With their bright visuals and slickly animated babies or colourful trains, you might think you’ve stumbled onto your average online video teaching kids their ABCs.
But it might actually be an example of nonsensical content mass-generated with artificial intelligence known as AI slop spreading on YouTube that’s alarming child development experts and advocacy groups like Fairplay. This month, the group sent YouTube’s parent company Google a letter raising concern about artificial intelligence slop, and requesting changes to how AI videos are displayed and distributed on the platform.
AI slop “harms children’s development by distorting their sense of reality, overwhelming their learning processes, and hijacking their attention,” the letter states.
Italian brain rot videos and other AI “slop content” previously ensnared older audiences, but experts say baby slop is a more recent wave of generative AI-created videos directly aimed at toddlers and preschoolers.
The videos might have bright colors, repetitive sounds and familiar characters from popular movies and TV shows, said Keri Ewart, an associate professor of education at Wilfrid Laurier University who researches AI literacy.
However, she says such content “doesn’t really have any developmental, cognitive coherence or intention to it.”
What does AI baby slop look like?
A closer look at the videos often reveals classic signs suggesting they’re AI-generated.
Characters or objects may suddenly become something else (a red car appears different from scene to scene) or behave in bizarre ways (a child on a scooter disappears into the ground or streams of bubbles underwater float down instead of up). Real letters and numbers might have gibberish mixed in.
The expanded YouTube video description might reveal that the video includes “altered or synthetic content” — but not every creator discloses this.
Because the toddler and preschooler brain is still in the early stages of understanding the world and developing language skills, Ewart suggests that some AI videos are so confusing that children “cannot make any type of meaning from that.”
According to child content creator Carla Engelbrecht, these kinds of videos can sometimes model dangerous behaviour — for example, somebody grabbing a hot pan with their bare hands.
“I’ve seen content where it is modelling behaviours you don’t want a child to do,” said the California-based creator who’s worked with Sesame Street, Netflix and PBS Kids for 25 years.
Dr. Michelle Ponti, chair of the Canadian Pediatric Society’s digital health task force and lead author of its screen guidelines for kids under five, is blunt when describing these videos.
“It’s garbage and young children cannot learn from garbage,”said the London, Ont.-based pediatrician.
“We know what promotes early learning and that’s that face-to-face contact with a loving caregiver, that back-and-forth interaction with an actual human that can make the connections and help that child learn,” she said. “AI slop is just a mishmash of junk.”
What is too much screen time for kids?
Ponti recently co-signed the Fairplay letter, which calls on Google and YouTube to make changes, like clearly labelling all AI-generated content and prohibiting AI-generated content on YouTube Kids.
In January, YouTube CEO Neal Mohan said “low-quality, repetitive” AI content is on the company’s radar.
In a statement to CBC News this week, a YouTube spokesperson said the company has “high standards” for YouTube Kids content and limits AI-generated content in the app to “a small set of high-quality channels.”
According to the spokesperson, creators on the main platform must disclose realistic-looking AI content and indicate if they’ve used YouTube’s AI tools. The spokesperson also noted that a labelling strategy for YouTube Kids is in progress.
CBC News attempted to contact several YouTube channels creating videos aimed at preschoolers that appear to be AI-generated, but did not receive any response in time for publication.
In high-quality kids’ programming like Mr. Dressup, Mister Rogers’ Neighborhood and Sesame Street where language is purposeful, Ewart says “repetition with intention” provides “opportunities for children to develop comprehension.”
Traditional TV and online programming that incorporates child development typically follows multimedia, design and storytelling principles, according to Engelbrecht.
For instance, to support literacy, you would spell out a word, show that object and say it onscreen to emphasize the concept for a young brain. It shouldn’t be too fast-paced or busy, either.
“If I want you to focus on ‘the dog is in the yard,’ but then there’s a busy street scene with clouds and planes flying overhead,” Engelbrecht said, those additional elements can be a distraction that drowns out the original intent.
Calgary parent Sarah Rimbey Werklund is adjusting her household’s screen time after she witnessed her nearly seven-month-old Ben become mesmerized by mere seconds of a baby sensory video — which she quickly turned off. She’s seen other parents showing online videos to their kids to get them through challenging tasks, like combing out tangles or brushing their teeth.
Because AI videos are becoming so prevalent online, Rembey Werklund says they can seem unavoidable, like “an avalanche that’s just rushing towards us.”
While there are parental controls on YouTube Kids accounts and parents should be aware of what kids are watching, she notes they simply can’t vet every single second or monitor every hour of the day.
“It’s just changed so much from when we were little kids [watching] Saturday morning cartoons.”
Ponti agrees this goes beyond a question of parental guidance and says appropriate regulation “with child development at the forefront” is needed.
Engelbrecht, who has worked with AI tools for years, has recently dabbled in AI animation, which she says has informed her work helping others improve AI literacy, detect slop and create better videos.
With AI models already being adjusted to reflect copyrighted content, Engelbrecht doesn’t think adding safety guidelines on content for children is too far a stretch.
“A straight ban of AI content — I think the toothpaste is definitely out of the tube,” she said. “But we can still be talking about content standards and quality and applying that.”
She says the current challenge is that platforms and algorithms prize fast-moving videos and visual transitions — which doesn’t align with video production and child development best practices.
However, she says if you prompt generative AI to include such principles, “it will actually do a pretty decent job.”
She believes AI can be used to make online kids content that strikes a middle ground between fast food and Michelin Star cuisine.
She also noted that the studio behind the hugely popular online video series Cocomelon teaming with UCLA scholars to incorporate child development research into their content also has the potential to bring “a lot of influence and effect.”
