She adds, “The novel idea of putting infants and AI to the same task allows researchers to better describe infants’ innate knowledge about other people and suggest ways to integrate that knowledge into AI.”
“If the goal of AI is to create flexible, commonsense thinkers like human adults,” says Brenden Lake, an assistant professor in NYU’s Center for Data Science and Psychology, “then machines must draw on the same basic abilities that infants have for setting goals and thinking.” are meant to detect preferences.” and one of the authors of the paper.
It has been well established that infants are fascinated by other people – as evidenced by the amount of time they spend watching others observe their actions and engage with them socially. In addition, previous studies focused on infants’ ‘commonsense psychology’ – their intentions, goals, preferences, and understanding of the rationality inherent in others’ actions – have indicated that infants are able to assign goals to others and respond to others rationally. Expect to chase goals efficiently and effectively. The ability to make these predictions is fundamental to human social intelligence.
What is Commonsense AI?
In contrast, “commonsense AI” – powered by machine-learning algorithms – directly predicts actions. This is why, for example, an ad featuring San Francisco as a travel destination pops up on your computer screen after you read a news story on a newly elected city official. However, what AI lacks is the flexibility to recognize the different contexts and situations that guide human behavior.
To develop a fundamental understanding of the difference between the capabilities of humans and AI, the researchers conducted a series of experiments with 11-month-old infants and compared their responses to those of state-of-the-art learning-driven neural-networks. model.
To do this, they deployed the pre-installed ‘Baby Intuition Benchmark’ (BIB)– Six tasks that test common sense psychology. The BIB was designed to allow testing of both infant and machine intelligence, allowing comparisons of performance between infants and machines, and importantly, an empirical basis for building human-like AI provides.
Specifically, infants on Zoom viewed a series of videos of simple animated shapes moving around a screen, similar to a video game. Shape Actions simulated human behavior and decision-making through retrieval and other movements of objects on the screen. Similarly, the researchers built and trained learning-driven neural-network models—AI tools that help computers recognize patterns and emulate human intelligence—and tested the models’ responses to the exact same videos.
Are infants good at recognizing human-like motivations?
Their results showed that infants could recognize human-like motivations even in simplified actions of animated figures. Infants anticipate that these actions are driven by hidden but coherent goals – for example, on-screen retrieval of the same object regardless of location and movement of that shape efficiently even when the surrounding environment changes. Is.
Infants demonstrate predictions through their prolonged viewing that violate their predictions – a common and decades-old measure to measure the nature of infants’ cognition. Adopting this “surprise paradigm” to study machine intelligence allows direct comparisons between a quantitative measure of algorithmic surprise and a well-established human psychological measure of surprise-infant gaze time.
The models showed no such evidence of understanding the underlying motivations of such actions, which suggests they are missing key fundamentals of common sense psychology that infants possess.
Dillon said, “A human infant’s basic knowledge is limited, abstract and reflects our evolutionary heritage, yet it can accommodate any context or culture in which that infant may live and learn.”Life Style