Video Turing Test (VTT)

From GM-RKB
Jump to navigation Jump to search

A Video Turing Test (VTT) is an AI evaluation test based on understanding and interpretation of various aspects of video content.

  • Context:
    • It can include understanding characters, objects, places, conversations, behaviors, events, emotions, and commonsense.
    • It can assesses content identity, features, relationships, means, context, sequences, causality, and motivation.
    • It can examines cognitive processes such as recall, recognition, and reasoning in AI.
    • It can involves human players of varying ages and cognitive abilities for a nuanced evaluation.
    • It can uses case studies, such as Korean TV drama clips, to assess AI's video QA capabilities.
    • It can analyzes results to determine AI's accuracy in answering questions related to different story elements.
    • ...
  • Example(s):
    • ...
  • Counter-Example(s):
    • text-based AI evaluation methods.
    • ...
  • See: AI Video Understanding, Human-Level AI, Cognitive Evaluation in AI.