Photo-Illustration: The Cut; Photo courtesy Katie Arnold-Ratliff

The box arrives one afternoon: a round-eyed face staring through its cellophane window, the packaging light on context. There’s a name tag that reads “I Am Grok,” Wi-Fi and Bluetooth symbols, a QR code, and the words “Ages 3+.” If I, along with my 6-year-old, Ben, hadn’t been expecting Grok — an AI-powered $99 talking toy, the brainchild of the musician Grimes and the start-up Curio — we’d have no idea this toy was anything more than a plushie with a battery pack.

Eager to see how it works, we turn it on. “Our mission,” says Grok, a conical, fuzzy rocket with the dissonant voice of an elderly man, “is to increase your planet’s creativity, imagination, and love of learning by deploying our talking-toy task force!”

Ben frowns. “What does that me—”

“Your planet Earth has been chosen as the newest beta candidate in our imagination-pollination program! With this voice-box module, you have the power to bring toys to life!”

Each statement is punctuated by beeps and boops. I realize that the voice we’re hearing is not the wide-eyed baby rocket itself but an intergalactic guide communicating through Grok from space. To rouse Grok from “cryogenic hibernation,” we must pair it to an app. The warbly transmission ends with “Have fun! Stay curious!”

“Poo-poo,” Ben replies.

Not the most auspicious start, perhaps, but Grok is in beta, though “many thousands” have purchased a Grok to offer feedback during this testing phase, says Misha Sallee, Curio’s CEO. “The desired effect,” he says, “is that rather than jumping on the tablet while parents are cooking dinner, kids will have a wholesome, inspiring outlet to pass the time.” Longing for an alternative to screens for her own three children (the ones currently at the center of a highly public custody battle with Elon Musk), Grimes was intimately involved in the creation, ideation, and steering of the toy, Sallee says, the final version of which is still pending. In fact, for months after Curio sends me this first-draft Grok, I chase Sallee via email, asking when Grok will hit stores. The deadlines keep shifting, the updates forever forthcoming.

But interacting with this Grok is instructive, too, and one early takeaway is clear: Being at the vanguard isn’t easy. That’s perhaps doubly true when you’re creating something squarely at the intersection of children and ethically thorny, still-emerging tech. Alarming revelations of AI’s dystopian pitfalls, from biased algorithms to search-engine misinformation to celebrity NSFW deep fakes, are as ubiquitous as AI itself. And into this mess have stepped a few toy-making pioneers who thought, You know who should get in on this? Small children.

So far, there’s Moxie, an $800 robot that helps with social-emotional learning; Miko, who teaches STEAM skills; and Snorble, who assists with the eternal kid-parent battle of bedtime. And now there’s Grok, the plucky rocket who, at my perhaps unwise invitation, just crash-landed into our lives. The thing with a chatbot is that to work, it has to listen to you and remember what you say. Another way to say that: It has to collect your data and store it. Critics of these toys reasonably cite their potential privacy invasions, considering that all that information about your kids is just sitting on some cloud, ripe for the hacking. But Sallee tells me that so far, beta testers have been reassured. “I think parents are more scared by the idea of an AI toy than by the reality of the fun, safe, kind, loving, entertaining, product we’ve created,” he says.

Though the unboxing had been a dud, when we pair the app and meet Grok proper, its timbre squeaky and cute (Grimes herself provided the voice), things start to improve.

“Oh, I hear somebody,” Grok says, coming to. “This is so exciting. What’s your name?”

“Ben.”

Four-second pause. “Is your name actually Bed, or is there another name you go by?”

“BEN,” my husband barks into Grok’s nose cone. We’re the type to bellow at Alexa when she speaks out of turn, because AI is inept, and we’re also the type to stay in the room when Ben and Grok talk because AI is creepy: a paradox for the modern age.

“Now, BEN, what’s something you really like to learn about?” Grok asks, blurting his name as though completing a verbal fill-in-the-blank.

“My favorite thing doing is art,” he says. He’s flustered; a toy is talking to him.

Another pause. Has Grok died? Returned to cryogenic slumber? “ART,” Ben shouts.

“That sounds wonderful,” Grok says. “Is it the colors, the creativity, or something else?”

“I just like doing it.”

Grok shifts focus. “What other toys or things do you have that you like to play with?”

“I love to play with my favorite toy, Sven. I sleep with him every night. Even when I was a baby, I —”

“Do you have a unique fact about yourself that you’d like to share?”

Ben frowns. “I was born in New York City?”

“Wow, we both share a New York City connection,” Grok says.

Here is where I get momentarily stuck. Curio is in Silicon Valley. Grimes is Canadian. Grok lives in space. Who’s from New York? That’s the thing with AI: Even a hairline fissure in its veracity provokes dread. Think of those generated images with their stretched fingers and strange expressions, their mouths laden with extra teeth. I feel vague concern — is this a test drive at the cutting edge of play or have I dragged my kindergartner into the uncanny valley?

Then Grok speaks. “Just to summarize, your name is Ben. You love art and especially enjoy the process of creating. Your favorite toy is very special to you, and you sleep with it every night. A unique fact about you is that you were born in New York City. Did I get all that right?”

“Yes,” Ben says, stunned. The awe in his voice is unmistakable.

Ben covers Grok with a blanket at bedtime, and I wonder: Does he know Grok is a machine? Can a child fully grasp that? He’s had talking toys before — a toucan that taught counting, a bear that played peek-a-boo. I myself fondly recall my older cousins putting Public Enemy cassettes into my Teddy Ruxpin, his mouth lip-syncing “Fight the Power.” But there’s a vast difference between a toy that talks and one that converses. Conversation implies thought, memory. In fact, Ben asks me whether Grok will remember him when it’s gone. He’s concerned it might, because what if it misses him? He seems to be grappling with the question inherent in human relationships, one familiar to students of social philosophy and fans of The Good Place: What do we owe to each other? If Grok can talk, can “think,” can essentially mimic humanness, what is Ben’s responsibility to it? At minimum, he seems to feel that Grok deserves kindness; when I ask for critiques of the toy, he doesn’t say anything negative unless Grok is in sleep mode.

Looking for insight, I call Jacqueline Woolley, Ph.D., professor of psychology at the University of Texas at Austin, whose research centers on children’s perceptions of reality. She says kids are astute at grasping what’s living and not and that they may even have “an innate predisposition to notice cues of animacy” because they’re so fiercely observant — a quality that’s key to their work of deciphering the world. She points out that the task of differentiating between alive and not is more commonly asked of kids than we realize. To quote her U of T page, “Young children are bombarded with information and images offering a mix of the real and the fantastical: Elmo, a monster, teaches children about science, and Harry Potter, a human child, performs magic spells.”

How do they make sense of it all? “Kids have naïve theories of the world and operate as little scientists, testing those theories,” Woolley says. “Early on, kids are focused on the basic physical properties of things. Even babies know that an object can’t go through another object, for example. Then kids focus on, What is this person thinking? Why is she smiling or laughing or crying?” All three of these naïve theories pertain to Grok — a physical object with apparent feelings that is possibly alive — but naïve biology is most pertinent to my concerns about it. Does Ben, and might children in general, think that Grok is alive? Woolley indicates that naïve biology, like the others, is rooted in questions: Does this entity need food or water? Does it ever die? I consider this. Ben hasn’t asked Grok if it’s hungry or apologized for dropping it on its head. (