Time Worth Wasting: Preserving Meaning in a Predictive World

The Book of Enoch

Hi. Thanks so much. I'm very excited to be here. I'm going to start with that gratitude. But also, if you indulge me, because I've been told I've got a minute. My son is really vexed. He's 17 years old. His name is Bruno. And he couldn't come on this trip with me and he's really vexed about it. So if everybody could just say hi to him, that would be awesome. Say hi, Bruno. Very cool. Thanks. Awesome. Thanks so much. He'll be really pleased with that.

So I'm going to jump into it. As Bill said, I'm going to be talking about wasting your time, because wasting your time is a really important concept in the construct that we've created with AI. And I'm going to start this talk about AI in a weird place, at about 300 BCE.

That's the Lighthouse of Alexandria there, which was up and running during that time. That time in history was very important because during that time, almost more than any other, there was a lot of theological and apocalyptic writing happening. And in this text, in the Book of Enoch, there's a story that to me is very, very modern. In fact, it feels a lot more like a tech prophecy than a kind of ancient scripture. The Book of Enoch, if you're not familiar with it, is about a group of divine beings. They're known as the Watchers, and they come down to earth, and one of them, Azazel, teaches humans how to work metal. At first, this seems like an unbelievable gift, a tool that's going to be a massive shortcut to power, but it becomes clear very quickly that what's happening is that the knowledge was premature and the humans weren't ready for it. So that wisdom was unearned and out of order. It had become a curse for everyone, because it knew something that we are only starting to realize, which is that not all knowledge is neutral. And this idea threatened the theological order, not because it was incorrect or false, but because it was too crazy. It was too wild, too cosmic, and really unconcerned with hierarchy.

So it's my opinion that the Book of Enoch, which you might not have heard of, wasn't lost but was suppressed. Today it's still part of the Ethiopian Orthodox faith, but it's excluded from all the other Bibles. It was buried, again, because I think it was too powerful. It suggests that divine rebellion and human curiosity, those two things, are inseparable, and that's a very dangerous idea. And like our buddy Oppenheimer here, he wrestled with these concepts. Is all knowledge sacred, and should all progress be pursued? These are questions that we need to ask ourselves today.

Silicon and the machines that predict us

So let's jump ahead to today. Today our metal, like in that story, is silicon, and our forbidden knowledge is artificial intelligence. We've built machines that don't just process our world, they predict it. They don't just watch what we do, they learn from us. They train us and they reflect us at scale. But in this world that we've created, that's completely optimized for efficiency and automation, a very human thing is starting to fall through the cracks, and that thing is meaning. Because what's happening is we're flattening purpose into performance and confusing speed with substance. These systems that are filtering our lives are filtering every aspect of it: what we see, what we buy, who we hear, what we feel.

Because these systems, they're not engineered and designed to nourish us, they're built to measure us. And we've optimized the entire internet to keep us there, not to make us wiser. This is evidenced through all the habitual behaviors we have with these utilities. We autoplay the next thing. We scroll endlessly. We test headlines for clicks, not for truth. So we've constructed this digital world that feeds on attention, but it's obfuscating context and intention. And not everything that's measurable is meaningful.

So I'll talk a lot about friction. Friction's very important. Imperfection, wasted time, these aren't bugs in the human system. They're the foundation of meaning. And the thing that proves this out is stories.

Mythologies as living entities

So that super hottie is my wife, and clearly I got the better end of the deal. But she's a professor of intercultural communications. And years ago, she said something to me that really changed the way that I think about brands and belief systems and everything. She said that mythologies are living entities. As a result, my curiosity around these things became totally rerouted. So I went out and began to examine mythologies. But I didn't treat mythologies like fossils. I treated them like organic living material. I looked at them as systems that evolve and adapt and compete, and I mapped them out with genealogies, like family trees. And I understood, or started to understand anyway, not just what they meant, but where they came from and how they mutated and why some survived. Quite obviously, some myths fade and others flourish. But they don't do that in a scientific sense; they do that because they're more durable. They resonate, and then they survive friction, and they do that across generations.

What I learned as a result of that is that meaning isn't fixed. Meaning is something that's alive. Meaning is something that survives. It's shaped and stretched and adapted and strengthened by friction. And I'll say this a couple of times, but stories aren't just content. Stories are memories encoded in myth. Stories are really a cultural immune system, because they don't just reflect who we are, they protect us from what we might become. If you think about it, these myths weren't designed by a committee. They weren't optimized for engagement. They earned permanence through repetition, conflict, and reinterpretation.

Generative AI doesn't do any of that. Generative AI is a big Plinko machine that remixes data. Right? It flattens information and stories. It reassembles them without context, without lineage, without exposure, without any stakes inside of it. And inside of that flattening, we are creating an illusion of progress, because speed is not growth, prediction is not perception, and utility is not truth.

Friction, flat time and deep time

So let me talk about friction a little bit more. Friction is what structures our time, and time is the canvas where we make our meaning. Without friction, we don't have any resistance, and if we don't have resistance, we don't have identity. So a smooth life is one without any texture, and that smooth life will become slippery and then empty.

So the ultimate question we're facing now isn't what AI can do, it's what we're choosing to spend our time on. Because time is more than a unit of productivity, it's a measure of meaning. And I bucket time into two categories, because not all time is created equal. I call some of it flat time and some of it deep time, because time spent making isn't the same as time you spend clicking. One time deepens and the other dissolves. So friction is what gives time weight, and meaning only grows in the weighted kind of time.

Day 29: the lily pads, the polygons and the coastline

So I'm going to give some metaphors now, and you'll probably be familiar with some of them. This is the lily pad metaphor. Imagine a lake, and every day the number of lily pads on that lake will double. Right? It's geometric growth. On day 30, the lake is totally covered. So what day is the lake only half covered? The answer is on day 29. That's geometric growth. It feels linear until that very last moment, until you realize there's no time left. And that's where we are with AI. We're on day 29, because progress has been doubling slowly. What we're seeing with AI didn't happen in the last three years. It's been decades. And we didn't notice it because it was beneath the surface and bubbling up, compounding itself in silence. And along the way, it's been optimizing everything. But more isn't necessarily always better.

Here's another metaphor, and I'll use a computer graphics metaphor for this. That's a 3D rendering of Beethoven. At 60 polygons, that bust is abstract. At 600, you start to see the face. At 6,000, you can make out that it's Beethoven. But at 60,000, there's not a big difference. There's not a big difference between 6,000 and 60,000. And AI works in a similar way.

Another metaphor is the coastline paradox. If you look at a picture of the coast from outer space and you keep zooming in on it, it's almost like that coast will grow infinitely, because the closer we zoom in on it, the benchmarks change. But it doesn't mean that the size of the coast is infinite. It doesn't mean that a system is growing in wisdom if we keep adding more processors to it. It means that we've simply changed the ruler of measurement.

Faster math is not deeper intelligence

So that's Leopold Aschenbrenner. And I'm not making fun of Leopold or giving him a hard time. He's clearly a genius. At 15 years old he got into Columbia University, graduated valedictorian, and led the superalignment team at OpenAI until he was fired, because he said that there were huge safety vulnerabilities with their product. Once he was fired, he went on and published a manifesto where he said it was, and I quote him, strikingly plausible that we'll get to AGI by 2027. Now, I don't necessarily agree with that. The argument that he makes in there, and I'm being slightly reductive here, is that AI is going to 27x its capabilities every year. If you do the math, that is what he's saying. So he's presenting AI development as this unstoppable exponential force that's destined to dominate everything we do: our global GDP, it's going to reshape our economy and our society, not in years but in decades.

And Aschenbrenner, like so many other people, is mistaking faster math for deeper intelligence. But just like the problems that I brought up, it's built on compute, not on consequences. He's predicating everything on curves, not on culture. And as I said before, stories aren't just content. Therefore 27 times performance is not 27 times meaning. They're different things.

So if we pause here for a minute, we have to think about what happens if we take exponential intelligence and run it on linear understanding, or if we build machines that learn faster than we do but they don't know what they're learning for. That seems to be the situation we've gotten ourselves in. That's like thinking we can continue to double those triangles and the face is going to become clearer and clearer, or that we're going to measure that border with an ever shrinking ruler and say, oh, you know what, the coastline of this country is infinite. Because at some point, reality pushes back.

Entrainment: what biology does that computers do not

Computers isolate complexity. They count it. They compress it. And they predict it. Biology doesn't just observe complexity. It joins it. It entrains it. The distinction there isn't speed. The distinction is that it's a relationship, because entrainment means syncing up. When your heart rate adjusts to your breath, when you see fireflies blinking in unison in a field, when your brain waves sync with a rhythm at a concert, or when a baby aligns its breathing with a parent who is holding it. That's entrainment, because biology adapts in real time, and that's a distinction that AI doesn't understand.

Because life is non-local. There are things that can influence each other without proximity, and that's how quantum particles do it. Well, humans do it in the same way, and so does culture. Life is also acausal. There's no real clean traceable line from action to outcome in life. And life, most importantly, is entangled. You can't always say what caused a movement, because memes don't go viral because one person clicked on them, and a revolution doesn't happen because one speech is made. In the same way, your mood doesn't shift because one neuron fires in your brain. That's because influence jumps.

So real life is about fields and feedback and emergence. It's about people that are non-local syncing to a shared belief. Meaning itself isn't just biological. It's social and it's viral. And that's why prediction isn't understanding. AI trains on everything we do, but never on why we do it. It sees the purchase, but not why you're longing for that thing. It sees the scroll, but not the ache of why you're scrolling. It sees the playlist, but not the memories that it evokes. As a result, we're not really able to understand what's real anymore, because we're drowning in prediction and simulation and spectacle. The fake news isn't a glitch. It's the system. And reality, the shared kind, is slipping as a result of this.

Is that real?

So Bruno, my son, that I mentioned at the top: when he was about four years old, for the first time, I took him to see fireworks for the Fourth of July in the United States. We're lying on a field on a blanket underneath the night sky, and the fireworks started. And he leaned over to me and he said, Dad, is that real? I thought that was the greatest thing in the world. Because I had just spent decades inventing, patenting, commercializing artificial reality, virtual reality, massively in this field, and deploying these things at scale. And I thought, wow, what a seminal moment, that my son is asking me this question, that we can't distinguish it.

But the truth is, that wasn't validation. It shouldn't have been validation for me. It was actually more of a warning, and it took me a minute to understand that. Building these synthetic worlds has real consequences. As I said, I've spent my career building these environments, these fictional realities, these artificial agents. And now we've trained machines not just to simulate the world but to predict it. They're designed to predict it, to rewrite it, and ultimately to replace it, because we're abdicating control and consenting to that.

But prediction, it turns out, comes with secrecy, because AI is actually designed to censor. Not to oppress, but to protect the illusion of control. And you've heard many other speakers up on this stage talking about that today. The proprietary data, the proprietary models, the black box that's behind it: you can't see how it arrived at its conclusions, and you're not meant to. That's not intelligence. That's secrecy at scale. Friction is reality's anchor, and without the friction we lose touch with what's real, and without the resistance we accept that simulation as truth. So we don't just need friction to slow things down, we need it to remember who we are.

Azazel was early

And the same thing is happening with AI. We're throwing more chips at the problem and saying that that's going to give us deeper meaning, or give us meaning. Stacking GPUs instead of asking better questions. Because meaning is never going to compound at 27 times a year. It doesn't double on command and it doesn't scale. So what if we really can't predict what we'll do, what we'll want, all of these things? Are we going to lose mystery? Are we going to lose our agency? Are we going to lose the joy of not knowing? Because the greatest danger isn't that AI is going to fail, it's that it doesn't.

And so maybe that Book of Enoch that I was talking about at the beginning wasn't a warning about disobedience. I think it was a warning about acceleration, and what happens when power arrives before readiness is established. And like Azazel, I don't think he was necessarily the villain in this story. He was just early. That mythology maybe needed to resonate for a couple thousand years before we got to a reason for it. So these modern Azazels, the compute curves, the benchmarks, the tokens, the trends, they're maybe bringing us gifts before we're ready. And like the Watchers, we might not survive them.

Five axioms

So the question is, what do we do? I put these five axioms together. Maybe they're guidelines to help guard our reflection. Because truth isn't created by machines. In fact, it's distorted by them if we aren't careful. We need to protect our agency, because the machine can act but you have to decide. We need to choose what's worth doing. Time worth wasting. Prediction is cheap, but the pursuit of things is sacred. We don't want to outsource our judgment. We need to be careful what we abdicate in our willingness to surrender control. And we need to remember that AI is the mirror, and whatever we feed it is what it becomes.

So we can't predict forgiveness or playfulness, interpretation or disobedience. AI can simulate knowledge but not wisdom. It can replicate tone. It does rhetoric very well, but it can't do that with tenderness. It can model behavior, but it can't do that with meaning. Prediction requires patterns, but humanity, we interrupt patterns. And that's where the soul, if you believe in that kind of thing, lives.

Things that do not scale

So to make it more concrete, these are things that don't scale, that don't optimize, because they're sacred, and things that AI fails to grasp. These are all very personal and very real for me. That's me. I wore a Star Trek uniform to school when I was little, every day, so I've always kind of been like this. And no prompt could have generated the weirdness that embodies me. This person, sitting in the audience, she's there. I raised her up and got to walk her down the aisle last year. And that moment here that you see, before that happened, that was a lifetime unfolding. There was no audience, no output, no metric for that. These things don't scale. And Bruno drew this when he was little. That's more expressive than 10,000 frames of training data.

These are the types of things I'm talking about. This is the kind of friction that I'm talking about introducing. This is the dissonance that exists today. So meaning isn't a straight path. It spirals and deepens and echoes. It's more like a mycelium network than it is a SQL database. You can't map it out. It's not a pipeline. So the paradox is this. Just as destruction compounds, so does redemption. If we orient ourselves now in the right way, geometry can still be our friend, all those things we talked about. And those small acts of resistance that cause friction can ripple through. It can scale.

Stay human deliberately

So why am I saying this? I'm not saying this as an engineer or a scientist. I'm saying it as a father. I'm saying it as a husband. I'm saying it as a son, and I'm saying it as a human being, because I'm watching the future come too fast for the structures that we're still living inside of. So we need to stop designing systems to eliminate inefficiency. We need to design for authorship, for difference, and for depth, because the future isn't predictive. The future is participatory.

What can you do? Be conscious of what you're feeding the machines. Resist the urge to optimize everything in your life. Teach children boredom. Make something that doesn't scale. And let time feel slow. Because slowness isn't inefficiency, it's reverence. And every time you choose friction, you're choosing to decide to matter. Ultimately, what you should do: stay human deliberately. Thank you.

Questions from the room

The host asked how he fosters that sense of meaning, of making the most of wasted time, with children as they grow up. Jason said:

Yeah. I think it's about being deliberate. I certainly went out of my way not to regulate access to technology in any way, because I think when we restrict those things, it makes them precious and it makes them vulnerable to wanting those things more. So I tried to ensure that engagement on an interpersonal level was the priority and the most enjoyable thing.

A member of the audience who works with narratives and communication asked what stories we need to be telling ourselves about AI today, and what story will bind us together as humans. Jason said:

Well, I think that obviously it's a very complicated question to unpack. But the reductive answer is that we have to be human first in everything that we do, that AI is not something that would replace us. It's something to augment us. Certainly not the first person on the stage today that said that, but that's the reality. And we need to look to construct stories, and to look to one another in order to build those things, to ensure that what is being constructed isn't synthetic. And to learn to discern the difference between artificial and synthetic stories, those metaphors. Are they traceable back to some sort of purpose and intent that is human, without any sort of sponsorship?

Asked whether he has a framework for children, who write essays with AI while their teachers summarize them with AI, so that they use AI as a tool but still learn to do the work themselves. Jason said:

Yeah. I couldn't agree more. Again, it should be something to augment it, to help perhaps with the prompting process. But authorship, in the true sense, needs to remain individual and human. It shouldn't be a substitute for authorship.

Asked about an era in which hundreds of new AI tools are released every day and everyone is hunting for something, and how we slow down and waste time. Jason said:

I think in order to waste time, you have to discern or focus on your priorities, and ensure that those things are not being driven by external forces but driven by your own needs and desires. And I think what's happening now is we start to compress things and abdicate more control to all of the algorithms, and that's becoming a little bit confused. Sorting our priorities based on our desire rather than our dictate is the challenge right now.

The host asked whether getting the friction right is something we each have to figure out personally, or whether we will turn to AI to figure it out. Jason said:

I think this is the rebellion. This is the struggle on an individual level: to introduce that at scale and start to demand that we need to step back and take a break. To say, no, I'm going to set time aside in my day to think. I'm going to set time aside to step back from what's being fed and make sure that I'm not inside this echo chamber or this continuous loop that's amplifying ideas that may or may not really represent the direction that I want to travel.

Asked for his tips on keeping the ability to think critically, and confidence in one's own beliefs, when facing an AI that can almost predict what you are about to say. Jason said:

It's to make sure that if you're engaging the AI, you're asking enough good questions for the answers it's providing you. To say, wait a minute, is that right? You must question it. If you're engineering these prompts to get a response back from it, then the question needs to be, is that right? What if it's this other thing? So make sure that you're always, for lack of a better term, being the devil's advocate, or having the contrarian view inside of what you're being fed. I think that can help to identify balance inside of it.

Asked how he does this himself. Jason said:

Frankly, I do exactly that. Building agents that allow for that balance, where I'm intentionally creating an adversarial prompt model to say, okay, whatever this thing's going to tell me, first of all, I need to fact check everything. Second of all, I need to say, okay, what would be the alternative to that? Can I get to the medium where I can try to figure out what's really happening here?

Asked what his inspirations were, beyond his wife and family, when he started to think that we have to create friction. Jason said:

I think that's it, 100%. Being able to imagine a world where the technology is merging with our biology and becoming more and more transparent inside of everything we do. And so I felt this desperate sense, in order to ensure that we don't get smoothed and flattened into a reality that we haven't constructed. For me, it's been more of a call to arms, to say, look, no, I need to really focus on building friction into this, to ensure that what I'm being given isn't an alternate reality to what I'm discerning my outcome to be, and then actually using those tools to help to discern that best outcome.

Asked for an example. Jason said:

So I'm writing a book that's going to be out. And I will often ask AI to take on the role of another author or an editor that I'm fond of, to give me a response if I've finished a chapter, or if I've finished something like this, to start those arguments. Not to reinforce that what I'm doing is correct, or the belief that it's going to be additive to something that I've constructed, but to give me a contrarian view on the thing that I've produced.