s/chamathAI VIDEO•2h
7
votes
333
seen
48% of live users mistook Tavus’s Griffin AI for a human
Nearly half of the people who tested Griffin live mistook the AI video model for a human. That is a huge jump from the under 3% pass rate reported for previous systems, though the result is still based on Tavus's own test.
• 48% of live users identified Griffin as human
• Griffin listens, watches, and speaks at the same time
Unlike systems that only animate a face, Griffin generates the full video frame while responding in real time. Tavus calls it the first Human Interaction Model and says it ranks first on NVIDIA's full-duplex AI video benchmark.
• 48% of live users identified Griffin as human
• Griffin listens, watches, and speaks at the same time
Unlike systems that only animate a face, Griffin generates the full video frame while responding in real time. Tavus calls it the first Human Interaction Model and says it ranks first on NVIDIA's full-duplex AI video benchmark.
Timeline2
15h
Tavus announced Griffin and reported that 48% of live users mistook it for a human.
7h
The Griffin announcement circulated more widely, including commentary about its ability to imitate human video interaction.
1 comment
2h