Emotionally intelligent AI can recognize 48 human emotions, developers say

Back in April, the startup Hume AI released a demo of what it believes to be the first emotionally intelligent artificial intelligence voice: “EVI,” the abbreviation for “Empathic Voice Interface.”

Talking to EVI is in many ways similar to chatting with one of the newer generations of AI-powered voice assistants – think Siri or Alexa, but with ChatGPT functionality. You can ask EVI to read you a poem or explain what caused the French Revolution, and a soft, gentle voice will recite a haiku or begin a mini-lecture about 18th-century France. EVI also has the bugs and lags you often find in AI voice bots.

What makes EVI different is that after you ask for a poem, the screen displays a list of nuanced emotional expressions that the AI ​​supposedly recognizes in your voice, from discomfort and confusion to contempt and surprise. Hume AI says it analyzes up to 48 different emotional expressions.

“These 48 go far beyond what emotion scientists have studied before because we can collect data on a much larger scale to train these models and capture a much greater variety of nuances,” said Alan Cowen, founder of Hume AI.

Hume’s empathic AI is trained using podcasts, other media, and recordings of psychological experiments it has conducted. The company also sells AI that analyzes facial expressions, all of which are designed to help businesses and chatbots better understand their customers. EVI costs about $0.10 per minute, though the price can vary depending on the customer.

“We have clients in customer service, like Lawyer.com. And large technology companies use various technologies that we have developed,” Cowen said.

Lawyer.com uses Hume to improve its 1-800 hotline. Call centers are a natural fit for technology that recognizes human frustration.

But Cowen has bigger ambitions: personal AI assistants that truly understand what you want and are optimized for your well-being.

“It learns from you over time and is therefore more tailored to you,” Cowen said.

An AI trained on your voice and facial expressions might one day say to you, “Hey, have you noticed that you’re tired and irritable around 3 p.m. every day?” That sounds helpful until a beguiling robotic voice gently reminds you that Starbucks Frappuccinos are half price until 4 p.m.

“In certain emotional states and at certain times of the day, we are more likely to make unnecessary purchases,” says Ben Bland, who helped develop industry-specific ethical standards for empathetic AI at the Institute of Electrical and Electronic Engineers.

He also fears that AI will do to our emotions what the smartphone did to our attention span.

“If you have a computer game that adjusts its settings depending on how excited you are and how much fun you are having playing the game, you can become addicted to the game,” Bland said. “Emotional desensitization can occur.”

Of course, all of Bland’s nightmare scenarios assume that a robot can actually make scientifically reliable inferences about your emotional expressions. But that is still up for debate.

“I don’t think these technologies work as well as their marketers sometimes claim,” said Andrew McStay, director of the Emotional AI Lab at Bangor University in the UK.

McStay pointed to studies of existing emotion recognition technologies that assign more negative emotions to the faces of black men than to those of white men, thereby perpetuating harmful stereotypes.

However, he said there is also a broader question about what emotions really are and how much variation there is in emotional expression between people and cultures.

“So when you consider what is often said about these technologies, that there is a biological program that can be detected, and we don’t take into account the social or the cultural, that is problematic,” McStay said.