September 2024 Summaries

Emotion science is a rapidly advancing field that explores how we experience and express emotions. It sits at the intersection of neuroscience and psychology and aims to understand our emotional lives, reactions to different situations, and communication of internal states. The history of emotion science can be traced back to foundational research on facial expressions conducted in the mid-20th century, with roots also going back to philosophers like William James and Charles Darwin. Emotion science is critical for AI as it enables machines to understand and learn from human emotional behaviors, leading to more meaningful, empathic interactions. By leveraging insights from emotion science, we are moving towards a future where AI can interpret the subtleties of our emotional lives and respond appropriately, enhancing our well-being and enriching our human experience.

Emotion AI, also known as Affective Computing or Empathic AI, is an area of research focusing on machine learning models capable of identifying and categorizing emotion-related behaviors. While AI can capture patterns in human expressions, it's crucial to recognize that these technologies interpret behaviors in probabilistic terms, not certainties. Real applications of these technologies focus on observable data and shared human interpretations of visible emotional expressions. Empathic AI is a more accurate term for this field, as it uses an understanding of how we express ourselves to produce better responses. The science behind human expressions has been studied extensively by emotion scientists over the past 50+ years, focusing on facial and vocal expressions. These models offer objective measurements of expressive behaviors that people readily associate with various feelings when they see or hear them in isolation. Empathic AI has the potential to enhance how we communicate with technology, improve how technology affects our well-being, and enable more meaningful connections among humans while respecting privacy.

Voice-to-voice foundation models are the latest major breakthrough in AI, enabling users to speak with AI through voice alone. The world's first working voice-to-voice models are Hume AI's Empathic Voice Interface 2 (EVI 2) and OpenAI's GPT-4o Advanced Voice Mode (GPT-4o-voice). These systems have many capabilities in common, such as processing audio and language, outputting voice and language, and understanding a user's tone of voice. However, EVI 2 is optimized for emotional intelligence, maintaining compelling personalities, customization, and designed for developers, while GPT-4o-voice supports more languages. Voice-to-voice models are set to transform various sectors like customer service, mental health, education, and personal development by providing a more efficient interface for virtually any application.

EVI 2 is a new voice-to-voice foundation model that enables remarkably human-like conversations with users. It can converse rapidly, understand tone of voice, generate any tone of voice, and adapt to user preferences. Currently available in beta, the model is designed for building into applications via API and is incapable of cloning voices without modifications to its code. Developers can adjust EVI 2's base voices along continuous scales, allowing them to create tailored voices for specific apps and users. The upcoming improvements include increased reliability, learning more languages, following complex instructions, and using a wider range of tools.