Tavus' AI avatar fools 48% of test subjects into thinking it's human
tavus' griffin ai model convinced 48% of participants it was a real person after a one-minute video call, showing significant improvement over previous systems. this raises questions about ai's potential impact on human interaction.
tavus has introduced griffin, a human interaction model designed to understand and carry on face-to-face conversations in real time. griffin processes speech, facial expressions, tone of voice, gestures, and pauses while both receiving and generating video. in a tavus study, 48% of participants believed griffin was a real person after a one-minute video call, compared to previous systems that maxed out at 2%. griffin scored 3.83 points in an independent nvidia test, with actual humans scoring 3.92 and the previous best ai model hitting 2.80. a preview called griffin-lite is available to select testers, with potential use cases including tutoring, practicing difficult conversations, and camera-based tech support. tavus, a san francisco ai startup founded in 2020, has raised about $64 million.