Are You Fine?

Conversations with large language models

Did you see that the banana sold for $6.2M? Amazing. And so, so funny. During the auction, the agents representing the remote crypto buyers all looked like they were ready to break out laughing. But enough about bananas. Let’s talk about AI.

Now that these tools have gotten so smart, I thought I’d test one of the large language models, or LLMs, with a deviously difficult question:

When I ask my wife how she's doing, she says, “Fine.” Is she really fine?

The LLM returned an extensive response within seconds. Here’s the edited version:

When your wife says she’s “fine,” it might be helpful to look at the context and her tone to get a better sense of what she really means. Here are some possible interpretations of “fine” and what you might consider doing:

She’s Actually Okay: Sometimes “fine” genuinely means that she’s doing well. If her tone is positive and she seems content, it’s possible that she’s just giving a brief, honest answer.

She’s Hiding Her True Feelings: Sometimes “fine” can be a cover for underlying issues. If she’s been acting differently lately or you sense something is off, she might be using “fine” to avoid bringing up her true feelings. “Fine” can be a way to deflect.

Wow, right?! Keep in mind that I’m not corresponding with another human. I’m conversing with a complex statistical model that looks at my words and rapidly spits out a different mix of words. Daniela and I go into detail about how this works in our book The Mind’s Mirror, but here’s how I see it in pictures.

Let’s start with person-to-person conversation:

Curly (on the left) has an idea, picks some words to share that idea, then speaks them out loud. Flare (on the right) hears the words, thinks, then develops his own idea, X, and comes up with a different set of words to share that back with Curly. Each idea is a product of that person’s mind, experiences, hairstyle, etc.

Now here’s a conversation with an LLM:

Flare has an idea, expresses that in words, then sends them to the LLM. This AI tool, typically based in a cloud data center, ingests the words, runs them through its black box, then spits out some other words that are strangely similar to the sorts of things a human would say. Only Flare isn’t interacting with another human mind.

He’s talking to a black box.

Weird, right?

OK, back to the marriage advice. After our initial exchange, I asked some follow-up questions, and the LLM suggested observing my spouse’s behavior (creepy), creating a supportive environment (fair), being patient (valid), and sharing my own feelings (impossible; that’s why I carry a harmonica.)

The tool also offered a few possible comments. Most were bland. One — "You seem a bit off today" — was incendiary. Clearly this AI has never been married.

Generally, I was impressed, but since I have the mind of a circus clown crossed with a community college professor, I followed up like so:

What if she's eating a donut at the same time?

Now, to be clear, I’m the only one in our marriage who eats donuts, but here’s a slimmed-down variation of the subsequent exchange.

• Make a Light-Hearted Comment: “Ah, a donut! That’s a good sign. But really, how’s everything going?”

• Express Interest Gently: “I’m glad you’re enjoying that donut. I was just wondering if there’s anything on your mind or if you’d like to chat?”

• Be Non-Judgmental: Ensure that your questions or comments come from a place of care.

That makes sense. I guess what complicates this situation is that she's also running really fast.

This situation is pretty unusual, so it might not be the best time for a deep conversation. It’s worth acknowledging the context and finding a way to connect that respects her current state. Let her finish her donut and her run before diving into a serious conversation. She may be juggling too many tasks.

It's interesting that you suggested she might be juggling too many tasks, because she's also juggling while running and eating the donut and insisting she's fine.

In a scenario where she’s juggling, running, and eating a donut, a combination of humor, empathy, and patience is key.

• Running and Juggling: It’s possible she’s using the situation to focus on the fun side of life rather than any underlying issues.

• Make a Light-Hearted Comment: “Wow, you’re really juggling it all right now! I’m impressed. Are you sure you’re okay with everything you’ve got going on?”

By acknowledging her multitasking and offering support, you can gently open the door to a deeper conversation if she’s ready.

The tool suspected I was creating a humorous scenario, which is actually wild, since it suggests an understanding of reality, or at least what constitutes normal behavior, but I insisted this was all really happening. The model then suggested that running alongside her would be supportive. I indicated that I liked this idea, but preferred to travel in a horse-drawn chariot dressed in the garb of a gladiator. Finally I requested a picture capturing this special moment in our lives. The model obliged:

That’s me in the robe, with the microphone, singing Don’t Stop Believing.

Yes, the horse in the background has five legs, I appear to be lacking a left foot, the mix of floating food is strange, and my wife has six left toes, but disregard those little details for a moment. The model knows me well enough to give me a microphone even though I made no mention of karaoke in my prompt. Plus, it generated this strange little illustration in less than twelve seconds. (I’ll explain how this works in a follow-up post on AI art, but it’s in the book, too. And it’s really strange.)

Along with the art, the tool recommended a soundtrack (Wagner, Journey), a post-race dinner date, and an inspirational quote from Nelson Mandela: “The greatest glory in living lies not in never falling, but in rising every time we fall.” Which is weird, because I'm pretty sure Rocky said something like that to his son in Rocky Balboa.

Finally, the LLM closed with these words of wisdom:

Enjoy every moment of your chariot ride, and have an amazing time supporting your wife! 🌟🏇🚀

Okay, so there might not be a human mind behind the generation of that particular mix of words and emojis, and it might have been spat out of a mysterious black box, but I will henceforth be using this blessing in all future wedding toasts.

A note from our sponsor: If you’re looking for a holiday gift for a kid, consider my two latest adventure novels, Sea of Gold and/or its sequel, Sea of Gold: The Duke’s Curse. And let me know if you liked this post by clicking the heart below, or answering this little poll.

Thank you for reading!