According to a new study published in the journal Cognition, infants eclipse artificial intelligence when it comes to determining what drives people’s motivations. The study, led by a team of New York University psychology and data science researchers, highlights fundamental differences between cognition and computation and points to the shortcomings of today’s technologies. The study also suggests what improvements could be made by artificial intelligence to more fully replicate human behavior.
In a statement released by New York University, Moira Dillon, the paper’s senior writer, said adults, and even young children, can easily make reliable inferences about what drives other people’s actions. She added that current artificial intelligence finds it difficult to draw these conclusions.
Dillon also said that the novel idea of putting infants and artificial intelligence head-to-head on the same tasks allows researchers to better describe infants’ intuitive knowledge of other people and propose ways to incorporate that knowledge into artificial intelligence.
Brenden Lake, one of the paper’s authors, said that if artificial intelligence aims to develop flexible, rational thinkers like human adults, then machines should draw on the same core skills that infants possess in recognizing goals and preferences.
What is sound psychology?
Infants often stare at others for long periods of time to observe their actions and engage in social engagement with them, indicating how fascinated they are by other people.
Studies of the common sense psychology of infants have been conducted, relating to their understanding of the intentions, goals, preferences, and rationality underlying the actions of others. The studies found that young children are able to ascribe goals to others and expect others to pursue goals rationally and efficiently. This ability to make predictions forms the basis of human social intelligence.
What is artificial intelligence with common sense?
On the other hand, common sense artificial intelligence powered by machine learning algorithms directly predicts actions. This is why a person reading a news story about a newly elected San Francisco city official may receive an ad showing the same location as a travel destination. Or, when a person searches for specific types of clothing online, they receive advertisements for similar clothing on social media.
The disadvantage of artificial intelligence is that it lacks the flexibility to recognize different contexts and situations that guide human behavior.
How the study was conducted
The researchers conducted a series of experiments with 11-month-old infants and compared their responses to those obtained through state-of-the-art learning-driven neural network models to develop a fundamental understanding of the differences in the abilities of humans and artificial intelligence.
To achieve this, the researchers used the previously established “Baby Intuitions Benchmark” (BIB). Within the framework of BIB, six tasks are carried out to research everyday psychology. BIB was designed to test both infant and machine intelligence. This allows for a comparison of the performance of infants and machines and provides an important empirical basis for building human-like artificial intelligence.
Infants on Zoom were asked to watch a series of videos featuring simple animated shapes that move around the screen in a manner similar to a video game. The shapes’ actions simulated human behavior and decision-making through on-screen object recall and other movements.
The researchers also built and trained learning-driven neural network models, which are artificial intelligence tools that help computers recognize patterns and simulate human intelligence. The researchers then tested the neural network models’ responses to the same videos the infants were forced to watch.
Infants outperformed artificial intelligence
According to the study, the infants recognized human-like motivations even in the simplified actions of animated shapes. Infants could predict that these actions would be driven by hidden but consistent goals. For example, the infants were able to efficiently predict the retrieval of the same object on the screen regardless of the position it was in, and the movement of that shape even when the environment changed. Infants are able to demonstrate such predictions by observing events over long periods of time.
The researchers found that the neural network models provided no evidence for understanding the motivations underlying such actions. This showed that artificial intelligence lacks the most important basic principles of common sense psychology that infants possess.
Dillon said the basic knowledge of a human infant is limited, abstract, and reflects evolutionary heritage, but it can accommodate any context or culture in which the infant might live and learn.
