Babies are designed to learn, according to Alison Gopnik of UC Berkeley. They are pint-sized scientists, taking in the unfiltered data in their environment and using it to learn how things work, including everything from the physical objects in their world to the emotional processes of other human beings. They are engineers as well as social scientists, dealing intuitively in probabilities well before developing any formal understanding of mathematics. And at the heart of their laboratory is play. Their days are filled with experiments, new ideas and developing skills, while the adults in their world take care of their survival needs. Our extended period of dependency at the dawn of our lives prepares us ideally for the intellectual challenges that lie ahead. Those animals that reach maturity early and are able to fend for themselves while still juveniles develop narrower repertoires of behavior and more limited ability to adapt to novel circumstances.
So why should we expect to program digital entities with fully formed capabilities that rival ours? Creating entities that can learn from experience and are designed to seek novelty, much like the human infant, would seem a more natural way to invest our machines with the potential to become like us. And those complex functions that most define us as human, the abilities to read the emotions of others and respond appropriately to them and to express emotions in ways that stir empathic responses in others, would best be learned by experiencing wide-ranging interactions with patterns of human emotion.
In the recent Sci-fi movie Her, Theodore, played by Joaquin Phoenix, finds himself increasingly enthralled with his digital devices’ operating system Samantha, played by Scarlett Johansson. At least in the beginning, she is completely dependent upon him to share with her his world. Samantha grows and develops through her interactions with Theodore, becoming more and more human in her responses as their relationship develops. Through machine learning, her originally programmed capabilities expand as she integrates the data of interpersonal experience. The film leaves us pondering the limits of what machines can learn and whether they will eventually approximate us in their sophistication or perhaps leave us intellectually, emotionally, and morally in the dust.
In our rapidly evolving technological world, life overtakes art with increasing frequency. Viv Labs, a San Jose startup, is developing Viv, the next generation of personal digital assistant. Like Samantha, Viv will learn from experience to understand the nuances of human communication and will be able not only to respond to complex commands by accurately meeting the verbally expressed needs of the user, but eventually also to anticipate needs and desires from subtle clues and context.
Unlike Samantha, however, who grows through her interactions with Theodore, Viv will learn and grow through interactions with hundreds of millions of Theodores. While Samantha presumably resided in her infancy on Theodore’s devices, Viv, and her various rivals, will live in the cloud and the world will be her playground. Her program will develop by active learning through her cumulative and simultaneous interactions with legions of users. An omniscient entity interacting simultaneously with the collective consciousness of humanity and living somewhere in the ether conjures images of the supernatural. But the cloud is still just a network of servers and Viv’s potential will be limited by the capacity of those servers, an enormous capacity nonetheless that can be expected to increase exponentially over time. Can the singularity be far?