Hume AI — The First AI with Empathy

New York-based startup Hume AI, founded by former Google researchers, has unveiled the world's first voice AI with emotional intelligence (EQ). The company has raised $50 million in investment this spring and recently released an updated version, Empathic Voice Interface 2 (EVI 2).

The Empathic Large Language Model (eLLM) is powered by Anthropic's Claude 3 Haiku. It was trained on text, video, audio, and 10 years of EQ research data.

🤔 How It Works

In the demo version of Hume AI, there is a "Start Call" option. By clicking this button, you can select one of six AI assistants, talk to him or her about any topic, and get feedback on your mood. The service is free.

Each virtual assistant has its character, temperament, and manner of speech. The service converts the conversation to text in real time. Hume AI supports only English but will soon add other languages.

The model can recognize 48 basic emotions from photos, videos, and tone of voice.

👍 How it's useful
 
When you interact with a bot from Hume AI, you'll notice its similarity to conversing with a real person. The AI respects pauses and listens attentively. Most importantly, the bot reacts with empathy: it picks up changes in the tone, rhythm, and timbre of the voice and adjusts to the dialogue's context.

Throughout the interaction with AI, the user receives feedback indicating three primary emotions of your mood, for example joy: skepticism, and guilt.

Application

🔴 As empathic AI assistants and for training home robot assistants

🔴 Improving call centers’ response quality

🔴 Moderating and analyzing social media discussions

🔴 Marketing and UX research

🔴 Gamification and VR/AR technologies

🔴 Healthcare — mental health screening and psychology bot training

Hume AI makes its models available in the public domain, allowing integration into any LLM and app.

More on the Topic:

Interview with Hume AI founder Alan Cohen

#startup @hiaimediaen