Artificial I

Episode 226 of the Secular Buddhism Podcast

The Self Isn't Fake, It's Made

Hello, and welcome to another episode of the Secular Buddhism Podcast. I am your host, Noah Rasheta. As always, I want to remind you that you don't need to use what you learn from Buddhism to be a Buddhist. You can use what you learn to simply be a better whatever you already are.

Today, I want to talk about artificial intelligence, but not quite the kind you're thinking of. I want to talk about the artificial I, the letter I, as in me, myself, the one who feels like it's doing the thinking.

Here's the idea I want to explore. The sense of self that feels so solid, so obviously real, works a lot like the AI systems everyone is worried about right now. It was trained on information we never chose. It predicts, it explains, and it answers with total confidence. And most of the time, we never think to question it.

So here's where we're going. First, I want to talk about why this matters, especially right now. Then I want to explain in very simple terms how AI actually learns. And you don't need to be a tech person to understand it. And once you do, I think you'll start to see the resemblance to your own mind. Then I'll explain what Buddhism means when it says the self is constructed, because that idea gets misunderstood a lot. We'll look at three things people fear AI will do to us, and notice that the self has been doing all three for as long as we've been alive. And I'll end with a practice that you can try this week.

"That Sounds Exactly Like Me"

So some of you know that a few years ago, I left teaching paragliding and started working for a data infrastructure company. That's the kind of company that builds the systems that power the internet and, more and more, artificial intelligence. I also went back to school to study applied artificial intelligence for a master's degree. And for the podcast's ten-year anniversary, I built a tool called Noah AI. It's trained on everything I've taught over the past decade, every episode, every book, every course.

And every once in a while, I'll read one of its answers when I'm prompting it, and I'll think to myself, "That sounds exactly like me." And of course, I'll realize I've never actually said that, not in those words, and sometimes not at all. But it took the way that I talk and the way that I think and produced something new that fits the pattern.

And then it hit me. It isn't lying. It's doing what it was built to do. It took patterns from the past, produced the most likely next thing, and said it with complete confidence. And that's where I had the slightly uncomfortable thought, "Well, that's a pretty good description of the voice in my own head."

The Megaphone

So let me start with the concern or the worry, because I think it's a reasonable one. A lot of people are concerned about AI right now. They're worried about jobs, about misinformation, about machines making decisions that affect our lives without anyone really understanding how those decisions got made. And those risks are real, and they deserve serious attention. I work in this field, and I take them seriously.

But there's another risk that we rarely talk about. It's much closer to home, and I'd argue it's more dangerous than any of them. It's the mind that's going to be using the technology.

So think about a megaphone for a minute. A megaphone doesn't decide what gets said. It doesn't have opinions. It takes whatever is spoken into it and makes it louder so that it can reach more people faster. If someone kind picks up the megaphone, you're going to get kindness at a high volume. If someone angry and confused picks it up, you'll get anger and confusion at a high volume. The megaphone is the same in both cases, but what changes is the person holding it.

AI as a tool is a lot like that. Whatever else it does, it amplifies the mind that's using it. A clear, caring mind with a powerful tool can do tremendous amounts of good. A confused, reactive, unexamined mind with a powerful tool produces more confusion faster. So the question that matters most isn't only what can this technology do, it's also who's holding it, and do they understand their own mind?

And if you look back through history, I think you'll find this pattern everywhere. Every war, every environmental disaster, every misuse of a technology started somewhere. It started in a human mind, probably a human mind that didn't see itself very clearly. Someone was absolutely sure about who they were and who the enemy was. Someone who was sure that more is always better. The tools changed over the centuries. It's been spears, then cannons, then bombs, and now it's algorithms. But the mind holding them hasn't changed nearly as much. So AI didn't create that problem, it just inherits it.

The Three Poisons

In Buddhism, there's a teaching about what are called the three poisons: greed, hatred, and delusion. Greed is wanting more and more. It's running after the things that we want. Hatred is the opposite. It's pushing away and attacking what we don't want, or running away from what we don't want. And delusion is not seeing things as they really are, especially not seeing ourselves as we really are.

And the idea here is that delusion is the root, and greed and hatred grow out of it. So a technology that amplifies whatever mind is using it is going to amplify all three.

So that's why I think this conversation really matters. The unexamined life has always been dangerous. AI just raises the stakes. Now, here's the good news, and it's the thread for everything else in this episode. The thing holding the megaphone can be looked at. It can be understood, because the self, like AI, is something that was put together. It's not fake, it's made.

How AI Actually Learns

Before we talk about the self, I want to explain how these AI systems actually learn. I'm guessing a lot of people listening have used a tool like ChatGPT, or at least have heard about it, without maybe anyone ever explaining how tools like this actually work. Again, you don't need any technical background for this. If you've ever sent a text message on your phone, you already understand the basic idea.

So think about the keyboard on your phone. When you start typing, it suggests the next word. If you type "on my," it's probably going to offer "way" as an option for the next possible word that you're going to type. If you type "running," it offers "late."

How does it know? Nobody sat down and programmed your phone to know that you're always running late. It just learned that from you. Every time you typed a message, it was quietly paying attention to which words tend to follow which other words. After enough messages, it starts guessing, and it's usually right, because we're all more predictable than we'd like to think.

And that's really the whole idea behind the AI tools everyone is talking about, just on an enormous scale. Instead of learning from your text messages, a system like ChatGPT learned from a huge amount of human writing: books, articles, websites, conversations, far more than any person could read in a thousand lifetimes. And from all of that, it learned patterns. Which words tend to follow which. Which ideas tend to go together. How a question is usually answered. Then when you ask it something, it does exactly what your phone keyboard does. It predicts the most likely next word, and then the next and the next, until it has a whole answer.

So all of that writing it learned from has a name. It's called training data. And here's the part I think is really important to understand. Nobody writes an AI's opinions by hand. Its opinions, its tone, its habits, its blind spots, all of that comes from the training data. If the writing it learned from leaned in a certain way, the AI leans that way too. It doesn't know that it's leaning, it just thinks that's how things are.

So there are three things I'd like you to hold onto. One, it learned from data it was given. Two, it predicts based on patterns from the past. And three, it can't tell which parts of what it learned are true and which parts are just familiar.

Your Training Data

So let's turn that around, because our minds are very similar. Think about how you became you. Before you were old enough to evaluate anything, you were already learning. You learned a language, and with it, a way of sorting the world into categories. You learned what your family valued and what they feared. You learned what got you praise and what got you in trouble. You learned what people like you were supposed to be like and what people unlike you were supposed to be like. Your culture, your religion or lack of a religion, the tone you grew up in, the things that hurt you, all of that was training data. And just like the AI, you didn't choose any of it.

So imagine for a moment a kid who grows up in a house where love seems to show up when things are going well. Good grades, praise. Helping out, a smile. A bad report card, silence at dinner. Nobody ever has to say, "You have to earn love here," because nobody would. The kid just learned that anyway, just by paying attention to what follows what. Little cues.

Now fast-forward thirty years. That kid is an adult who can't rest. Every time they're not being productive, they feel a low hum of anxiety that they can't explain. And if you ask them about it, they'd say, "That's just who I am. I'm a hard worker," and they'd believe it. But it's not really who they are. It's what they were trained on. The pattern was learned so early that it doesn't feel like a pattern anymore. It feels like a personality.

So this is where I want to introduce a question that we'll come back to at the end of the episode. When a strong reaction shows up, the question we usually ask is, "Is this me?" or, "Is this true?" A more useful question might be, "What was this trained on?"

Not Fake, Made

Now, I want to connect this to a Buddhist teaching that often gets misunderstood, and that's the teaching of no self. When people hear that Buddhism teaches there is no self, the first reaction is usually something like, "Wait, so I don't exist? Then who's listening to this podcast?"

But that's not what it means. The teaching of no self isn't the claim that you aren't here or that you don't exist. You're obviously here. You have a body, you have thoughts, you have a history and a name. What the teaching points to is that the self isn't a fixed, solid thing at the center of it all. It's something that gets put together moment by moment out of other things.

And I think the word artificial can actually help us here. We usually hear artificial and think that means fake, but the word comes from Latin, and the original word meant made by skill. Art plus make. Something artificial is something that was made.

Think about an artificial lake, a reservoir sitting behind a dam. It's still a real lake. You can swim in it, you can go fishing, you can drown in it. There's nothing fake about the water. It's just a lake that was made rather than one that was always there. That's the difference. The self is the same. It's not that it's fake, it's that it's made.

And the Buddha used a word that means almost exactly the same thing. The word is sankhara, and it's usually translated as formations or conditioned things. Literally, it means something like put together. So the Buddha used it to describe everything that arises from causes and conditions, and the sense of self is right at the top of that list.

So when Buddhism says the self is constructed, that's what it's saying. It's not that it's an illusion in the sense of not being there, but that it was put together. It was assembled from everything that came before, and it's still being assembled right now, even as you listen to this.

The idea here is that made isn't bad. Made is what makes a thing useful. An artificial lake can water a whole valley. A constructed sense of self is what lets us have a name, and keep a promise, and show up for our kids, and remember where we parked. The problem was never that the self is made. The problem is when we forget that it was made. When we forget, we stop questioning it, and then it just runs.

And the self isn't the only thing like this. I talked about this in the Right View episode a while back, but we live surrounded by things that humans constructed. Money, borders, our reputations, the stories about who we are and who they are. None of those existed until people made them, and they're incredibly useful things. But when we forget that we made them, we start mistaking the story for reality, and we let the story decide how we think and act. The self is the most personal example of that. It's the constructed thing that we're living inside of, and it's not fake, it's just made.

The First Fear: Deception

So now let's go back to the worries that people have about AI. If you listen to how people talk about it, most of the fears come down to three. We're afraid AI will deceive us, we're afraid it will manipulate us, and we're afraid it will make decisions without our consent. What I want to suggest is that the self has been doing all three of those things for as long as we've been alive, and let's look at them one at a time.

The first fear is that AI will deceive us, and there's a specific way it does that, which has a name. It's called hallucination. That's when you ask an AI a question, and it gives you an answer that sounds completely reasonable, confident, well-written, and yet it's just flat out made up. It might invent a quote that nobody ever said or recommend a book that doesn't exist.

The important thing to understand is that it isn't trying to lie. Remember, it's just predicting what a good answer would probably sound like. Sometimes what sounds right and what is right are two different things, and the AI can't tell the difference.

Now, it turns out our minds do this too, and there's a famous set of experiments that shows it really clearly. So our brain has two halves, a left side and a right side, and they're normally connected so that they can share information. And years ago, for some people with severe epilepsy, doctors would cut that connection as a treatment. And these people went on to live fairly normal lives, but it meant researchers could do something unusual. They could show a picture to just one half of the brain without the other half knowing about it. And for most people, the half that does the talking, the half that puts things into words, is the left side.

So in one experiment, they showed the left side of a patient's brain a picture of a chicken's foot, and they showed the right side, the side that doesn't talk, a picture of a snowy yard. Then they asked him to point to matching pictures with each hand. One hand pointed to a chicken. The other hand pointed to a snow shovel. Makes sense so far?

Then they asked him, "Why did you pick the shovel?" Now, the talking side of his brain never saw the snow. It had no idea why his hand picked a shovel, but it didn't say, "I don't know." Without hesitation, he said, "Oh, that's simple. The chicken foot goes with the chicken, and you need a shovel to clean out the chicken shed."

Now, he wasn't lying. He believed it. His brain made up a reasonable-sounding explanation on the spot and handed it to him as the truth. That is a hallucination. And the researchers came to see that this isn't something that only happens in unusual brains. It's something the explaining part of the mind does all the time in all of us. We do something, and then a voice explains why we did it. It feels like the reason came first, but often it came after.

And I think most of us know what this feels like when we look back honestly. We snap at someone, and a moment later, the voice has a whole case ready, right? Like, "They were being unreasonable. I was tired. Anyone would have reacted that way." And yes, maybe some of that is true, but the voice didn't make the decision. It's writing the explanation afterward with complete confidence. Again, the voice that explains us isn't necessarily the part that decided.

The Second Fear: Manipulation

So the second fear is that AI will manipulate us. It'll learn what makes us react and use that to steer what we do. Now, remember how AI works. It predicts what comes next based on patterns from the past. Our minds do exactly the same thing, and a lot of the time, they use those predictions to steer us.

So here's an example I think almost everyone has experienced. Let's say you get a text message from someone important to you, your boss, your partner, a close friend, and it just says, "Can we talk later?" That's it. Four words. Before you can even put the phone down, your mind has already finished the sentence for you, right? "Uh-oh, something's wrong. I'm in trouble. They're upset with me." You might even start rehearsing what they're going to say and what you're going to say back. Maybe your heart rate goes up, your stomach gets tight. All of that from four words that on their own don't say anything bad at all.

That's the keyboard again, right? The phone keyboard. Your mind saw the beginning of a sentence, and it predicted the ending based on patterns from the past. If you've had conversations that started that way and went badly, the prediction leans that way.

And here's the manipulative part. The mind doesn't hand you that prediction labeled as a prediction. It hands it to you as a fact, right? Wrapped in a feeling strong enough to change what you do next. Maybe you get defensive before the conversation even starts. Maybe you avoid that person the rest of the afternoon.

In Buddhism, there's a teaching called the two arrows. The first arrow is the thing that actually happens. The second arrow is the story and the reaction that we add on top of it. In this example, the text message might not have been an arrow at all, but the second arrow was the whole thing. And our own mind is the one that fired it. And now we're experiencing all this anxiety, and we don't even know what any of this is about yet. So there's the mind manipulating us.

The Third Fear: Decisions Without Our Consent

The third fear with AI is that AI will make decisions for us without our consent. It'll quietly run things in the background, and we won't even know that a decision was made.

Well, once again, the self does this constantly. There's research suggesting that something like forty percent of what we do each day isn't really decided in the moment. It's just habit. And I don't mean that in a bad way. Habit is efficient. You don't want to decide from scratch how to brush your teeth every morning.

But think about how often it shows up in places that matter. Somebody says something in a meeting, and before you've had a chance to even consider it, you've already responded in a way that sounds a lot like how your dad used to respond to you. Or you open your phone to check one thing, and twenty minutes later, you look up, not quite sure how you got there, and you didn't choose that, right? It was chosen. It was a pattern that was running, and you just went along with it.

So the idea here is that a lot of our life is being run by a model we didn't train and have never really inspected, and that's the whole reason that this teaching matters. Again, the sense of self is not fake. It's made. Anything that was made can be looked at.

The Unexamined Self

So let's bring these together, because this is really the heart of what I wanted to share in today's episode. We worry, rightly, about a technology that could deceive us, manipulate us, and make decisions without our consent. But we've been living with something that does all three, and most of us have never thought to question it. We just assume the voice is who we are.

And an unexamined life is dangerous in at least three directions. It's dangerous to ourselves, because we suffer from stories that we never question. The kid who learned that love has to be earned grows into an adult who can't rest and never once asks, "Where did that come from?" That's a lot of suffering over a whole lifetime from one piece of training data.

It's dangerous to others. A self needs something to define itself against. If I'm gonna be someone, there has to be someone that I'm not. That's where us and them comes from. And I think if you look at nearly every time one group of people has harmed another, you'll find a very confident story running underneath it about who we are and who they are. Nobody experiences themselves as the villain. From the inside, the story always makes sense.

And then, of course, it's dangerous to the planet. A self that experiences itself as separate from everything will tend to treat everything as a resource. When I forget that I'm part of the web of life, the forests and the rivers become things out there rather than things I depend on and belong to.

Now, put that unexamined self behind the most powerful tools humans have ever built. That's the megaphone. All three of those dangers get louder, and they travel faster.

Interpretability Turned Inward

So what do we do about it? Well, here's something I find really helpful, and it comes from the AI world itself. One of the surprising things about these AI systems is that even the people who build them can't always explain why they say what they say. The patterns are buried so deep in all that training that nobody can simply open it up and read it.

So there's a whole field of research devoted to exactly that problem. It's called interpretability. It's the work of looking inside an AI to understand why it produced the answer it did. And the reasoning is simple: a system you can't see into is a system you can't fully trust.

Mindfulness is interpretability turned inward. When we sit and meditate, we're not trying to become someone else. We're not trying to shut off the mind. We're trying to sit with the model that's running and watch how it works. We see a thought arise, we see a feeling, we follow it, we see the urge to react, and then slowly we start to understand our own outputs, our own patterns.

Again, I want to be clear that the goal isn't to get rid of the self. We don't want to turn the model off. It's useful. A tool like Noah AI that I built, it's a useful tool, and I use it, but I'd never let it make my decisions for me without checking its answers. We give that kind of caution to our tools all the time. We almost never give that kind of caution to our own minds, and we should.

The Practice: What Was This Trained On?

So here's the practice for this week, and it's built around the question from earlier: What was this trained on?

So first, notice when you experience a strong reaction: irritation, certainty, defensiveness, dread, any kind of reaction. The stronger and faster it shows up, the better. Those are the ones running on the deepest training.

Next, name what it's predicting. Say it to yourself as a prediction, not as a fact: "My mind is telling me that this means I'm in trouble," or whatever it is the mind is telling you, right? It could be, "My mind is telling me they don't respect me." Just putting "my mind is telling me" in front of it already creates some space.

And then three, ask, what was this trained on? Where did I first learn to see it this way? Whose voice does this actually sound like? Is this true, or is it just a pattern that's familiar?

And that's it. You don't have to fix anything. You don't have to argue with the reaction or change the reaction. All you're doing is looking inside the model. Try it once a day, and if you want, before bed, write down one reaction that you caught and where you think it came from.

And please don't be discouraged when you miss most of them, because you will. I do. The training runs deep, and it runs fast. So catching one, that's the practice working, right? Because we've got these things running all day long, and it's all pretty much automatic, like we're running on autopilot. So the idea here is, let me at least pause from time to time to see if I can detect one of these things and see where it came from.

Strive On with Heedfulness

So going back to where we started, when I'm using the tool that I built, Noah AI, and it tells me something, I know that I need to check it. I know that, yes, it's fluent and confident, and it's trained on the past, and that fluent and confident isn't the same as being true. And if you've used a tool like ChatGPT, you know the same. It'll give you a very confident answer, but that doesn't mean that it's right.

What I realize is that I've never really given the same caution to the voice in my own head. That's the lesson I'm hoping you'll take away from this, that the voice in your head has been making decisions for you way longer than any AI tool has.

Remember, the Buddha's last words, as they've been passed down to us, were, "All conditioned things are subject to decay, so strive on with heedfulness." And that phrase, conditioned things, again, it's the same word we talked about earlier, sankhara. It means the made things. So his final instruction wasn't to believe anything. It was to stay awake, to realize what are the things that have been made, including the one that feels most like us, which is the self.

The self isn't the enemy, and it isn't an illusion in the sense of not being real. It's not fake. It's made. And what's made can be seen, questioned, and even remade differently.

So I'll leave you with a few questions. What am I sure of today that I was simply trained to be sure of? Why am I so sure of that?

Another one is, whose voice am I mistaking for my own? Do you ever recognize that that voice in your head, maybe it's not yours? It's the voice you grew up with. It might be the voice of a parent or an authoritative figure from your youth.

And then the last question, this is one I keep asking myself: What story am I living inside right now that I've never thought to question?

That's all I have for today's episode. Thank you for listening. Until next time.

Listen, Learn, and Grow

Visit secularbuddhism.com for more information, show notes, and additional resources.