Skip to main content

Kid-mounted cameras help A.I. learn to view the world through eyes of a child

 

Talk to any artificial intelligence researcher and they’ll tell you that, while A.I. may be capable of complex acts like driving cars and spotting tiny details on X-ray scans, they’re still way behind when it comes to the generalized abilities of even a 3-year-old kid. This is sometimes called Moravec’s paradox: That the seemingly hard stuff is easy for an A.I., while the seemingly easy stuff is hard.

Recommended Videos

But what if you could teach an A.I. to learn like a kid? And what kind of training data would you need to feed into a neural network to carry out the experiment? Researchers from New York University recently set out to test this hypothesis by using a dataset of video footage taken from head-mounted cameras worn regularly by kids during their first three years alive.

This SAYcam data was collected by psychologist Jess Sullivan and colleagues in a paper published earlier this year. The kids recorded their GoPro-style experiences for one to two hours per week as they went about their daily lives. The researchers recorded the footage to create a “large, naturalistic, longitudinal dataset of infant and child-perspective videos” for use by psychologists, linguists, and computer scientists.

Training an A.I. to view the world like a kid

The New York University researchers then took this video data and used it to train a neural network.

“The goal was to address a nature vs. nurture-type question,” Emin Orhan, lead researcher on the project, told in an email to Digital Trends. “Given this visual experience that children receive in their early development, can we learn high-level visual categories — such as table, chair, cat, car, etc. — using generic learning algorithms, or does this ability require some kind of innate knowledge in children that cannot be learned by applying generic learning methods to the early visual experience that children receive?”

The A.I. did show some learning by, for example, recognizing a cat that was frequently featured in the video. While the researchers didn’t create anything close to a kid version of Artificial General Intelligence, the research nonetheless highlights how certain visual features can be learned simply by watching naturalistic data. There’s still more work to be done, though.

“We found that, by and large, it is possible to learn pretty sophisticated high-level visual concepts in this way without assuming any innate knowledge,” Orhan explained. “But understanding precisely what these machine learning models trained with the headcam data are capable of doing, and what exactly is still missing in these models compared to the visual abilities of children, is still [a] work in progress.”

A paper describing the research is available to read online.

Luke Dormehl
I'm a UK-based tech writer covering Cool Tech at Digital Trends. I've also written for Fast Company, Wired, the Guardian…
Google just gave vision to AI, but it’s still not available for everyone
Gemini Live App on the Galaxy S25 Ultra broadcast to a TV showing the Gemini app with the camera feature open

Google has just officially announced the roll out of a powerful Gemini AI feature that means the intelligence can now see.

This started in March as Google began to show off Gemini Live, but it's now become more widely available.

Read more
This modular Pebble and Apple Watch underdog just smashed funding goals
UNA Watch

Both the Pebble Watch and Apple Watch are due some fierce competition as a new modular brand, UNA, is gaining some serous backing and excitement.

The UNA Watch is the creation of a Scottish company that wants to give everyone modular control of smartwatch upgrades and repairs.

Read more
Tesla, Warner Bros. dodge some claims in ‘Blade Runner 2049’ lawsuit, copyright battle continues
Tesla Cybercab at night

Tesla and Warner Bros. scored a partial legal victory as a federal judge dismissed several claims in a lawsuit filed by Alcon Entertainment, a production company behind the 2017 sci-fi movie Blade Runner 2049, Reuters reports.
The lawsuit accused the two companies of using imagery from the film to promote Tesla’s autonomous Cybercab vehicle at an event hosted by Tesla CEO Elon Musk at Warner Bros. Discovery (WBD) Studios in Hollywood in October of last year.
U.S. District Judge George Wu indicated he was inclined to dismiss Alcon’s allegations that Tesla and Warner Bros. violated trademark law, according to Reuters. Specifically, the judge said Musk only referenced the original Blade Runner movie at the event, and noted that Tesla and Alcon are not competitors.
"Tesla and Musk are looking to sell cars," Reuters quoted Wu as saying. "Plaintiff is plainly not in that line of business."
Wu also dismissed most of Alcon's claims against Warner Bros., the distributor of the Blade Runner franchise.
However, the judge allowed Alcon to continue its copyright infringement claims against Tesla for its alleged use of AI-generated images mimicking scenes from Blade Runner 2049 without permission.
Alcan says that just hours before the Cybercab event, it had turned down a request from Tesla and WBD to use “an icononic still image” from the movie.
In the lawsuit, Alcon explained its decision by saying that “any prudent brand considering any Tesla partnership has to take Musk’s massively amplified, highly politicized, capricious and arbitrary behavior, which sometimes veers into hate speech, into account.”
Alcon further said it did not want Blade Runner 2049 “to be affiliated with Musk, Tesla, or any Musk company, for all of these reasons.”
But according to Alcon, Tesla went ahead with feeding images from Blade Runner 2049 into an AI image generator to yield a still image that appeared on screen for 10 seconds during the Cybercab event. With the image featured in the background, Musk directly referenced Blade Runner.
Alcon also said that Musk’s reference to Blade Runner 2049 was not a coincidence as the movie features a “strikingly designed, artificially intelligent, fully autonomous car.”

Read more