Research shows that when it comes to common sense cognition, infants outperform AI algorithms – at least for now. New York University psychologists are now striving to make Artificial Intelligence (AI) smarter by seeing if this technology can master commonsense psychology.
Dr. Moira Dillon spearheaded the study, which was recently published in the journal Cognition. Her team tested 11-month olds versus three computerized neural networks in various commonsense scenarios. She uses an example when one reaches for a coffee mug, humans can intuit that the person wants to drink even before it reaches their lips, whereas AI algorithms lag behind in predicting such behaviours.
To bridge the gap, the NYU team wanted their study to provide clues for man-made intelligence to better comprehend common sense. They constructed a visual scenario where a neighbour Bob made a straight line to the mailbox. With this, even an 11-month old would be able to understand that Bob would be expecting something important. On the other hand, a neighbour taking a stroll around the block before heading back to the mailbox conveys a different message; one of wanting to get some fresh air or simply exercise.
However, uttering a simple command to any chatbot, for instance ‘Write a sequel to Titanic starring dolphins’, can often fluster these algorithms and throw off the most creative replies.
This groundbreaking study by NYU is already shedding light on bettering AI and helps these machines become more human. But until then, kids still know the common sense stuff better.
NYU, also known as New York University, is one of the largest private non-profit research universities in the United States. Located in the heart of New York City, NYU-Tandon School of Engineering is a global leader in both research and education.
Widely known for her pioneering work in neuroscience, cognitive and developmental psychology, Dr. Moira Dillon is an associate professor of psychology and neural science at NYU. With a research focus on real world embodied cognition, she has made pioneering contributions in the field of machine learning and robotics.
