Particle.news
Download on the App Store
4 ARTICLES
·
2w ago

AI Models Fall Short in Understanding Human Social Interactions, Study Finds

Johns Hopkins researchers reveal that over 350 AI models fail to match human accuracy in interpreting dynamic social cues, highlighting critical gaps for real-world applications.

When it comes to ‘reading the room,’ humans still have a leg up.
The results provide a sharp contrast to AI’s success in reading still images, the researchers said. Credit: Neuroscience News
Portrait of person and ai robot

Overview

  • The study, presented at the International Conference on Learning Representations, tested over 350 AI models against human performance on interpreting three-second social interaction videos.
  • Humans consistently outperformed AI models in understanding social dynamics, with AI struggling to predict human ratings and brain responses accurately.
  • Language models performed better than image and video models but relied heavily on descriptive captions, exposing limitations in direct video analysis.
  • Researchers attribute AI's shortcomings to its design, inspired by brain areas for static image processing, which lacks mechanisms for dynamic scene comprehension.
  • The findings underscore risks for AI in autonomous vehicles and humanoid robots, emphasizing the need for interdisciplinary approaches to enhance social understanding.