Artificial intelligence has evolved rapidly in recent years, with AI tools increasingly capable of generating realistic images, videos and human-like conversations. Now, a US-based startup has claimed that its AI avatar can pass for a human in a job interview.
According to a report by the New York Post, Bay Area-based startup Tavus said its latest AI model, Griffin, fooled 48% of the 54 participants who interacted with it during a simulated job interview, leading them to believe they were speaking to a real person.
The company said that the results suggest that Griffin can pass the Turing Test, which measures whether a machine can exhibit behaviour indistinguishable from that of a human. Participants who believed the interviewer was human were 79% confident in their assessment, on average, while those who identified it as a bot were 81% confident.
How does the AI avatar work?
According to Tavus, Griffin generates every pixel of every video frame in real time using a single reference image. The technology generates not only the avatar’s face, arms and fingers but also details such as the movement of a chair, the shadows it casts and the background. The avatars are available in male and female appearances and are designed to mimic natural human behaviours, including breathing and blinking.
Hassaan Raza, CEO and co-founder of Tavus, said face-to-face communication involves more than spoken words, with facial expressions, tone, gestures and timing playing an important role.
{{/usCountry}}Hassaan Raza, CEO and co-founder of Tavus, said face-to-face communication involves more than spoken words, with facial expressions, tone, gestures and timing playing an important role.
{{/usCountry}}“Humans are evolutionarily designed to communicate face to face. We speak as much through our words as we do our expressions, tone, gestures and timing. So much of human communication is non-verbal. It’s an art, a dance,” Raza said, as per the New York Post.
Raza said that Griffin addresses some of the unnatural responses that can make AI-generated avatars appear less realistic, including smiling while hearing bad news.
(Also Read: Microsoft CEO Satya Nadella warns against blindly trusting AI, calls for ‘emergency brake’)
AI model can interrupt, respond in real time
Tavus said Griffin can react in real time, interrupt a person, adapt to new topics and handle interruptions without losing its train of thought.
In one demonstration, the AI avatar teaches a participant how to solve a Rubik’s Cube, showing that it can respond to visual cues when needed. In another demonstration, the model is seen successfully playing a game of Simon Says.
The company said the technology could be used in educational settings, including schools, where AI avatars could assist students or serve as tutors.