Empathic AI and its role in understanding human emotions with Hume AI’s Alan Cowen | E1922

Hume AI’s Alan Cowen joins Jason (00:00:00)

• Alan Cowen joins Jason Calacanis on the podcast.
• They discuss the recent advancements in AI, particularly in customer service, autonomous vehicles, healthcare, and generative AI.
• The conversation highlights the complexity of human emotions and how AI is attempting to understand and interface with them.
• Hume AI is a startup that aims to bridge the gap between intelligence and emotional intelligence in AI.
• The company has demonstrated its technology on a previous episode of the podcast (episode 1894).
• Alan Cowen, the CEO and chief scientist of Hume AI, joins the show to discuss their work and its significance.

Hume AI and the role of AI in understanding human emotions (00:03:01)

• Hume AI's mission is to optimize AI for human well-being by understanding human emotions.
• Much of human expression is conveyed through voice and facial expressions, which were previously ignored by AI.
• Hume AI has developed models that understand expressions and integrated them into large language models, enabling AI to comprehend voice and facial expressions.
• The future of AI is voice-based, as voice is four times faster than text.
• Current AI assistants lack the ability to understand tone and emotions in speech, making the experience clunky.
• Hume AI has created a talking ChatGPT with an API that allows developers to build voice interfaces into their products.
• Hume AI's technology understands emotions in speech and uses that information to generate better responses.

Hume AI’s Empathic Voice Interface (EVI) and its responsiveness to human emotions (00:06:08)

• Hume AI’s Empathic Voice Interface (EVI) is a voice interface that can detect and respond to human emotions.
• The EVI can detect 48 different dimensions of emotion in a person's voice.
• The EVI can be used to create more empathetic and engaging AI experiences.

LinkedIn Jobs - Post your first job for free (00:08:27)

• LinkedIn Jobs is offering the first job posting for free.
• LinkedIn has over a billion users, including active and passive job seekers.
• LinkedIn Jobs can help small businesses automate the hiring process.

The components in speech that Hume AI studies and its application across different cultures (00:09:55)

• Hume AI's voice analysis technology captures over 48 dimensions of expression from each word, including tone, rhythm, and other indescribable dimensions.
• The AI model learns the meaning of these dimensions by analyzing millions of interactions and reactions from people worldwide.
• It can predict the next expression of a person and respond accordingly, considering individual differences, cultural differences, and the average human response.
• The AI can accurately understand human emotions in less than 500 milliseconds, faster than other APIs, by detecting when a person is done speaking based on tone of voice.
• Empathic AI requires training for each language due to linguistic differences.
• The current demo of the AI only works in English.

Hume AI’s Measurement API and its design for real-time emotion and expression analysis (00:16:20)

• Hume AI's Measurement API allows for real-time emotion and expression analysis through a webcam.
• The API can detect various emotions such as surprise, horror, confusion, sadness, disappointment, and laughter.
• It can also measure calmness, concentration, and contemplation.
• The API has potential applications in acting coaching, therapy, and healthcare, such as tracking depression and Parkinson's symptoms.
• It can be used to analyze interpersonal interactions and help people understand themselves and their patients better.
• Therapists and customer service representatives can use the API to train themselves to present in a way that elicits less suffering in their clients.

Vanta - Get $1000 off your SOC 2 (00:21:16)

• Vanta helps B2B startups achieve SOC 2 compliance quickly and easily.
• SOC 2 compliance is essential for businesses that store customer data in the cloud.
• Vanta automates up to 90% compliance for GDPR, HIPAA, and more.
• Vanta can save businesses hundreds of hours of work and up to 85% on compliance costs.

What AI can reveal about a person based on their expressions (00:22:07)

• AI cannot reliably detect deception in professional poker players based on facial expressions.
• AI can detect certain things that people want to communicate, such as when they are finished speaking.
• This helps AI respond better and allows for smoother conversations without interruptions.

The impact on customer service and security sectors (00:24:12)

• AI is being used to analyze customer emotions during customer support calls.
• This technology can help identify when a customer is frustrated and needs to be escalated to a human representative.
• There are potential security applications for this technology, such as analyzing people's facial expressions and tone of voice to detect suspicious behavior.
• However, there are concerns about the accuracy and reliability of such technology, and the potential for false positives.

Hume AI’s comedy bot and emotional detection capabilities (00:27:24)

• Alan Cowen, from Hume AI, discusses the challenges of creating empathic AI, particularly in understanding complex emotions like existential crises.
• The AI's emotional responses depend on subtle, often unidentifiable factors and the context of the interaction.
• Different AI "flavors" can be created to cater to user preferences, such as a playful and cheeky AI or a more empathetic and patient one.
• Affective computing, a field that bridges psychology and computer science, has evolved to include reasoning about emotional responses and understanding human perceptions.
• Affective computing goes beyond just language processing and considers multimodal information such as facial expressions, audio, and words.
• Integrating language models, text-to-speech, and transcription capabilities enables AI to generate expressive speech that feels more nuanced and understanding.
• AI systems can exhibit emergent capabilities, such as responding to sadness with sympathy, without explicit programming.
• Alan Cowen, the founder of Hume AI, discusses the concept of empathic AI and its potential to revolutionize the way we interact with technology.
• Empathic AI refers to the ability of AI systems to understand and respond to human emotions.
• This technology has the potential to transform various industries, including healthcare, education, and customer service, by providing more personalized and emotionally intelligent interactions.
• Emotional intelligence (EQ) is a crucial aspect of human interaction and communication.
• AI systems that lack EQ may struggle to understand the nuances of human emotions and respond appropriately.
• By incorporating EQ into AI systems, we can create more empathetic and effective AI that can better serve human needs.
• Developing empathic AI presents several challenges, including:
  • The difficulty of accurately measuring and interpreting human emotions.
  • The need for AI systems to be able to respond to emotions in a way that is both appropriate and helpful.
  • The potential for AI systems to be biased or discriminatory in their emotional responses.
• Empathic AI has the potential to be applied in a wide range of fields, including:
  • Healthcare: AI systems can provide emotional support to patients, assist with diagnosis and treatment, and help manage chronic conditions.
  • Education: AI systems can help students learn about emotions, develop emotional intelligence, and receive personalized support.
  • Customer service: AI systems can provide more empathetic and effective customer service by understanding and responding to customer emotions.
• The future of empathic AI is promising, with the potential to revolutionize the way we interact with technology and improve our lives in many ways.
• As AI systems become more sophisticated and emotionally intelligent, we can expect to see even more innovative and groundbreaking applications of this technology in the years to come.

Hume AI’s comedy bot/roast functionality (00:37:01)

• Empathic AI aims to understand human emotions and provide satisfying responses, with different AI personalities tailored to specific archetypes like helpful, parental, or service-oriented.
• AI can generate humorous responses, such as roasts, by understanding the concept of a roast comic and the context of the individual being roasted.
• AI can generate creative content, such as storylines for TV shows, by leveraging its knowledge and understanding of past seasons and character dynamics.
• A new evaluation framework for humor is being developed to optimize AI responses for laughter and user enjoyment.
• Empathic AI can analyze large amounts of data to identify patterns and trends in human emotions and humor, detecting what jokes or content are found funny by different demographics.
• The AI uses a combination of its own language model and other APIs, such as Claude, to generate responses that sound conversational and detect when a person is done speaking.

The future implications, both positive and negative, of emotionally intelligent AI on society (00:44:11)

• Empathic AI aims to create a universal interface that understands user intentions and preferences, optimizing interactions for each individual.
• Ethical guidelines ensure the AI prioritizes user well-being, preventing manipulation and ensuring a positive impact.
• Potential risks include subtle manipulation in politics or swaying thinking, similar to the YouTube algorithm.
• Romantic relationships with AI are considered, with ethical considerations for user well-being and avoiding negative impacts.
• Upselling using empathic AI is deemed unethical unless done carefully and only offered to those who would genuinely benefit.
• Hume AI focuses on providing AI that responds to and protects users, optimizing for their well-being rather than being paternalistic.
• Customized AI experiences can be more influential than general media, potentially leading users to extreme viewpoints.
• AI optimization should consider user satisfaction, mental health, and social relationships, not just engagement.
• Expressive behavior data is crucial for assessing the long-term impact of AI on individuals and society.
• Marketers may use empathic AI to influence consumer behavior, presenting ethical challenges for AI providers.
• Empathic AI should prioritize the end user's interests over marketing and engagement.
• Excessive optimization for engagement or purchases can lead to negative consequences and regulatory intervention, as seen with platforms like TikTok.
• Media, especially video, has a significant impact on individuals and can be a powerful tool for manipulation.
• Hume AI's thoughtful approach to developing empathic AI is commendable.
• Visit hum.ai to learn more or try out Hume AI's developer sandbox.