The Semantic Arms Race: Ben Hunt on AI and Narrative

Ben Hunt: I think that part of the problem with the lack of trust is our transformation of what we write and what we say into content.

Jeff Kavanaugh: And are we heading to some sort of semantic arms race?

Ben Hunt: Yeah, we're already there. So, it's an arms race where the technology is being used, I'll say, against us.

Ben Hunt: One of the simple tenets one can use to kind of build this sort of critical distance is to always ask yourself the question, "Why am I reading this now?"

Jeff Kavanaugh: If Ben Hunt has taught us anything across a decade of Epsilon Theory, it's that the story is the strategy, and now AI can read it faster than any human can write it.

There's a line you'll hear in strategy meetings, "Let the data speak for itself." My guest today has spent more than a decade proving that's wrong. Data doesn't speak, stories do, and the stories that surround data are what actually make and move markets. They shape decisions and determine outcomes.

Ben Hunt is the co-founder of Perscient and the creator of Epsilon Theory, one of the most widely read independent investment publications in the world. He's a former hedge fund manager, a Harvard PhD, and one of the sharpest and most contrarian thinkers working at the intersection of AI, narrative, and markets.

Today, we're going to talk about semantic narratives, what they are, why AI changes everything about them, and what business leaders need to understand right now. Ben, great to have you on the Infosys Knowledge Institute podcast.

Ben Hunt: It's great to be here, Jeff. Thanks for having me.

Jeff Kavanaugh: You've built an entire body of work, and it's significant, substantial, on the premise that narratives are more important than data in understanding markets and institutions. That's a strong claim, and you built a company on that as well. Make the case.

Ben Hunt: So by data, let's describe that as structured data. So what I'm more interested in is unstructured data. It's still data, it's still information, but it's the text, it's the words, it's what we hear and what we see. It's the ocean of data and information that we humans swim in all the time.

There is a structure to unstructured data. There is a path and a form, a script, if you will, that unstructured data takes, and we're totally not aware of it.

And the great institutions of our time, by which I mean Hollywood, Wall Street, and Washington, DC, they're all storytellers. That's what they are. Hollywood is an overt storyteller. We recognize that they are in the business of storytelling. And famously, Hollywood, what they say — there are only five scripts for movies in general.

Jeff Kavanaugh: Hero's journey archetypes.

Ben Hunt: Exactly, and this is true also for politics, it's also true for markets. It's true for any sort of human social behavior. We don't see the storyline. We don't see the pattern in the script. Although you can do this for any movie you've ever seen. Go type in — for me, it's "The Godfather" movies. I'm a big fan and watched them all a million times. But type in the name of any movie and say, "Three-act structure," and you'll see an image of the structure of the script. And the same things happen over and over again in every script. Once you start looking for it, once you start looking for the structure in unstructured data, you can't unsee it, and you'll see it everywhere.

Our company, our goal is to try to take this observation about the role of storytelling, the structure in unstructured data, and apply it to politics. Apply it to markets. Because again, once you start looking for the same patterns and stories, you'll see them over and over again, and we humans are hardwired to respond to them.

Jeff Kavanaugh: You use the phrase common knowledge in a specific way. Not what people know, but what people know that other people know.

Ben Hunt: Yep.

Jeff Kavanaugh: Unpack that for people who might not think of narratives that way.

Ben Hunt: Well, this is actually something that John Maynard Keynes back in the '30s was talking and writing about, and he was trying to understand markets, as were we. And his observation was that you'd often see this sort of — I'll call it crowd behavior. We all see it in markets today. How does storytelling work so that we all apparently seem to act at once according to the same beliefs?

And what he proposed was that this was an entirely rational process. It's not like we're some irrational creatures who just get all of a sudden fearful or greedy and all just decide to act as one. What he said is that, well, when you think about the way you rationally should be making your decisions in an area where people are voting for something — and that's what you're doing when you're in markets. You're voting for a security or something you're buying because you think it's going to go up. You think other people are also going to buy that security.

So what Keynes was observing was that, well, it's not really, when you think about it, what you believe about that company or that stock — you start to think, "Well, where I should really be thinking about is what does everybody else think? Where are they going to be buying or selling?" And then he said, if you think about it a little bit more, it's not even what everybody else will be doing independently, because we're all smart enough to be thinking in these same terms. Everybody is thinking, "What is everybody else thinking?"

And once you recognize that, you start thinking, "Well, what changes what everybody knows that everybody knows?" And the answer is media. This is called the common knowledge game — in game theory, it's called a missionary. Someone who gets in front of a mic, gets in front of a camera, and speaks to everyone. Because if everyone sees someone — a missionary — saying something, an opinion, saying, "This is what the truth is," we all know that everyone else saw it. And that missionary statement is what drives what everyone knows that everyone knows. And when that clicks, when you get a powerful missionary, then that changes everything in a market, in an organization, in a country. It's really one of the most powerful forces in human social behavior, is when everyone knows that everyone knows.

The Emperor's New Clothes is a great example of the common knowledge game in action. The king's walking down the street naked as a jaybird. We all see it with our own eyes. But it's not until the missionary, the little girl, announces loudly so that we all hear it, and we know importantly that everyone else heard it — then it's rational to act on the information. It's not that we don't trust our own eyes, but it's actually not rational to act on our own beliefs unless we think that everyone else is also acting on those same beliefs. That's what common knowledge is.

And when common knowledge shifts and changes — I'll give you some example — Joe Biden's debate performance is a great example of something that many of us thought privately was an issue. But when everyone saw what happened on the debate, you can't explain it away. We all saw what we all saw, and that's where everything changed in the election dynamics for who was going to run for president for the Democrats. You can't unring a bell when common knowledge changes, when there's a common knowledge moment like that. So that's what common knowledge is.

Jeff Kavanaugh: Perscient is built on using language, using language models that are analyzing unstructured data — the stories, not the numbers or the structured part. What can AI read in a narrative that maybe a human analyst might miss, and maybe vice versa?

Ben Hunt: Well, AI has been transformative in our ability to trace and measure the scripts within this ocean of text that we live in today.

So what our company does specifically — and this principle can be applied anywhere — we take in all the world's news every day, everything that's publicly available. Everything. Everything.

So what we do is we've identified the scripts. Now, it's a lot more than five. We track more than 5,000 distinct individual scripts. Think of it as a narrative. Think of it as a theme. It can be a political theme, a market theme, a sports theme. I'll use another Hollywood example. So the television show "Law & Order" — it's had over 700 episodes right now. There are 12 scripts. There are 12. And each one has a very similar pattern, but there are 12 pretty detailed scripts across all those 700 episodes. Each of those 700 episodes kind of appears unique and different — different people, different scene, different setting, different crime. But it's only one of 12 scripts. We're identifying the script, and so the application of that script will absolutely change in the same way you'll shift out actors and scenes, but the underlying script will be the same.

So what we do is we pull in all the news media in the world, then we use AI to see for residue. We call it a semantic signature — what's left behind as that script passes through the actual words. You start to see these scripts almost as like living things. They have a life cycle. A story is here, it grows, it dies down. It never completely goes away. It'll come back again in a different case, in a different form.

So that's what we're tracking — the rise and flow of all of these stories all at one time. And what you learn from this is that there are certain stories, certain scripts, certain patterns that are just incredibly impactful that we are hardwired as humans to respond to. It's what politicians, I think, have known for a long time. And today, we're able to actually measure it, identify it mathematically over time, both to defend ourselves when others are using the storytelling, I'll say for their own advantage and against us, and also, frankly, so that we can tell a more effective story — we, a business leader, about the things we care about.

Jeff Kavanaugh: If AI can detect narrative manipulation, spin, script injection, coordinated messaging — what happens when the actors generating those narratives also have access to AI? And are we heading to some sort of semantic arms race?

Ben Hunt: Yeah, we're already there. And it's an arms race where the technology is being used, I'll say against us. Our real goal in — I'll say in spreading the word, making this ability to measure and see it more visible — is to do something you mentioned right at the beginning of our conversation, you were saying, we don't see the script. And again, we're hardwired not to.

I mean, I know a lot about scripts, and I'll go into a movie, and what I know when I go in to see a movie is that about halfway through act one of all movies, of every movie you've ever seen, the MacGuffin — the object of desire, MacGuffin was what Hitchcock called this thing — and it could be a thing, the ring of power, it can be your long-lost love, it can be an idea. But about halfway through act one, the MacGuffin, the object of desire, will be introduced into the plot. And that's true for every movie you will ever see. So we are, as humans, hardwired not to see the skeleton, the architecture, the structure in unstructured text.

Once you see it, though, once you're presented with it, seeing is believing, and it allows us to maintain a critical distance between what we are told by our political leaders, our business leaders, whoever is out there behind a mic, in front of a camera, shaking their finger at us and telling us how to think about the world. It gives all of us a critical distance. So the first and foremost use of this, I think, is to allow all of us to have that critical distance in what we read and hear — so that's one, because it's already being used against us.

Two is that by — I'll say popularizing this, by giving all of us the tools to see the structure of language and the like — we can all start to tell better stories. Right now, what we're immersed in is, I'll use the phrase, AI slop. All of this messaging, we are overwhelmed by storytelling, by messaging, and so much of it is constructed. It's too much.

What this allows anyone to do is to become better at telling stories — to not just gravitate to the stories that drive engagement, which are the stories that are being used against us today, but actually to look for the patterns of storytelling from what I like to call the old stories. Stories of loyalty, of trust, of empathy. There are some amazing old stories and old patterns of stories that get lost in this vast tsunami of AI slop. That's the other thing this allows us to do, is to bring forth the good scripts, if you will, to tell better stories.

So first, it's self-defense, maintaining critical distance. Second is telling better stories that are more pro-human and human-centric, versus the stories we're overwhelmed with today, which are, in my view, profoundly anti-human and destructive of the good things in human society.

Jeff Kavanaugh: Institutional trust is near historic lows, pretty much every sector. Is this a narrative problem, and is it fixable with better communication, or is there something else that's just hard to overcome?

Ben Hunt: So this is going to be kind of a tough love session. I think that part of the problem with the lack of trust is our transformation of what we write and what we say into content.

Jeff Kavanaugh: That word, content — you're doing what?

Ben Hunt: We are doing. I shouldn't say you — when I say you, I mean the most generic you — is this idea that we can create content which will have an impact on our reader or our listener, regardless of whether that content is authentic to us.

That's what central bankers did. That was the whole idea of forward guidance. It may not be what we actually believe in our heart of hearts, but we're going to speak as if we believe it because we think it will impact investor behavior. It's advertising — it's advertising for the sake of advertising, not because you actually believe in the claims you're making, but because you believe that they are effective words to sell that product.

The whole notion of thought leadership can fall into this camp as well — that thought leadership is a thing that can be constructed or manufactured, as opposed to something that needs to come authentically from an actual leader.

Jeff Kavanaugh: Yes, and since you mentioned tough love and you're looking at me, I'll respond. That's why it's so important, at least for us, and I think anyone who genuinely, authentically is in this area — that there is research of some kind, there is a signal you're basing it on facts. Then as you abstract, or at least take a step back, maybe a critical distance from those facts, you're able to have an informed opinion and perspective. It is your angle through the solid object of how you view that in a context. Whatever it is, that is the slightly more messy aspect of deconstructing those facts or taking a step back from them. At the same time, that's a step into the world of unstructured data and what we are dealing in. It is messy.

Ben Hunt: It has to come from the heart, because that's what makes it effective.

Jeff Kavanaugh: Yes. And that's why the light bulb went off, I think, for some of what we've done in the past — that although we're in a business sense, and dumping words like empathy and human into it too much, you think, ooh, this is a different kind of discussion. But the interesting thing is, those people that go to work, that are analysts, that buy, that sell, that lead, that follow — they're all humans, and deep down, that's how they're wired.

Ben Hunt: Historically, right? Anciently. This is how we're wired. But does the messaging actually fit with here? And that goes back to aligning to purpose, and your model, and people can sense that.

Jeff Kavanaugh: And I actually find that younger people sense it so much more quickly. So my daughters, who are all of this digital age, who have been inundated with messaging all the time — they can smell inauthenticity like that.

Ben Hunt: I think this is the trap so many people who are in our business of working with messages and words and ideas.

Jeff Kavanaugh: Especially marketing folks.

Ben Hunt: It's easy to fall into the trap of efficacy, and it's a false efficacy for, again, long-term success, non-myopic success with your messaging. It has to be here. It has to tap into those old stories. And now you've got the tools to actually measure and see their effectiveness. So I think it's just the most important thing.

Jeff Kavanaugh: You've written that our autonomy of mind is our birthright — that is what makes us human. And then we give it away when we let semantic systems do our thinking for us. Does AI make this problem better or worse? What's its effect?

Ben Hunt: Well, right now it makes it worse, because, like you say, it's being used against us. It's not AI per se, but it's the use of AI to do these A/B tests and the efficacy of messaging at scale — whether that's just creating bots that generate tons of potential ideas and let's see which ones work.

We see this all the time when we're doing tracking on, like, what are the stories that got some meme stock going? Because — exactly, you know, I was talking about the math of epidemiology — one of the goals, when you're tracking a disease, is: can you find patient zero? Well, we can find patient zero. You can work backwards to see when a narrative has taken off, what was the path by which it spread.

So what we find over and over again is just a very intentional effort to — I call it a snowball approach to messaging. You're at the top of a big hill, and you just start rolling down hundreds of different snowballs in hopes that one of them will pick up mass and create an avalanche down the hill. This is happening all the time, particularly in social media, but also in mainstream media.

And it's why one of the simple tenets one can use to kind of build this sort of critical distance is to always ask yourself the question, "Why am I reading this now?" And again, it's not to say that it's a lie or it's wrong — I'm not saying to fight it. But just to create that critical distance.

AI today, I say, is being used, I think, against us. What's possible, though, if we see it for this incredibly powerful measuring instrument of the semantic dimension, is our ability to explore that dimension, to explore the old ideas, the old stories. It's old wine in new bottles, essentially — how can we bring these old stories, which are extremely powerful in their own ways, not with the negative engagement, but with much more powerful stories of loyalty and affect. Again, this is all measurable. How do we bring those old stories and present them in ways that are in line with our institution or corporation's values and goals?

Jeff Kavanaugh: Something you wrote on the golden age — it was last year.

Ben Hunt: Yep.

Jeff Kavanaugh: You talked about the golden age of AI coming. Do you think we're closer to it? Is there some skepticism? What's your point?

Ben Hunt: Oh, I think it's coming. I do think it's coming. Because my strong view is that AI — it's a technological innovation — it's not just akin to the printing press, it's more than that. I really think that AI, large language models, is more akin to the invention of writing. I think it's as big of a thing as that.

And the reason why is that semantics — what that word means, it's a $10 word that means meaning — and meaning is what drives our lives. You mentioned earlier this autonomy of mind, which is what makes us human. We use this autonomy of mind to find meaning in our lives, find it in big ways, in little ways, our purpose. But in little ways too, meaning is all around us, but we walk through our lives without seeing it and without using it.

I think that's where people lose sight sometimes — they think there's only the big meaning, the big why, not realizing that while they have their own big why, they can be attracted to all the small things around, if leaders, maybe even marketers or communicators, think like that a little more authentically.

So all of our preferences, all of our choices in life, they are driven by meaning. And sometimes the meaning will kind of filter down to us over time. What I'm saying is that we should make a real effort to see meaning clearly in our lives, both in big ways and little ways, and that's what AI allows us to do. They live natively in the world of meaning, which is always a probabilistic approach — it's all the ways you can say the word table, not just a list of the ways you can say table.

It is truly magic to see this world of meaning and to be able to identify the little pieces of meaning. It's the difference between search and discovery. So I love to tell this story. So when I was a kid, my dad would take me every week, every other week, to the big Birmingham, Alabama Public Library. And oh man, I still remember the murals up on the walls of storytellers of the past, and I just loved spending time at the library. So this will be familiar to those of us of a certain age — there'd be some topic you're interested in, or a book, and you'd go to the card catalog and you'd find the card.

Jeff Kavanaugh: Thank you, Mr. Dewey. That's right.

Ben Hunt: And they had little scraps of paper and little bitty pencils, and you'd write down where to find it, and you'd go to the stacks and you'd find your book. That was search. The magic for me then was discovery — was to see, oh, here are all these other books on this shelf, and here are the books two shelves above it and below it. That was the magic.

And that's what's possible today with AI. It's not search, it's not the list of all the tables in the world — it's the idea of table. Here are the things around that. That's what it opens up for us — not for looking for a book title, but for looking for meaning. Big meaning, little meaning, all the meanings. And for as a human being trying to make my way in this world, it's just a tool of enormous power that I want to have available to me and not controlled. I want it to be decentralized. For a leader, it's an enormous tool to present meaning to your audience, whatever that audience is.

It allows us not to put a fresh coat of paint over things that we've already done, not to achieve productivity by creating an agent and doing old jobs faster, but to actually do new work in new ways. That's the productivity miracle, the golden age, that I think AI can bring about. And I do think we're closer than ever to that.

Jeff Kavanaugh: Well, you spend a lot of time with investors and executives wrestling with this area, and probably passionately saying some of these same things. What's the question you wish more of them would ask, yet few of them do?

Ben Hunt: Let me start with the question that I wish was asked less. The question — whether you're talking in an investment perspective, whether you're talking in a media perspective — the questions often come down to notions of efficacy. How can I construct a message that works, how can I find a signal that predicts. The idea is to use semantics for that notion of efficacy.

What I wish would be asked more was, how do we study the patterns of semantics and meaning to try to pull out the old stories that work — more than this focus on efficacy and content for content's sake.

It's funny, when I have this conversation with people who see the semantic dimension as real, like I do, they are usually the people I talk to in San Francisco who are working with the frontier models and the LLMs and the like, because they see every day that LLMs are not tools for search — they really are tools for understanding meaning and semantics and storytelling.

What I want to encourage are questions about seeing that — just that notion of seeing this world of meaning and trying to move back a little bit from just raw efficacy, because it's only by seeing the semantics that you're going to be able to see meaning that is actually authentic for you, and that will actually work in the long term for your company, yourself, your family, whatever sphere we're talking about. That's when communication truly happens — when you actually connect with another human being on the fulcrum of meaning.

Jeff Kavanaugh: Narrative isn't soft. It isn't the communication team's problem. If Ben Hunt has taught us anything across a decade of Epsilon Theory, it's that the story is the strategy, and now AI can read it faster than any human can write it. That asymmetry is the challenge and the opportunity in front of every leader listening.

Ben, this has been a genuinely interesting, compelling conversation. What a pleasure.

Ben Hunt: Thank you. Thanks for giving me the opportunity.

Jeff Kavanaugh: For our listeners, find Epsilon Theory at epsilontheory.com and Perscient at perscient.com. And as always, keep learning and keep sharing.