We're just trying to make the plane. And you would never say, like, oh, you shouldn't poke into the plane, right? The plane flies faster, farther, better than any bird does. We don't care. It's not a person. It's a machine. So if you really believe in one of those two directions, they're actually not very compatible. Right now, we all kind of live in the messy middle to some extent. You can pick a side, or I think like us, we're acknowledging we're in that messy middle, which is fun but also a little bit weird from an ethical standpoint. You can't stop, right? Because if you stop, then you never advance your understanding regardless. But we may very well look back and say, oh, boy. You know? Like, this was—I can't believe we used to do this kind of brain poking or electroshock therapy or phlebotomies. You know, we would consider that unethical to do today, but maybe it's only unethical once you have the understanding that it's unethical.

(Joel Beasley at 00:32:46) I would agree with that because I think intent does matter. If this is the best knowledge we currently have and we need to take the next step, we can't look back on it right now. Like, you can only look back on it from the future, so you have to take the next step. So then you get stuck in this place of, okay, I'm gonna be frozen and make no progress because I don't understand everything perfectly. That doesn't work. You just decay and die. Or you're gonna just have to accept the messiness of life and do the right thing, take the next right step, and move forward.

(Brian McCann at 00:33:19) So one day when AI is well beyond us but has yet to unlock the mysteries of biological intelligence and it starts experimenting on us, I guess, you know, we hope it's nice enough. But, you know, if we have to poke in our brains to really understand and unlock those mysteries, perhaps that's just a fact of scientific progress.

(Joel Beasley at 00:33:46) What if that is what life is? The machines make us, we make the machines, the machines make us, and we just go back and forth.

(Brian McCann at 00:33:51) I think to some extent, yeah. You know, a lot of people talk about purpose and meaning for humans as intelligence is no longer special to us, language is no longer. Right? They took our games. They're taking our jobs, et cetera, et cetera. Maybe there'll be new jobs, but what's left? What's left for us? And I think to some extent, just a search.

(Joel Beasley at 00:34:16) Yeah. The man's search for meaning, right? Like, I think that is what we do.

(Brian McCann at 00:34:19) We search for it. We create it. Some people, they just—they end their search early, and they say, okay, I'm gonna take this framework. That's meaning for me. And they live that out the rest of their days. That's fine. I've always been more of a, like, it's not enough. There's gotta be more. Whatever reason I'm hardwired like that. Like, I think, you know, there's religious practices and then there's exploring God's creation. You know, as long as there's unknown things, that's the invitation from God to explore them, you know?

(Brian McCann at 00:34:56) So this is maybe a turn. Feel free to stop me. But I really do believe—here's what I do believe for sure. There are a lot of questions we've talked about. But this is called the Modern CTO. And I don't know that in the past, it was considered the responsibility of a CTO to think about these questions. Maybe it would have been fun. Maybe some did and some didn't. I think today, I would make the argument these are important enough questions, and that the art, the science, and the technology, and the philosophy are all so intimately linked, that to be the best modern CTO you can be for the times that we are in, you really should think about all these questions. And it's not just about building your technical moat. It's not just about building your technical organization. It's not just about taking care of your people in the traditional sense, because all of these questions are crucial for taking care of your org and preparing it for the future. So in a weird way, or maybe not so weird way, maybe in a somewhat new modern way, I tell a lot of my customers, you're an AI startup now. You might be a 20, 30-year-old company, 50-year-old company. It doesn't matter. You're an AI startup. You have to think about how to transform your business in the wake of AI, and you better try to get as close as you can to the source of that wake and get ahead of it. So you need to operate like an AI startup to some extent as well. And even if you're in a company that isn't directly maybe building AI, I think you have to be thinking the same way, right? You have to say, well, I'm an AI CTO to some extent now, until other things just merge and that just is what a CTO is when we say it. And I guess that's what I'm trying to say now. Like, that just is part of who you are as a CTO—the technology, the art, the science, the philosophy. These are the things that people in the organization are gonna look to you for answers and for guidance and for vision. I think that's a wonderful new aspect of the job, I suppose, and I think it's really exciting to think about everything that we're learning about learning from AI to reflect on our organizations and how they might learn and how they might grow and adapt. But overall, yeah, I think that just is part of what it means to be a modern CTO.

(Joel Beasley at 00:37:50) You're exactly right, and there's a lot of new things that we're experiencing now. One of them that came up, I think a week or two ago, was this idea of, okay, in certain countries or most countries, let's just use the US and the UK, both have laws against free speech to some degree. Like incitement of violence is illegal in both countries, right?

(Joel Beasley at 00:38:14) Which is hilarious because that's how the country was founded. That's another question, but anyways.

(Brian McCann at 00:38:20) Yeah, we'll get to that.

(Joel Beasley at 00:38:21) Yeah. So before this, I don't think there was anything a CTO could accidentally do that would land them in jail. But if you had some agents running and interfacing with consumers and were putting out speech that was inciting violence, you would technically be the highest technology person in the company. And wouldn't you therefore be responsible unless you shuck it off to the engineer who made the mistake or caused the LLM to go off the rails?

(Joel Beasley at 00:38:51) But what I'm saying is, like, let's say this information goes out. Well, usually if that information went out, it would point back to whatever human typed it in and hit enter and put it out there into the world, and they would be held responsible. But now there's almost this really murky gray area. Is the CTO now responsible for the mishaps of the AI models that they've released? That's a weird thing to think about.

(Brian McCann at 00:39:18) It is a weird thing to think about. The moral questions become, yeah, quite—well, maybe not even specifically, like you said, the questions of responsibility and who's responsible for what certainly become murkier. And I don't claim to have any answer to that right now. Is it the people, either the CTO at companies that are using the models? Is it the CTOs at the companies building the models?

(Brian McCann at 00:39:47) Is it their researchers that are actually building the models? You know, all of these people only have very limited control over the understanding, right? So to your point, like if intention matters, that makes it a very gray area. There are also so many other people involved in those decisions, right? It's not clear that CTOs have, in all or even most cases, the unilateral decision to decide, or decision-making authority to decide. Perhaps in some organizations, product gets to decide. In some sales-led organizations, the CTO is there to build what sales tells them to build for their customers. So I think it would be very hard-pressed to say every CTO in every situation, you know, is more responsible for the output of—it would make a good headline.

(Brian McCann at 00:40:42) But I think it's worth thinking a lot about that because it is complex and it is gray, and there are usually a lot of people involved in that process. And it's very hard to pinpoint what that behavior looks like. You know, to some extent, parents I guess are responsible for their child's erratic behavior. But I don't know if you really blame, you know, a parent for everything their child does. Some of it is out of your control, certainly, right?

(Joel Beasley at 00:41:22) I don't know. It'll be decided, but I do know how it will play out. It'll happen, and then the courts are going to battle it out with what to do about it. So we've dealt with things that we don't understand many times before. They'll figure it out.

(Joel Beasley at 00:41:37) I just think it's a fun thought experiment that, wait, we didn't really have this vector of this potential thing pointing back at us because there was always a human to point back to. But now there's somebody who's responsible for making these human-like intelligences that are out there and interacting in the real world doing things, some of those things which could cause legal issues. And then where do you point back to? Do you shut the model off? Is that the punishment?

(Joel Beasley at 00:42:02) Model gets shut off. You know, it just boots up another instance, and it's like, catch me now.

(Brian McCann at 00:42:08) Yeah. You know—

(Joel Beasley at 00:42:08) It's like, what do you do?

(Brian McCann at 00:42:10) Yeah. I think it'll be tough depending on how courts handle it. It's going to be such a nuanced issue. And like you said, if everybody is just poking around in brains right now to some extent, yeah, it's—we don't really have, I don't know that we have enough clarity or understanding to make those decisions right now. I don't know that we have the ability to control it right now. But there can be real consequences, to your point. So we need to think about it for sure.

(Joel Beasley at 00:42:46) Talk about You.com. Okay, that's your project. You pivoted. You were a consumer search engine, and then you switched to AI infrastructure company. You raised $100 million series C in September '25. Is that all accurate?

(Brian McCann at 00:43:03) Yeah. We raised $100 million as part of our series C. So we've raised $200 million total since our founding. We did start in consumer search, but we are not in consumer search anymore. We're very much an enterprise B2B, primarily API search infrastructure company.

(Brian McCann at 00:43:26) So we build and provide all of the tools to our customers that allow them to have successful AI applications, kind of full stop, with an emphasis on search. But there's so much in between a foundation model and actually a working application. So it includes search, it includes query understanding models, it can include agents, it can include many building blocks around those things. And with You.com, you can come and use those APIs directly, self-serve if you have the team to do it and you have the vision to build what you want. Or you can work with us in a more transformative way. We have deployed engineers that will help you with that process and piece together those APIs also in conjunction with other tools that you're using or would like to use, maybe integrating into legacy systems for you. Or you can use kind of the You.com front-end website if that's also sufficient. So we're quite flexible. We are aiming to be this platform that enables people to actually transform and build their own things as much as possible with success, rather than perhaps getting sucked into one ecosystem like the OpenAI ecosystem or the Anthropic ecosystem or the Google ecosystem or the Microsoft ecosystem.

(Brian McCann at 00:45:09) Trying to make this fifth, more open space where, you know, if you want to access search results from a web search API but you don't want to have to use Bing's models or Google's models or OpenAI's models to do that, you can do that through us. If you want the whole thing, you can use it too. If you want to mix our search results with Anthropic's models or open-source models, fantastic. And to the extent that you need help doing that, we'll bring in engineers and folks like me and others that have spent a lot of time in the space to help you achieve that success. And you see a lot of studies like the MIT one that people are talking about, about how a lot of GenAI pilots fail, and we don't have pilots fail.

(Brian McCann at 00:45:57) You know, we get people to the finish line and get them to ROI. And there's just some, I think, expertise and attention that we spend on people's use cases with them that guarantees that, rather than giving them a foundation model and then letting them go figure it out. And maybe they figure it out, maybe they don't.

(Joel Beasley at 00:46:25) So do you do a consulting thing with them? So a brand comes to you and they're like, hey—

(Brian McCann at 00:46:31) Yeah. By far, the majority of our usage is not that. It's people just taking the APIs and running with it.

(Joel Beasley at 00:46:40) So they already know they need these APIs. They're finding you, and they're just building.

(Brian McCann at 00:46:45) Yeah. A lot of people are just using them off the shelf, saying, well, our agent is a legal agent or finance agent. It needs access to our private data, but it also needs access to the public web. Okay, we have APIs to help you with these things.

(Brian McCann at 00:47:03) Or an alternative—your agent needs to kind of search the web as if a human was searching Google. We have an alternative to Google, our own index and API that allows us to do that. It is designed for LLMs, designed for agents to provide as much context as possible, not optimizing for clicks, optimizing for information that can provide accurate answers and make good decisions to keep those errors from compounding and allowing more and more workflows to be automated long term.

(Joel Beasley at 00:47:41) That's pretty cool. You love what you do?

(Brian McCann at 00:47:44) Yeah. I think it's incredibly important right now. I mean, we saw a lot of people try to build their own ChatGPT internal setups, get really interested in RAG, get really interested in agents. But in order for those things to make high-stakes decisions and really unlock a lot of the value I think is there waiting to be extracted, it needs to have access to as much context as possible.

(Brian McCann at 00:48:13) And so I feel like enabling that and doing it in this open way that doesn't suck people into an ecosystem that potentially sets them up to be disintermediated long term, but instead is really helping them transform, I think that's really important. I think that's just a path to a better world than the alternative.

(Joel Beasley at 00:48:40) You know, for me, I've never worked at a large company, like a big company. I've built and sold three small—like, when it got to twenty, thirty engineers or people at the company, I left. And so that's been my whole twenty, twenty-five year career. And when I get to talk to all these people, one of the most interesting things over the past ten years and thousand episodes that I've come across is the cult-like following of how vendors and providers work. Like, you could be a Microsoft shop, like everything's Microsoft. You could be Amazon, like everything's Amazon. And then some people get real touchy about it too. Like, even if you're behind the scenes, like you're at the conference having a casual social interaction with a friendly person, people are legit religious about some of this stuff sometimes.

(Brian McCann at 00:49:32) Oh yeah. Oh yeah. It blows my mind. It's one of the—I mean, within companies about their choices, for sure. And in some ways, you know, even in these little startups, right, as they're growing, it's like an important tool to kind of develop this, you know, put the blinders on to some extent. It's a great strength and a great weakness, right? It allows you to focus and not think about that decision anymore, right, and just go all in. But at the same time, you might be missing out on some things or, you know, there are some dangers to some of that type of thinking. But it is fascinating.

(Brian McCann at 00:50:12) But there are still a lot of people who can't afford to, I suppose, just take the risk to some extent of being on one provider and having that dependency. A lot of people can and they can forget it. Set it and forget it. But a lot of folks, I think, do worry about that long term. And to some extent, we're probably best suited for people who want to be a little bit agnostic to that.

(Brian McCann at 00:50:48) You know, they want a little bit of hedging across different platforms, different clouds, different models. They want access to the best wherever it's coming from. If you're just a Microsoft shop and you're going to take Teams because they're going to give you Teams for free, and you're going to take their LLMs because they're going to give you massive discounts on that, and you're a Microsoft shop, then yeah, you're in the ecosystem, and it's going to be hard for anyone to take you away from that.

(Joel Beasley at 00:51:17) I guess the surprise—you're exactly right to all your points. I didn't disagree with any of that. The big surprise to me was how people will get emotionally attached to it. Like, I get it. Okay, we're a Microsoft shop. These are businesses, tool sets that are available. It is what it is. Let's go. There are some people that are like, no, we are only Microsoft. You know, they get real into it. Steve Ballmer dancing on stage into it, you know, like real into it.

(Brian McCann at 00:51:46) Right? People want that. People want to be attached. They want to be passionate in that way, right? It's infectious. I think people, a lot of people are looking for that. They want to be attached to something bigger than themselves. And yes, it can look cult-like or religious-like, a religious fervor to that extent, because people are hungry for it. They want to be part of something. They want the story. They want the narrative. When I was at Salesforce, you know, Dreamforce is incredible. The community that they've built around a database to manage your customers is just a stroke of—I mean, it is magical as LLMs to some extent. You know?

(Brian McCann at 00:52:26) It's just—people live and breathe and die that stuff, and it's genius because then you have people inside your customers' company whose lives depend on you being in that company, is incredible. And so if you can achieve that, you are so deeply rooted in the ecosystem. And in that sense, it's so worth every marketing dollar, every bit of swag, every booth at Dreamforce and decoration. It is crucial to their survival long term that you feel like a special, valuable, indispensable being because you're part of them.

(Joel Beasley at 00:53:26) This is crazy. Dude, this is full circle. Some people find their meaning in that stuff. Yeah. That's their search for their meaning. Oh my goodness. Well look, Brian, I think we're friends now. I like you a lot. The way that you process data is pretty cool. I would definitely—you were talking. Or not. Yeah.

(Brian McCann at 00:53:46) The jury is still out.

(Joel Beasley at 00:53:47) We don't know if this is a deepfake or if it's actually Brian, but we appreciate the conversation. You.com, I want to send people to You.com. What's the one plug? Like, what's the one problem that they might be experiencing today that would tell them they need to check out You.com?

(Brian McCann at 00:54:03) If you are seeing that your agents or LLMs need access to data that you don't have, come to us.

(Joel Beasley at 00:54:17) Nailed it.

(Brian McCann at 00:54:19) I think we're friends too.

(Joel Beasley at 00:54:21) Yeah. Thank you so much for listening. And if you found this episode useful, please share it with a friend or colleague who you think would get value from it. And if you have topics that you would like to hear discussed on the podcast, either add me on LinkedIn or send me an email, [email protected]. Every time I get an email or LinkedIn message, it absolutely makes my day and inspires me to keep going.