Hume AI Launches “Emotionally Intelligent” Chatbot EVI

New York-based AI startup Hume AI has introduced Empathic Voice Interface (EVI), a conversational AI chatbot with emotional intelligence, capable of understanding and responding to human emotions.

EVI leverages a sophisticated empathic large language model (eLLM) to decipher nuances in tone, emphasis, and pitch in users' voices, enabling it to generate responses that resonate emotionally. What distinguishes EVI is its integration of empathic responses into various applications through a single API, empowering developers to create emotionally intelligent apps across different domains, from transcription services to customer support tools.

Alan Cowen, Hume AI CEO and Chief Scientist, mentions in a company blog that he sees empathic AI as essential to aligning AI with human well-being: “The main limitation of current AI systems is that they’re guided by superficial human ratings and instructions, which are error-prone and fail to tap into AI’s vast potential to come up with new ways to make people happy. By building AI that learns directly from proxies of human happiness, we’re effectively teaching it to reconstruct human preferences from first principles and then update that knowledge with every new person it talks to and every new application it’s embedded in.”

Features

Using a diverse dataset from over 30 countries, Hume's models are trained to recognize cultural nuances, aiming for unbiased interpretation. While Hume primarily relies on its in-house model, it supplements it with external large language models for complex content. This technology simulates human conversation styles, even incorporating pauses and laughter.

Its features, including end-of-turn detection and incorruptibility, ensure smooth and natural conversations without common AI interaction issues. The potential applications of EVI are vast and exciting, envisioning AI assistants, customer support agents, and even virtual therapists capable of providing genuine emotional support.

Limitations

However, detecting nuanced emotions through voice and text remains challenging due to subjectivity influenced by various factors. Hume acknowledges its limitations, as demonstrated by inaccuracies when tested on casual inquiries like lunch preferences.