It’s been almost four years since I started using artificial intelligence. It’s practically part of my life: I’m an AI champion at my company, I code with AI, and I use it for correspondence and server maintenance.

Just now, it dawned on me that the AI has never told me, ‘I don’t know.’ Truly, I’ve been using it almost from the very beginning, but I’ve never encountered that response. In theory, it’s capable of it. With me, something has always arrived. If the first explanation isn’t enough, you can ask for a longer one: ‘Like I’m five.’ ‘Explain it in Esperanto.’ If it’s too long, a shorter one; if you don’t like it, a different perspective. Whether what I get is any good is a separate question, not relevant here. What interests me now is how quickly I got used to this situation. I ask a question, and a response is guaranteed. I realized that I’m already counting on it while typing the question.

We didn’t use to have anything like that. Back then, we were accustomed to ‘pass.’ To not knowing. We couldn’t possibly know everything.

Am I happy about this? No? I don’t even know myself. I like having access to things, and I particularly like not having to organize an entire afternoon’s program around a single point of interest. And honestly, a whip and a curse to anyone who romanticizes those three weeks of sweating in an uncooled library. Just because your shirt sticks to your back doesn’t make what you find any truer. There is nothing to romanticize here. We tend to retroactively bundle miserable working conditions together with performance, then love them both because ‘hey, we worked hard for it.’ Sure, we worked hard, but it could have been cool inside, too.

Because those old, post-analog, ‘still pounding away on a typewriter’ folks drilled it into us: you cannot make easy and fast progress. Because that’s a sin. Science and art only count for something if you pedal until your fingertips are bloody.

Still, thinking about it: research did have a certain scent, and so did the thought preserved among mildewed paper pages. Worrying whether that single copy would still be there? Whether I still had loose change for the copier? Whether this chapter would fit? The snowball method. That was the scientific name for the ‘trial and cherry’ method, where we tried to backtrack the question from the teacher’s sample study. Because you had to start somewhere, and there was no guarantee at all that you’d find an answer there, or the kind of answer that earns top marks. One reference led to another, and in the end, perhaps the title of a book you couldn’t get your hands on, because reasons. The search had a spatial extension. At the edges of my work, the things I never quite reached could safely remain behind. I knew, at least roughly, where that territory began, about which I only had a hunch.

Polite phrases like these are familiar from old essays. I wrote them to my dying breath, too. Because saying ‘I haven’t the foggiest’ isn’t salon-capable in the academic sphere. Instead came, ‘There is no scope here for a detailed discussion of the question.’ Also: ‘This would exceed the framework of the present paper,’ and the queen of water-treading: ‘Investigation of this requires further research.’ One neatly fenced in what one dealt with and left the rest outside. There could be professional discipline in this, and perhaps a bit of relief. You didn’t have to attach a personal confession to every gap in your knowledge. The space ran out, the time ran out, the writing had to be finished. The other person understood this, too. They had to finish something as well. A case just came to mind now, told by a history teacher of mine at university: a historian colleague found something at the very last minute that completely blew apart the argument system he’d built in a book. What did the good father figure do? He put the book back and gave his lecture the next day. The following week, he dug out the book and wrote a refutation of his last week’s lecture. Today, he wouldn’t stand a ghost of a chance of slipping out of the noose like that.

The AI would know.

I’d be interested to know how many of those sentences were backed up by the much simpler sentence: I don’t know this. Not necessarily out of deceit, but rather because ignorance had its own social attire. You could put on a proper sentence to go with it. Through it, I acknowledged the existence of the topic, indicated that someone else would have work to do there, and moved on. The deficiency didn’t vanish; I placed it in the text, it got an acceptable excuse, and I didn’t necessarily have to feel ashamed of it.

The space runs out now, too. Time as well, very nicely, by the end of the day. Only now, something keeps nagging at the back of my mind: I could have asked about this, too.

You really could have spared this much time. You didn’t have to travel for it, order a book, wait for someone else to return it. It would have fit into those two minutes while you were figuring out how to phrase that it doesn’t fit. And with that, a tiny debt of mine was instantly created. The answer exists somewhere, or at least some kind of answer can be generated, and I didn’t bother.

But how long can we torture ourselves with this? Why do we expect perfection from ourselves? We can’t get a PhD in every study. And we can’t always chase Hemingway in every piece of writing either.

This is the part that really occupies me. What kind of mental state results when behind every missing answer we see a missed opportunity? When not-knowing suddenly acquires intent. You could know, so for some reason you don’t want to know. Maybe you’re lazy. Or you don’t care enough, and then own up to it. The help is right there, it works, and it’s even polite. From this point on, it’s hard to claim that the world is big and I am just a single person in it with finite attention. That sounds like such an old-fashioned excuse. A search box was just handed to us for the sheer scale of the world.

Besides, nobody needs to specially nag you about this; you can pull it off all by yourself. The saved time leaves room for yet another question, and the previous answer also had a detail you didn’t quite understand. Well, ask about that one, too. You can ask for it in a different formulation, ask for an example. Every single step is reasonable. All of them together can demand an entirely different workday, a different evening, a different kind of attention. Where is the point where I can stop without feeling negligent?

Is there some question I simply don’t want to deal with today? Does that remain an acceptable reason? If I want to keep some of the free time I’ve gained, even if a few patchable holes remain in my erudition, I’d like it if at the end of an explanation I could decide that enough is enough for today—and that I shouldn’t need a better justification for it than simply: enough is enough for today.

And whose knowledge is this, then? If I can’t recite it instantly, is it no longer mine? It used to be said often that you don’t need to keep everything in your head; you need to learn where and how to look it up. I thought that was a sensible statement, and I still do. A book’s table of contents isn’t compiled so that decent people can learn it by heart. With the fulfillment of that sentence, however, a somewhat strange situation has arisen. I can find help for every question of mine, and they even help me formulate what I’m looking for.

Have we arrived? Is this Canaan? Because if knowing how to search is enough, then we are suddenly overflowing with academics. We only need to write the inaugural speeches, but the AI helps with that, too. It would be good to know precisely where we begin to take back the promise. Searching was still fine, using the found answer too, but now it feels almost as if the whole thing was too easy. How many of my own operations must I perform on it to convince myself: I know this?

If I read it, I understand it. At least while reading, it feels that way. Then I close it and try to recount it, perhaps someone asks back about some entirely insignificant detail. Well, that’s where the situation gets more interesting. The realization that I’ve seen this sentence before can be quite convincing. It’s also easy to feel that the thought behind it is with me, in a usable state. Only that one uncomfortable little detail is missing: what I do with it if the question is framed just a little differently.

Fisher, Goddu, and Keil’s 2015 study of nine experiments examined how we estimate our own knowledge after internet searches. As a result of searching, participants tended to think more highly of what they could explain even without outside help. That was still just internet searching. What interests me is how easily the boundary can blur between what I have access to and what I consider my own. How much does it change when the helper can actually converse with us?

Moreover, I haven’t gathered my own knowledge entirely alone up until now. We learn from books, rely on other people’s sentences, and sometimes an explanation comes to mind years later with the exact same example that helped us understand it. That doesn’t make me doubt that I’ve learned something. You can learn from an AI’s response, too. That’s precisely why this whole thing bothers me: I can’t draw a simple line with the certainty of my own knowledge on one side and the machine’s text on the other. From the same conversation, I might understand one thing and merely skillfully regurgitate another.

Meanwhile, I still need a certain amount of my own knowledge just to notice this distinction in the first place. If I’m at home in something, a too-comfortable explanation or an odd emphasis might catch my attention. If I arrive at a completely new field, how am I to know which crucial question was left out? I’m free to ask for a full overview. It will still be up to me to determine whether it’s complete. Quite a pretty task for someone who asked precisely because they don’t know the field. And of course, one can ask about that, too.

So, then? Does “not knowing” still count as valid? Can “laziness” or “I don’t feel like it” still be acceptable reasons?