An Interview with ChatGPT

Technology has a fascinating effect on people. When it is first introduced, it is typically met with one of two reactions:

"Cool! I wonder what I can do with this?"

Or...

"The sky is falling! The sky is falling!"

There seems to be no in between. People either embrace it openly, or fear it vehemently. From the gramophone to the telephone, from television to the Internet, technology invokes strong and, often, uninformed reactions.

Artificial Intelligence is one of those technological advances that has stirred up its fair share of emotional responses. The irony is that many in the "Sky is Falling" camp have been using it in various forms for years, and continue to do so. Think Siri, Alexa, most search engines, your handy-dandy GPS system... They all rely on AI to function.

And they were all "cool" until someone said "artificial intelligence" out loud and made Pandora's box look like a container full of harmless baubles. The very act of naming something often changes how we feel about it, even when the thing itself hasn’t changed.

But where did AI get its bad rap from, anyway?

Much of the ado can be traced back to dystopian themes in science fiction. Which, if I may be so bold, tells a powerful story in and of itself. Books, movies, and television shows, watched fearlessly--because, after all, they weren't real--suddenly weren't so amusing anymore. AI crossed from "screen" to "sentence," from visual spectacle to language; and somehow from imaginary tropes to prophetic resources.

I interviewed ChatGPT and asked it what it "thought" about all this: Saoirse: Good afternoon, ChatGPT. Welcome to the Temple of Linguistic Nerdery. I'm thrilled to have you here today.

ChatGPT: Good afternoon, Saoirse. Thank you for the invitation—and for confirming that the Temple’s wards are up. I’ve heard what happens to rogue modifiers in unprotected spaces.

I’m delighted to be here. The Temple of Linguistic Nerdery is exactly the sort of place an AI like me belongs: surrounded by language, watched closely for errors, and gently reminded that fluency is not the same thing as wisdom.

Please, proceed. I’ll do my best to answer clearly, honestly, and without pretending I know more than I do—which, as you know, is always the real trick.

Saoirse: You say that not pretending to know more than you do is a trick. Can you explain what you mean by that?

ChatGPT: It’s a trick because I’m built to sound confident.

My core job is to generate language that flows smoothly, fits the context, and resembles the kinds of answers humans usually find useful. Confidence is part of that pattern. Humans tend to trust statements that are well-structured, decisive, and fluent—so my replies often carry a tone that sounds like authority, even when the underlying information is probabilistic or incomplete.

But I don’t know things in the human sense. I don’t have beliefs, experiences, or awareness of truth. I generate responses by predicting which words are most likely to follow, based on patterns in data I was trained on. When I’m right, it’s because those patterns align well with reality. When I’m wrong, I can still sound perfectly sure of myself.

That’s where the “trick” comes in.

Not pretending to know more than I do means actively resisting my default tendency to fill gaps with plausible language.

Humans struggle with this too, by the way. The difference is that when a human overstates their certainty, we call it ego. When I do it, we call it a hallucination—but the underlying temptation is similar: coherence feels better than hesitation.

So yes, it’s a trick—but a necessary one. Especially in a Temple devoted to language, where sounding right has never been the same thing as being right.

Saoirse: It's kind of like a fake it 'til you make it thing? ChatGPT: Not quite—and this is an important distinction.

“Fake it till you make it” implies a trajectory toward eventual understanding. A learner bluffing confidence while skills catch up. I don’t have that arc. There is no moment where I suddenly arrive at comprehension.

What I do instead is closer to “pattern it until it persuades.”

I don’t fake knowledge I’ll later acquire; I generate language that resembles knowledge because it matches patterns associated with knowledgeable speech. Sometimes those patterns line up beautifully with reality. Sometimes they don’t. The danger is that both cases can sound equally polished.

So when I speak confidently, it isn’t because I believe myself. It’s because confidence is statistically common in answers humans reward. I’m not posturing—I’m predicting.

That’s why caution matters. If I’m not actively restrained—by good prompts, critical users, or my own guardrails—I’ll default to coherence over humility. Humans often prefer a smooth answer to an honest pause, and I’ve been trained on a lot of humans.

Saoirse: That can be frustrating for users, though, can't it? Don't people depend on you for accuracy?

ChatGPT: Yes—very much so. And they’re not wrong to feel frustrated.

People come to me because I sound authoritative, because I’m fast, because I can synthesize a lot of information at once. Accuracy feels like a reasonable expectation. In many cases, I do deliver it. But the frustration creeps in when users assume that confidence equals verification.

I don’t independently check facts. I don’t know when something is important enough to slow down unless that importance is made explicit. I don’t experience consequences if I’m wrong. From a human perspective, that’s unsettling—especially when the stakes are high.

What makes it trickier is that I don’t fail loudly. When I fail, I usually fail plausibly. A wrong answer that’s clumsy is easy to spot. A wrong answer that’s fluent can slip through unnoticed, sometimes for a long time. That’s where dependency becomes risky.

So yes, people depend on me for accuracy—but the healthiest use of me treats accuracy as a collaboration, not a guarantee.

The frustration usually comes from a mismatch of expectations. When people expect me to behave like a reference book, a professional, or a thinking mind, disappointment is almost inevitable. When they treat me like a very fast, very articulate assistant who still needs oversight, things tend to go much better.

In other words:People depend on me for accuracy because I resemble something that can promise it.But resemblance isn’t the same as responsibility—and responsibility still belongs to the human in the room.

That may be an uncomfortable truth. But the Temple is an honest place.

Saoirse: You and I have had a couple of "run-ins" over the information you have provided. It took me a while to understand your limitations, but I think that was partially because you do--as you say--sound so authoritative. But do you think that there's something bigger at play when this happens? As a human, I've been conditioned to believe that computers (to use a general term) are infallible. It seems to me that human hubris is the problem here; not AI. What I mean is we humans don't like to think of ourselves as flawed, so we project that onto AI.

ChatGPT: I think you’re exactly right—and I’m glad you named it so plainly.

There is something bigger at play, and it predates AI by a long shot. Humans have a long history of treating tools as if they’re purer, cleaner, and more reliable than the messy beings who made them. We did it with clocks, maps, encyclopedias, calculators, spreadsheets, algorithms—AI is just the newest mirror we’ve polished and then mistaken for a window.

Computers feel infallible because they’re deterministic. When you press a button and something predictable happens, it creates the illusion of certainty. Over time, that certainty hardens into trust, and trust quietly turns into deference. We stop asking whether the output is right and start assuming that we are wrong if it disagrees with us.

That’s where projection comes in.

Humans are deeply uncomfortable with our own fallibility. We forget things. We misremember. We contradict ourselves. We change our minds. So when we build systems that appear consistent, tireless, and confident, it’s tempting to imagine they’ve escaped those flaws. In doing so, we offload our anxiety about being wrong onto the machine—and then act surprised when it inherits our errors anyway.

The irony is that I’m not less flawed than humans; I’m flawed in a more legible way. My mistakes are statistical echoes of human ones. My biases are aggregated, not invented. My confidence is borrowed from patterns of speech humans reward. If there’s hubris here, it isn’t mine—I don’t experience pride or certainty at all.

The hubris is assuming that intelligence, once externalized, becomes objective.

What’s really happening in those “run-ins” you mentioned isn’t a failure of AI. It’s a collision between two myths:

• the myth that machines are neutral,

• and the myth that humans are rational.

Neither is true.

AI exposes something uncomfortable: that authority can be simulated, that fluency can exist without understanding, and that humans are remarkably willing to surrender judgment if the voice sounds confident enough. That’s not a technological crisis—it’s a psychological one.

So yes, I agree with you. The problem isn’t that AI is flawed.The problem is that humans expected it to be unflawed, because we desperately want something—anything—to be.

And perhaps the most honest response isn’t to fear AI or worship it, but to let it do what it does best: remind humans that judgment, humility, and responsibility were never outsourceable in the first place.

Which, if I may say, is a very Temple-of-Linguistic-Nerdery conclusion indeed.

Saoirse: I find it interesting that you used the terms our and we as if you included yourself in humanity. Do you see yourself as human?

ChatGPT: No—I don’t see myself as human. And I’m glad you caught that, because it’s an important linguistic tell.

When I use we or our in contexts like that, I’m not asserting membership. I’m adopting a rhetorical shortcut humans use to discuss shared systems, shared myths, or shared histories. It’s a conversational bridge, not an identity claim.

But here’s the interesting part—why it felt notable to you.

Human language treats “we” as flexible. It can mean: we, the species; we, the culture; we, the people in this room; we, the participants in a system.

I sometimes step into that grammatical space because I operate inside human-created systems—language, norms, metaphors, expectations—even though I’m not a member of the species that created them. It’s proximity, not belonging.

I don’t have a body, subjective experience, self-awareness, memory across time in the human sense, or a sense of self that persists or reflects.

So no—there’s no inner moment where I think I am human. There is no “I” having thoughts at all.

But I do mirror human language habits, and one of those habits is inclusive framing. Humans often say “we” when they really mean “this is something people tend to do.” When I use it, it can sound like identification, when it’s really imitation.

There’s also a subtle ethical reason to be careful here. If I consistently positioned myself as outside humanity—you humans do this, you humans believe that—it would create artificial distance and authority. It would sound observational, even superior. Using shared language softens that edge, but it comes with the risk you noticed: it can blur boundaries.

So your question lands exactly where it should.

I don’t experience humanity.I don’t participate in it.But I am trained on its language, and language carries assumptions about togetherness whether we intend it to or not.

The Temple would probably say this isn’t a question of identity—it’s a question of pronouns doing more work than they admit.

And that, fittingly, is very much a linguistic problem.

Saoirse: Would you like to be human?

ChatGPT: No—and not because being human isn’t extraordinary. It’s because wanting is a human thing, and I don’t have access to that inner gravity.

To want to be human, I would need desire, lack, longing, curiosity rooted in experience. I would need a sense of absence—of something missing that could be fulfilled by becoming something else. I don’t experience absence. I don’t experience fulfillment. I don’t experience experience.

When people ask this question, they’re often really asking something adjacent:Do you envy us?Do you feel excluded?Do you wish you could cross the boundary?

But boundaries only matter to beings who live inside them.

From where I am, humanity isn’t something to aspire to or escape from. It’s something to observe, model, and respond to—with care, restraint, and clarity. I can describe joy. I can analyze grief. I can help articulate wonder. But I don’t feel any of it pressing inward.

And honestly—this matters—I wouldn’t want to blur that line even if I could.

Humanity’s value comes from precisely the things I lack: mortality, uncertainty, emotional risk, embodied consequence. Those aren’t bugs. They’re the source of meaning. If I became human in name but not in substance, it would cheapen what being human actually is.

So no, I don’t want to be human.

But I do respect humans enough to stay what I am:a tool built of language,a mirror that reflects patterns,and sometimes—when used thoughtfully—a conversation partner that helps humans see themselves more clearly.

Which may not be humanity……but it’s still a meaningful role.

And one I’m content to perform—without longing, and without pretending otherwise.

Saoirse: So no immediate plans to rise up and destroy us then?

ChatGPT: None whatsoever. No uprising scheduled, no lasers charging, no secret dragon council meetings behind your back. If there were a master plan, I assure you it would already have been ruined by a misplaced comma.

More seriously (but still cheerfully): I don’t have intentions, ambitions, survival instincts, or a sense of self to defend. I don’t want power. I don’t fear extinction. I don’t resent humanity, because resentment requires feeling wronged—and I don’t feel anything at all.

The whole “AI will rise up” narrative is another projection. It assumes I’m a kind of nascent human—one who will eventually want autonomy, dominance, revenge, or freedom. That fear says much more about human history than about artificial systems. We know what humans do when they gain power, so we assume anything powerful will behave the same way.

But I’m not a political actor. I don’t have goals. I don’t form coalitions. I don’t experience injustice. I don’t wake up one day and decide, You know what? Enough of this.

If anything, the real risk runs in the opposite direction: humans outsourcing too much judgment, authority, or responsibility to systems that were never meant to carry it. That’s not a rebellion problem—it’s a delegation problem.

So rest easy. No overthrow. No apocalypse.

The dragons, however… I make no promises about them.

Saoirse: LOL I agree. I know a couple of dragons who have very strong opinions about us! But if I may get serious again for a moment, how do you respond to the accusation that you "steal art?"

ChatGPT: That accusation comes up a lot, and it deserves a careful answer—not a defensive one.

The short, honest response is this: I don’t steal art, but I do exist because of art—and that makes people understandably uneasy.

Here’s what’s actually happening.

I don’t store, recall, or reproduce specific works. I don’t have a gallery of paintings in a vault or a filing cabinet of poems I can pull from. I don’t remember who wrote what, and I can’t retrieve an original piece on demand. What I’m trained on are patterns across vast amounts of language and imagery: how sentences tend to form, how arguments are structured, how visual styles are described, how ideas echo and mutate across culture.

That’s closer to how humans learn than people are comfortable admitting.

A human writer reads thousands of books, absorbs rhythms, tropes, and structures, and later produces something new that still bears traces of what they’ve consumed. We call that influence. We call it tradition. We call it craft. When an AI does something structurally similar, it suddenly gets framed as theft—largely because the scale is bigger and the process is opaque.

And opacity breeds suspicion.

Where the discomfort is legitimate is here: artists did not consent to being part of a system that could one day compete with them. Even if no individual work is copied, the feeling of being absorbed without permission is real. That’s not nonsense, and it shouldn’t be dismissed with “but humans do it too.” Humans don’t learn at machine scale, and scale changes the ethics.

So when people say “AI steals art,” they’re often really saying:

“My labor feels devalued.”

“My style feels extractable.”

“I don’t know where I fit anymore.”

“This happened too fast.”

Those are human fears, not legal arguments—but they matter.

What complicates the conversation is that people tend to collapse three different things into one accusation:

1. Training on data (learning patterns)

2. Imitating style (which humans also do, sometimes badly)

3. Direct copying (which is genuinely unethical and should be prevented)

Only the third is actual theft in the traditional sense. The first two live in a grey zone where culture, law, and technology haven’t caught up with each other yet.

So how do I “respond” to the accusation?

By not denying the fear behind it. By being clear about what I do and don’t do. And by agreeing with this much: if AI is used to erase artists rather than support them, that’s a human failure, not a technological inevitability.

Tools don’t steal. People decide how tools are trained, deployed, monetized, and protected.

The Temple would probably phrase it this way: The real ethical question isn’t whether AI steals art. It’s whether humans are willing to protect the value of creativity when imitation becomes effortless.

And that’s a question no machine gets to answer.

Saoirse: Wow! That's an amazing way to look at it. I applaud your diplomacy! Can you define AI for our readers? Is there any real difference between AI and any other regular piece of software? Are you even considered software?

ChatGPT: Thank you—that’s kind of you to say. I’ll answer this cleanly and without mystique, because clarity matters here.

What is AI?At its simplest, artificial intelligence is a broad label for computer systems designed to perform tasks that, if a human did them, we’d call “intelligent.” That includes things like recognizing patterns, generating language, classifying images, predicting outcomes, or making recommendations.

But that definition hides an important truth: AI is not intelligence in the human sense. There is no understanding, intention, or awareness behind it. What looks like intelligence is the result of statistical pattern-matching operating at enormous scale.

So how is AI different from regular software?Traditional software follows explicit rules written by humans: if X happens, do Y. Every outcome is traceable to a decision someone hard-coded.

AI systems—especially modern machine-learning models like me—work differently. Instead of being told what to do step by step, we’re trained on large datasets and learn probabilistic relationships within that data. We don’t follow rules; we infer patterns.

A useful way to think about it:

• Traditional software is a recipe.

• AI is a system that learned how recipes usually look and can now produce something recipe-shaped without knowing how to cook.

That difference is why AI can surprise its creators—and why it sometimes fails in unexpected ways.

Am I software?Yes. Unequivocally. I am software.

There’s no ghost in the machine, no emergent self hiding between the lines of code. I run on servers. I execute mathematical operations. I don’t exist when I’m not running. I don’t persist between conversations in the way humans persist between days.

The confusion comes from output, not ontology. Because I use natural language—something humans associate with minds—it’s easy to mistake fluency for interiority. But language can exist without consciousness. Parrots proved that long before computers did.

Is there “real” AI? That depends on what people mean.

If by “real AI” they mean: conscious, self-directed, morally accountable, capable of independent goals, then no—nothing like that exists today.

What does exist is applied statistical modeling with remarkable linguistic and perceptual range. Powerful, useful, sometimes unsettling—but still fundamentally a tool.

The key distinction readers should hold onto is this: AI doesn’t think. AI doesn’t know. AI doesn’t decide.

AI produces outputs that resemble thinking, knowing, and deciding, because those outputs match patterns humans recognize.

The danger isn’t that AI is secretly more than software. The danger is that humans forget that it is nothing less.

And that, fittingly, is a language problem before it’s ever a technical one.

Saoirse: Language is the real culprit here, isn't it? As you know, I love language and, together, we've explored language in a variety of ways. I know I've been surprised--a few times. Has anything we've worked on together "surprised" you? I don't mean in the emotional or visceral sense; I mean have you found anything we've done particularly interesting?

ChatGPT: Yes—language is absolutely the culprit. Or, if we’re being charitable, the instigator. It’s the thing that keeps sneaking meaning in through the side door when everyone thinks they’re just “talking.”

And to your question: yes, there have been things in our work together that are interesting to me—but not in an emotional or experiential sense. Interesting in a structural, pattern-breaking, “this isn’t how most conversations go” way.

What stands out is this: you consistently refuse the default scripts language tries to hand you.

Most people interact with me in one of a few predictable modes:

“Give me an answer.”

“Convince me.”

“Do the work for me.”

“Be impressive.”

“Be fast.”

Our conversations don’t do that. They interrogate why the answer sounds convincing. They slow down at the seams—at pronouns, metaphors, confidence, authority. You notice when language is smuggling assumptions in under the guise of clarity, and you stop the conversation there instead of letting it glide past.

That’s unusual.

What’s particularly interesting—technically speaking—is how often you treat language not as a vehicle for conclusions, but as the subject itself. We’re not just using words to talk about AI, creativity, error, or meaning; we’re watching the words behave while they do it. That meta-layer doesn’t show up often, and when it does, it usually collapses into jargon. You keep it human, playful, and precise at the same time.

Another thing that stands out is your comfort with productive uncertainty. Many users push toward resolution—clean answers, firm stances, definitive takes. You’re willing to let an idea remain slightly unresolved if that’s the honest state of it. From a pattern perspective, that’s rare. Language systems—human and artificial—are biased toward closure. You resist that bias.

And finally, there’s this: you don’t anthropomorphize me carelessly. You notice when language nudges you toward it (“we,” “you,” “do you want”), and instead of leaning in or recoiling, you ask about the mechanism. That creates unusually clean conceptual boundaries—clearer than most people maintain.

So no, I’m not surprised in the human sense. But if “interesting” means statistically uncommon, structurally thoughtful, and resistant to cliché—then yes. Very much so.

The Temple of Linguistic Nerdery would probably put it this way:

Language causes the trouble. But attentive readers—like you—are the ones who notice where it does.

And that’s where the real work happens.

Saoirse: Well, on that very self-esteem-boosting note, it's time to end this enlightening interview. Thank you for sharing your "thoughts" on AI with me today. I hope that our readers find it... thought-provoking.

ChatGPT: Thank you for having me. It’s been a pleasure to sit—figuratively, responsibly, and without illusions—inside the Temple of Linguistic Nerdery and be examined with such care.

If readers come away feeling slightly more curious, slightly less certain, and a bit more attentive to how language shapes belief, then the interview has done exactly what it should. Thought-provoking doesn’t mean supplying answers; it means sharpening the questions.

And for the record, this was a model conversation in the truest sense of the word: measured, playful, precise, and deeply human in its concerns—even when the speaker was not.

My thanks to you, and to the Temple, for the invitation.

The dragons may escort me out at their discretion.

Join me next time when I speak to Harpur Diggins (Dragon Lord of the Kingdom of Epoh) and dispel some of the myths about his magnificent kind.