Rosalind Picard on Affective Computing, AI Ethics, Emotion, Privacy, and Human-Machine Connection

This podcast features Rosalind Picard, a pioneer in affective computing, discussing the evolution of the field she launched over two decades ago. Initially conceived as any computing related to, arising from, or influencing human emotion, the core focus has largely been on machines recognizing and intelligently responding to human emotional states. Picard highlights early examples like Microsoft's Clippy, which failed due to emotional unintelligence, contrasting it with a vision where AI could genuinely adapt to user frustration or delight. She emphasizes that social-emotional interaction is far more complex than traditional AI problems like chess or Go, requiring a broader range of skills and perspectives within computer science. A significant portion of the conversation delves into the ethical implications and societal concerns surrounding advanced affective computing. Picard expresses deep worry about the misuse of emotion recognition technology, particularly in authoritarian regimes like China, where it could be used for surveillance and control, penalizing even subtle expressions of dissent. Her company, Affectiva, actively declines opportunities that involve reading emotions without prior informed consent, advocating for strong safeguards and user buy-in. She also discusses the concept of adversarial examples and deepfakes, noting how technology can both enable and potentially counteract such surveillance, highlighting the ethical choices researchers face in how they allocate their time and resources. Picard argues for the need to put "brakes on" certain AI developments, shifting the focus from simply building AI for profit or academic publication to addressing societal inequities. She advocates for regulations that protect data ownership and extend lie detection laws to emotion recognition, especially concerning mental health predictions. She explains how non-medical data from smartphones and wearables, combined with machine learning, can predict future emotional and health states, underscoring the urgency of these privacy concerns. Despite these worries, she acknowledges the potential for AI, like personal assistants, to alleviate loneliness and foster deep connections, drawing parallels to engaging with characters in books. Ultimately, Picard stresses that AI is a human creation, and its design reflects human values. She challenges the notion of AI as an equal, emphasizing that humans control its capabilities and objectives. She warns against AI designed to manipulate or enforce submission, as depicted in dystopian visions like "Brave New World." Instead, she champions building AI that extends human intelligence, empowers the weak, and balances power dynamics, rather than exacerbating existing inequalities. The discussion also touches on the limitations of emotion recognition, noting that while physiological states (like heart rate from facial video) can be inferred, nuanced thoughts and feelings remain private.

actually originally when I defined the term effective computing it was a bit broader than just recognizing and responding intelligently to human emotion although those are probably the two pieces that we've worked on the hardest the original concept also encompassed machines that would have mechanisms that functioned like human emotion does inside them it would be any computing that relates to arises from or deliberately influences human emotion

social emotional interaction is much more complex than chess or go or any of the games I that people are trying to solve and in order to understand that required skills that most people in computer science actually we're lacking personally

I actually think there's a more important issue right now then the difficulty of it and that's causing some of us to put the brakes on a little bit usually we're all just like step on the gas let's go faster this is causing us to pull back and put the brakes on and that's the way that some of this technology is being used in places like China right now and that worries me so deeply

here in the US and my first company we started a Factiva we have worked super hard to turn away money and opportunities that try to read people's effect without their prior informed consent

I would like to see people own their own data I would also like to see the regulations that we have right now around lie detection being extended to emotion recognition in general

what most people don't know yet is right now with your smartphone use and if you're wearing a sensor and you want to learn about your stress and your sleep and your physical activity and how much you're using your phone and your social interaction all of that non-medical data when we put it together with machine learning now call to AI even though the founders of a I wouldn't have called it that that capability cannot only tell that you're calm right now or that you're getting a little stressed but it can also predict how you're likely to be tomorrow

we can read with the ordinary camera on your laptop or on your phone we can read from a neutral face if your heart is racing we can read from a neutral face if your breathing is becoming irregular and showing signs of stress

your thoughts are still private you're nuanced feelings are still completely private we can't read any of that

if you believe that the human race is better off being given freedom and the opportunity to do things that might surprise you then you want to use AI to extend people's ability to build you want to build AI that extends human intelligence that empowers the weak and helps balance the power between the weak and the strong not that gives more power to the strong