Luis Piloto, a computer scientist at Google-owned company DeepMind in London, and his collaborators have introduced an AI-inspired by research on how babies learn. The new AI called PLATO can think like a human baby.
PLATO stands for Physics Learning through Auto-encoding and Tracking Objects. It can learn simple physical rules about the behavior of objects- and express surprise when they seem to violate those rules: much like a baby does.
By observing babies’ gazes, developmental psychologists can assess how well they follow the motion of objects. The duration of the children’s gaze in one spot helps scientists determine how surprised they are when showing them videos of, say, a ball that abruptly disappears.
Scientists trained PLATO with animated videos of simple objects such as cubes and balls. They trained the system on around 30 hours of videos showing simple mechanisms such as a ball rolling down a slope or two balls bouncing off each other.
Piloto said, “Luckily for us, developmental psychologists have spent decades studying what infants know about the physical world and cataloging the different ingredients or concepts that go into physical understanding.”
“Extending their work, we built and open-sourced a physical concepts data set. This synthetic video data set takes inspiration from the original developmental experiments to assess physical concepts in our models.”
The system developed the ability to predict how those objects would behave in different situations.
Piloto said, “In particular, it learned patterns such as continuity, in which an object follows an uninterrupted trajectory rather than magically teleporting from one place to another; solidity, which prevents two objects from penetrating each other; and persistence of the objects’ shape. At every movie step, it predicts” what will happen next.”
“As it gets further into the movie, the prediction becomes more accurate.”
“When shown videos with ‘impossible’ events, such as an object suddenly disappearing, PLATO could measure the difference between the video and its prediction, providing a measure of surprise.”
“PLATO is not designed as a model of infant behavior, but it could be the first step towards AI that can test hypotheses about how human babies learn. “We’re hoping this can eventually be used by cognitive scientists to model the behavior of infants seriously.”
Jeff Clune, a computer scientist at the University of British Columbia in Vancouver, said, “Comparing AI with how human infants learn is an important research direction. That said, the paper does hand-design much of the prior knowledge that gives these AI models their advantage.”
Journal Reference:
- Piloto, L.S., Weinstein, A., Battaglia, P. et al. Intuitive physics learning in a deep-learning model inspired by developmental psychology. Nat Hum Behav (2022). DOI: 10.1038/s41562-022-01394-8