AI passes video Turing test using Griffin model

No text in the source message.

Griffin neural network passes video Turing test

Tavus has introduced the Griffin model, capable of simultaneously seeing, hearing, and responding to a conversation partner's actions in real time. During the test, more than half of the participants believed they were communicating with a human rather than artificial intelligence.

Video perception and generation capabilities

The neural network demonstrated the ability to help assemble a Rubik's Cube, rather than merely detecting its presence. The model ranked first on the NVIDIA VideoFDB benchmark in both video generation and perception categories.

This achievement evoked a sense of anxiety among observers.