Babies Defeat AI In Test To Establish Common Sense

Babies and artificial intelligence have just gone head to head in an examination of their common-sense perceptions. And it turns out babies have come out as champions in the test.
 
The experiment identified infants’ ability to judge people’s behavior in ways that machines cannot. The team pitted a group of 11-month-old babies against learning-drive neural network models for the study detailed in ScienceDirect.

First, the research team had the human subjects sit through a developed ‘Baby Intuitions Benchmark’ (BIB), created to evaluate infants and machine intelligence. From the test, the babies had to sit through a sequence of videos on Zoom that portrayed simple animated shapes moving on the screen. The video displayed the items being retrieved and other movements that mimicked human behavior and decision-making. 

On the other hand, the team trained the machine systems to recognize patterns, replicate human intelligence, and evaluate the model’s reactions to the same video.

Ultimately, the findings discovered that infants could anticipate human behavior even in simple animation. The kids recognized these actions as being motivated by certain goals. For example, how the shapes tried to move around efficiently in different environments and how the objects were retrieved from various points on the screen.

The models, instead, demonstrated that they could not understand the motives behind such instances and hence were deemed to have lacked the common sense that babies possess.

The test aims to bridge the gap between human intelligence and artificial intelligence by conducting studies such as this to help fill the areas that AI lacks.
 
 
 

[via TechTimes and SciTechDaily, Photo 113790805 © Lazarenka Sviatlana | Dreamstime.com]