But what if an AI could learn like a baby? AI models are trained on vast data sets consisting of billions of information points. Researchers at Latest York University desired to see what such models could do after they were trained on a much smaller data set: the sights and sounds experienced by a single child learning to speak. To their surprise, their AI learned rather a lot—due to a curious baby called Sam. 

The researchers strapped a camera on Sam’s head, and he wore it on and off for one and a half years, from the time he was six months old until a bit after his second birthday. The fabric he collected allowed the researchers to show a neural network to match words to the objects they represent, reports Cassandra Willyard on this story. (Value clicking only for the incredibly cute pictures!) 

This research is only one example of how babies could take us a step closer to teaching computers to learn like humans—and ultimately construct AI systems which might be as intelligent as we’re. Babies have inspired researchers for years. They’re keen observers and excellent learners. Babies also learn through trial and error, and humans keep getting smarter as we learn more in regards to the world. Developmental psychologists say that babies have an intuitive sense of what’s going to occur next. For instance, they know that a ball exists regardless that it’s hidden from view, that the ball is solid and won’t suddenly change form, and that it rolls away in a continuous path and might’t suddenly teleport elsewhere. 

Researchers at Google DeepMind tried to show an AI system to have that very same sense of “intuitive physics” by training a model that learns how things move by specializing in objects in videos as a substitute of individual pixels. They trained the model on a whole bunch of 1000’s of videos to find out how an object behaves. If babies are surprised by something like a ball suddenly flying out of the window, the speculation goes, it’s because the thing is moving in a way that violates the infant’s understanding of physics. The researchers at Google DeepMind managed to get their AI system, too, to point out “surprise” when an object moved otherwise from the best way it had learned that objects move.

Yann LeCun, a Turing Prize winner and Meta’s chief AI scientist, has argued that teaching AI systems to watch like children is likely to be the best way forward to more intelligent systems. He says humans have a simulation of the world, or a “world model,” in our brains, allowing us to know intuitively that the world is three-dimensional and that objects don’t actually disappear after they exit of view. It lets us predict where a bouncing ball or a speeding bike might be in just a few seconds’ time. He’s busy constructing entirely latest architectures for AI that take inspiration from how humans learn. We covered his big bet for the long run of AI here.

The AI systems of today excel at narrow tasks, equivalent to playing chess or generating text that appears like something written by a human. But compared with the human brain—essentially the most powerful machine we all know of—these systems are brittle. They lack the type of common sense that might allow them to operate seamlessly in a messy world, do more sophisticated reasoning, and be more helpful to humans. Studying how babies learn could help us unlock those abilities. 

Deeper Learning

This robot can tidy a room with none help

Robots are good at certain tasks. They’re great at picking up and moving objects, for instance, they usually’re even convalescing at cooking. But while robots may easily complete tasks like these in a laboratory, getting them to work in an unfamiliar environment where there’s little data available is an actual challenge.