Bridget has spent over 18 years as a consumer tech reporter, hosting daily tech news shows and writing syndicated newspaper columns. She's often a guest on national radio and television stations, including ABC, CBS, CNBC and NBC.
Generative AI software is popping up everywhere -- and now, it's even in teddy bears. Poe the AI Story Bear is the first talking stuffed toy that uses artificial intelligence software to generate and read aloud stories to children. It hits stores in August for $50.
Los Angeles-based toymaker Skyrocket is launching Poe as part of PLAi (pronounced "play"), its AI-powered toy brand. Skyrocket made a multiyear partnership with AI speech software company ElevenLabs to provide the voice of Poe and future PLAi toys. No two stories Poe tells are alike. Stories are generated using Microsoft Azure and ChatGPT 4o, a chatbot from OpenAI, and served up on demand with a kid-friendly app.
Poe is one of an emerging wave of toys powered by AI software. The trend comes as toymakers explore how AI can keep a toy feeling like new, and potentially hold a child's interest longer. The app and voice of this toy bear are still being finalized, but I was given an exclusive early look to take Poe home and see what it was like for my kids (and for me) to use an AI program as a plaything.
Poe, a stuffed plush that's powered by four AA batteries, needs an app to generate new stories. The audio files created are sent to the bear via a Bluetooth connection, and his mouth moves to "talk" when playing the audio. But Poe doesn't always need the app to play a story. You can save your favorite stories on the bear and play it back at any time by pressing its ears.
The app has a simple design. It walks a child through a series of pictures, representing characters, objects and setting, to choose the ingredients of their story. The options go beyond what's found in a typical fairy tale. Take, for example, the choice of characters: You can blend a story with a witch, social media influencer, alien warlord, zombie and archeologist. The more that's added to the story stew, the stranger it gets.
Behind the scenes, a story prompt is sent to ChatGpt 4o, the chatbot from OpenAI. The team at Skyrocket has a number of guardrails in place behind the scenes to make sure a story doesn't wander into controversial territory, such as talking about murder. Although it wasn't present in my early test, the final app should also allow parents to block out scary themes or characters -- and even be tailored to younger age groups. (Perhaps no zombies for the under-5 crowd.)
The bear's childlike voice is also fabricated by AI, which turns the text to speech. All of this happens within a few moments of hitting the "Make Up Story" button on the app. I found it generates a story in under 30 seconds.
The company says Poe will be able to tell stories in over 20 languages, and the voices from ElevenLabs will resonate with listeners in each language.
The bear itself has no microphones or cameras, and it is not connected to the Internet. However, the safety risk here is not knowing what words an AI chatbot will serve to an impressionable child. To curb those fears, Skyrocket says there are a number of filters and guardrails added in the programming to keep a story appropriate for kids.
Skyrocket CEO Nelo Lucich told me that there's a team adjusting the protections it puts on story creation prompts, and the company is listening to parent feedback. During tests, for example, a story used the word cancer as a synonym for destructive. The company decided this could lead to uncomfortable conversations, and a block was put on that word.
There are also limits put on the length of a story -- not just for the sake of a child's attention span, but because of cost. Generating AI stories is server intensive, and the longer the story, the more it would cost Skyrocket. Stories I generated ran for about 2 or 3 minutes.
There will also be limits to how many stories you can create with the app. The plan is for each Poe to come with credits to create 75 stories. But after that, a parent may need to buy additional credits to keep making stories. Lucich said there is talk of having extra stories pushed out for free over time, so you might not have to pay for new content.
My kids are ages 8 and 5, and I didn't need to explain generative AI to them before playing with Poe. I simply told them that the bear makes up stories from the things you tell it to include in the app.
After listening to few stories, I could tell sometimes my kids were having a hard time picking up on everything that was happening, but they did want to keep creating stories, eventually treating it like a game to see what they could make the bear say. (And cram everything possible into a story to be silly.)
To me, the stories it generates exist in a literary uncanny valley, with stories taking random turns to fit all the parameters of the prompt. It was sometimes difficult to follow every line of a story when the vocabulary felt out of place for a children's story, like the ChatGPT AI was overusing a thesaurus.
Take this moment from a scary time-travel story we made about a princess, a prince, and a thief:
The once vibrant kingdom had fallen under a dark curse by the evil Prince Barron. His sorcery summoned shadowy wraiths to haunt the land, chilling the hearts of Allara's people.
A ghastly apparition materialized beside her, the vengeful spirit of Queen Avelina, Allara's ancestor from centuries past. "The phantom hour draws near," Avelina's raspy voice warned. "The stars aligned to undo this ancient wrong."
Before Allara could respond, the sinister cackle of the thief Kiros echoed through the chamber. With a poof of smoke, he appeared clutching Avelina's chronosphere, a mystical orb used for chronological travel.
I'm told the final programming should help steer a story to be a better fit for a younger audience, and perhaps that also applies to a simpler vocabulary. The inflection and pausing of a voice also matters in getting kids to pay attention. My test bear's voice varied, and when the voice lacked emotion or flare, my 5-year-old son couldn't follow the plot. You wouldn't want to follow a story from the robot voice that calls you about a prescription that's ready for pickup. You want a voice actor.
Skyrocket shared a sample recording of the near-final voice from ElevenLabs, which I included in my video embedded above. That sample did seem to have the ability to pause and stress the right words at the right moments to mimic what a voice actor might do.
It's clear this technology is evolving quickly. Last year, the CEO of VTech told the Financial Times that he saw AI-powered teddy bears reading stories to kids by 2028. But here is Poe, doing just that in 2024. And more parents might use ChatGPT to tell bedtime stories even without the bear. Last month, at Apple's Worldwide Developers Conference, Senior Vice President Craig Federighi highlighted how Apple's future software will make it easy to use ChatGPT to create stories for children without paying for an account with OpenAI.
For over a decade I've covered the big toy shows and industry trends, and for the past two years, there's been a growing discussion around using AI chatbots as toy companions. There's already the $99 Miko Mini, which has an optional subscription fee. The price goes up the more engaging AI technology gets, such as with the $799 Moxie robot.
The experience of testing Poe makes me think about an iconic talking toy bear from my childhood: Teddy Ruxpin, the first animatronic talking toy. Released in 1985, Teddy Ruxpin moved his snout and blinked his eyes in sync with audio stories played from cassette tapes, and the bear came with physical books for kids to flip through and read along. The toy was a smash hit, and it cost a whopping $70, the equivalent of $204 today.
But the story behind Teddy Ruxpin involved a team of writers, musicians, and voice actors (who also sang). It was a fantasy adventure series about friendship, starring Teddy, and it spanned 60 story tapes that were translated into 13 languages.
So when I look at Poe, I see the art that made Teddy Ruxpin special has been exported to a machine. Poe doesn't have his own adventure series, but Skyrocket says the app could make sequels to the stories it generates. That means a child could make multiple stories with the same cast of characters -- although it's hard to say how much they will feel like the same characters when creativity is left to a computer.
Poe isn't good enough yet to take a parent's job of reading stories to their kids before bed. But I can see the positive of using this toy as a creativity tool, almost as something that could spark interesting family conversations. Maybe those strange stories get kids thinking about making their own stories.
It's worth watching how much the storytelling evolves as the bear gets closer to launch. And I'm curious to see if Poe inspires more AI toys down the line.
