Artificial intelligence (AI) systems are already way ahead of us in certain areas – playing Go, for example, or processing huge data sets – but in other aspects AI is still far behind human beings, if only by a few months after I’m born.
For example, even young infants instinctively know that an object passing briefly behind another should not disappear and reappear elsewhere. Presented with such a magical act, babies act surprised.
But such a simple rule of continuity, along with other basic laws of physics, was not so intuitive to AI. Now a new study is introducing an artificial intelligence called PLATO that was inspired by research into how babies learn – and can think a lot like a human baby.
PLATO stands for Physics Learning Through Automatic Coding and Object Tracking and is taught through a series of coded videos designed to represent the same basic knowledge that babies have in their first few months of life.
“Fortunately for us, developmental psychologists have spent decades studying what babies know about the physical world and cataloging the different ingredients or concepts that go into physical understanding,” says neuroscientist Louis Pilotto of the DeepMind artificial intelligence research laboratory in the UK kingdom.
“Extending their work, we have built and opened a dataset for physical concepts. This synthetic video dataset draws inspiration from the original development experiments to evaluate the physical concepts in our models.”
There are three key concepts that we all understand from a very early age: permanence (objects will not suddenly disappear); strength (solid objects cannot pass through each other); and continuity (objects move in a consistent manner through space and time).
The data set the researchers built covered these three concepts, plus two additional ones: immutability (an object’s properties, such as shape, do not change); and directed inertia (objects move in a way that is consistent with the principles of inertia).
These concepts were presented through clips of balls falling to the ground, bouncing off each other, disappearing behind other objects and then reappearing, etc. After training PLATO on these videos, the next step was to test it.
When the AI was shown videos of “impossible” scenarios that defied the physics it had learned, PLATO expressed surprise (or the AI equivalent of it): it was smart enough to recognize that something strange had happened that violated the laws of physics.
This also occurred after relatively short training periods, as little as 28 hours in some cases. From a technical perspective, just as in studies of infants, researchers have looked for evidence of violation of expectation (VoE) signals, indicating that the AI understands the concepts it has been taught.
“Our object-based model showed robust VoE effects across all five concepts we studied, even though it was trained on video data in which the specific probe events did not occur,” the researchers wrote in their published paper.
The team ran further tests, this time using objects different from those in the training data. Again, PLATO showed a solid understanding of what should and shouldn’t happen, demonstrating that it can learn and expand on its basic training knowledge.
However, PLATO is not yet at the level of a three-month-old baby. There was less surprise from the AI when it was shown scenarios that did not include any objects, or when the test and training models were similar.
What’s more, the videos that PLATO was trained on included additional data to help it recognize objects and their motion in three dimensions.
It seems that some built-in knowledge is still needed to get the full picture—and this nature vs. nurture question is something scientists are still asking about infant development. The research can give us a better understanding of the human mind as well as help us build a better AI representation of it.
“Our modeling work provides a proof-of-concept demonstration that at least some central concepts in intuitive physics can be acquired through visual learning,” the researchers wrote.
“Although research in some precoital [born in an advanced state] species suggests that some basic physics concepts may be present from birth, in humans the data suggest that intuitive knowledge of physics emerges early in life but can be influenced by visual experience.
The study was published in Nature Human Behavior.