Babies Beat AI in 'Commonsense Psychology', According to New Study

According to a recent study by a group of psychologists and data scientists, infants are better than artificial intelligence at determining what drives other people's behavior.

The findings emphasize key distinctions between cognition and computation and identify areas in which current technology is lacking, as reported first by SciTechDaily.

"The novel idea of putting infants and AI head-to-head on the same tasks is allowing researchers to better describe infants' intuitive knowledge about other people and suggest ways of integrating that knowledge into AI," Moira Dillon, an assistant professor at New York University's Department of Psychology, said in a statement.

Baby Intuitions Benchmark

The researchers ran a series of experiments with 11-month-old babies and compared the results to those produced by cutting-edge learning-driven neural network models to gain a fundamental grasp of the differences between human and AI capacities.

They used the previously developed "Baby Intuitions Benchmark" (BIB), a set of six activities that examine common sense psychology. BIB was created to evaluate both infant and machine intelligence and emphasize their differences.

On Zoom, the babies saw a sequence of videos that included simple animated shapes moving around the screen, much like in a video game. The retrieving of items from the screen and other movements of the forms mimicked human behavior and decision-making.

The researchers also created and trained learning-driven neural network models that aid computers in pattern recognition, replicate human intelligence, and evaluated the models' reactions to the same videos.

Their findings show that infants can discern human-like intentions even in the simplified movements of animated forms. Babies infer that these behaviors are motivated by subtle but enduring goals, such as the efficient movement of the same shape in different environments and the retrieval of the same object from any point on the screen.

Surprise Paradigm

By observing how long infants stare at events that contradict their predictions, it is possible to make inferences about the nature of their knowledge and the babies in the study were able to demonstrate such predictions.

The researchers applied this "surprise paradigm" on machine intelligence to quantify surprise measurement from an algorithm and the babies.

The models did not show any indication of understanding the motivations behind such actions, suggesting that they lack the principles of commonsense psychology that infants have shown.

"A human infant's foundational knowledge is limited, abstract, and reflects our evolutionary inheritance, yet it can accommodate any context or culture in which that infant might live and learn," Dillon said, quoted by SciTechDaily.

The findings of the study were published in ScienceDirect.