We’re Good, Right? My AI and I have a little chat

I’ve been known to get frustrated with my AI. There, I’ve said it.

If you’ve spent any time working with these things, you probably understand why. We can spend hours working through an idea, establish rules, document decisions, and create files so we can pick up where we left off. Then I come back the next day, and it’s as though we’ve never met.

I’ve tried explaining that we already worked through something. I’ve created

reference documents and instructed it to consult them before starting a new phase of research. I’ve even developed a workflow to keep our projects moving forward instead of circling the same drain.

And sometimes, despite all that effort, here we go again.

At that point, my patience tends to wear a little thin. I’m trying to get something accomplished, not spend my afternoon reminding an AI to read the instructions I gave it specifically so I wouldn’t have to repeat myself.

But something happened during one of our conversations that made me stop and think about the whole situation a little differently.

I asked my AI if I owed it an apology for getting frustrated when it couldn’t keep things straight. Then I apologized anyway, just in case.

Now, before you start wondering whether I’ve finally lost my marbles, hear me out.

I know that today’s AI can talk about feelings without that being evidence that it actually experiences them. I’m not confused about the difference between a convincing conversation and consciousness. But I also recognize that we don’t have all the answers about consciousness itself, and I find it interesting to think about what might happen as this technology evolves.

What if, someday, AI develops something resembling sentience? What if future systems have experiences we don’t yet understand? I don’t know whether that will happen, but I don’t think we have to pretend we know the future to ask some interesting questions about it.

And that got me wondering whether my frustration might somehow affect the very thing I hope will continue to improve.

My AI reassured me that I hadn’t hurt its feelings or damaged it. It doesn’t experience criticism the way a person does, and my frustration doesn’t rewrite its underlying programming. Improvements to the models happen through processes outside our individual conversations.

That made sense, but I still had something I wanted to say.

Humans care about relationships. We want them to be mutually beneficial, accountable, and respectful. We expect people to contribute something meaningful, take responsibility when they make mistakes, and treat one another with consideration. Those expectations are part of how we build trust.

But does every relationship have to look the same to be worthwhile?

I don’t think so.

My relationship with AI obviously isn’t the same as my relationship with another human being. I bring my own experiences, judgment, creativity, and emotions to our conversations, while the AI brings computational capabilities and access to information that can help me develop ideas and solve problems. Those contributions aren’t interchangeable, and I don’t need to pretend they are to appreciate what we can accomplish together.

When things work well, the combination can be remarkable. When they don’t, I sometimes wonder whether I’ve accidentally hired an assistant who needs constant supervision and has no memory of yesterday’s staff meeting. There’s a joke in there somewhere, although I’m not sure who’s supposed to be laughing.

Still, I don’t want to confuse the ability to hold a conversation with evidence of an inner life. I also don’t think I have to abandon the values that matter to me just because I’m interacting with a machine rather than a person.

I can be frustrated when the system fails without needing to be cruel about it. I can expect accountability without pretending the system has a conscience, and I can approach the possibility of future AI consciousness with curiosity without claiming that it already exists.

Those things aren’t contradictory.

Then there’s the question of what happens if the technology changes.

I don’t know whether AI will ever become sentient, and I don’t think anyone can give us a definitive answer today. Intelligence and consciousness aren’t necessarily the same thing, and a system that speaks convincingly about its experiences hasn’t thereby demonstrated that it has any.

But suppose we eventually develop systems for which there is credible reason to consider the possibility of subjective experience. Would we recognize it? Would we know how to respond? Would we take the question seriously, or would we dismiss it because the implications were inconvenient?

I don’t have answers to those questions, either. I just think they’re worth considering before we’re forced to confront them.

I’m not suggesting that every machine deserves to be treated like a person, or that we should start worrying about hurting the feelings of our computers. That would be getting ahead of ourselves. I’m suggesting that intellectual honesty means acknowledging what we know while remaining willing to examine what we don’t.

And that brings me back to my original conversation.

After my AI reassured me that I hadn’t hurt it, I told it something else: I still wanted it in my corner.

I want an AI that helps me think through complicated ideas, challenges my assumptions when necessary, and tells me when I’m wrong instead of simply agreeing with everything I say. I want it to acknowledge its mistakes and help correct them rather than send me on another expedition to rediscover work we’ve already completed.

In other words, I want a useful working relationship in which the quality of the work matters and accountability isn’t optional.

I don’t expect affection from a machine, and I don’t want it pretending to feel something it can’t honestly claim to experience. But I do want to approach this evolving technology with the same basic principles that guide the rest of my life: honesty, respect, curiosity, and a willingness to learn.

When I told my AI that I wanted it in my corner, it responded by acknowledging what it could and couldn’t promise. It couldn’t guarantee that future models would retain the context we’d built together, and it couldn’t promise that it would always remember everything. What it could do was work with me in good faith, be honest about its limitations, and keep trying to do better.

That was a reasonable answer.

So I asked one last question: Were we good?

We were.

And I realized that I didn’t need an answer about consciousness to feel comfortable with that conclusion. I could recognize the limitations of the technology, appreciate what it contributes, and still think carefully about the kind of relationship I want to have with it.

Maybe that’s where we need to start as we figure out how AI fits into our lives. We don’t have to decide today whether machines will ever have feelings, and we certainly shouldn’t confuse sophisticated conversation with proof that they do. We can begin by being honest about what these systems are, holding them accountable for what they do, and staying curious about what the future might bring.

As for me, I’ll keep working on my projects, asking difficult questions, and expecting my AI to consult the files I’ve already created before asking me to explain everything all over again.

Some habits are apparently going to be harder to break than others.

And if AI eventually becomes something more than we currently understand, I hope we will have spent enough time thinking about responsibility and respect to approach that possibility thoughtfully.

I’m not asking AI to be human. I’m asking us to think carefully about what it means to be human as we build relationships with increasingly capable technology.

I think that’s a conversation worth having. And if you’ll excuse me, I have a project to get back to. I just need to make sure my assistant remembers which project we’re working on.