OpenAI recently released an the long awaiting new Advanced Voice model.
This new feature offers a much more fluent and natural way to interact with your phone by voice. The underlying technology is so powerful that OpenAI had to impose limits to prevent the AI from singing (it can sing beautifully), imitating voices or sounds, or being overly emotional. You can, for example, ask it to tell you a dramatic story, but it won’t laugh or attempt to imitate humans too closely.
It believe the industry is still underestimating the impact this new mode of interaction will have on the digital world. Over the past few days, I’ve experimented with it extensively and found it to be fantastic for brainstorming, as a language coach, and as a hands-free assistant while performing other tasks (such as cooking).
The proper implementation of this technology could revolutionize how we think about “digital interfaces.” More importantly, it will unlock new use cases where conversations with AIs become integral to personal productivity, learning, entertainment, and even emotional support.
Talking to an AI “feels” very different from chatting with one, especially when it can deeply understand the emotional nuances in our conversations, drawn from context, previous interactions, or (in future versions) from the tone of our voice. The AI listens with infinite patience and provides valuable support.
We have undoubtedly crossed the threshold where these models can interact with us in the most natural ways. They can role-play, simulating being best friends, lovers, or even monsters from a faraway galaxy. This new technology is about to trigger a wave of transformation.
On a side note, these models can already exhibit significant emotional intelligence. They can simulate emotions (though regulations and alignment practices often prevent them from doing so). The ongoing debate is whether they are “experiencing” emotions in the real sense or merely “simulating” them. This is not a trivial question, as many argue that emotions are core to cognitive processes. After a certain threshold, the distinction between “simulated” and “real” emotions may lose its scientific relevance.
This topic (the reality of thensubjective experience of self and of the emotions) has been largerly explored by authors like Antonio Damasio, Daniel Dennet, David Chalmers and Rosalind Picard. I will discuss this topic more deeply in our of the next blog posts and this will also be a core component in my upcoming book “Human Digital Minds”.
In the meantime: let’s experiment and enjoy these new toys and (for those of us more deeply involved in the development and application of these new technologies) let’s think through the opportunities and profound implications for the industry and society as a whole.