UC Berkeley Psychology Professor Alison Gopnik discussed what children’s development can teach us about artificial intelligence and what AI can teach us about being human during this year’s Berkeley Social Sciences Distinguished Faculty Lecture.
Gopnik challenged the commonly-accepted notion of a “general intelligence,” instead proposing that intelligence arises from many distinct capacities. Rather than viewing intelligence as a quantitative, measurable value, she outlined three interdependent kinds of intelligences: exploration, exploitation and empowerment.
“Intelligence is a folk concept – it’s a mysterious essence that people either have more or less of. Supposedly, people who have more should be rewarded,” said Gopnik. “In reality, humans are made up of a variety of diverse intelligences. Certain factors that are portrayed as deficits are actually strengths.”
According to Gopnik, these intelligences reflect the human life cycle. She explained that exploration drives children to learn through experimentation, while exploitation allows adults to pursue goals and empowerment enables elders to share wisdom across generations.
“In contrast to AI, intelligence takes different forms across the lifespan,” said Gopnik. “These complementary kinds of intelligence are what has driven human success.”
Gopnik also presented caregiving as a unique form of intelligence. While the importance of caregiving has often been neglected, she noted, it’s actually a foundational skill that helps others develop the autonomy children need to pursue their own goals.
Children’s ability to learn, innovate and grow is dependent on receiving proper care, Gopnik explained. We see this modeled in the relationship between parents and children, but according to Gopnik, this is also applicable to relationships between doctors and patients, teachers and students or elders and the youth.
“Children are rewarded for the very fact that they’re learning to act on their environment,” said Gopnik. “However, if exploration is dependent on having adults around, or other people who are willing to put care in, what does this mean for AI?”
Linking these ideas to artificial intelligence, Gopnik argued that the way AI develops is also dependent on how children learn. She suggested that AI shouldn’t be trained solely through external reinforcement, but that it should also be designed to experience intrinsic rewards that encourage exploration.
If AI becomes more autonomous, it will need relationships of guidance and care, argued Gopnik. Just as caregiving nurtures human intelligence, it may also be essential for creating AI systems that align with and support human values.
“AI systems are just another method we’ve invented for passing on information, similar to elders. This could have great or terrible consequences,” said Gopnik. “Just like with children, we need to make sure they’re aligned with our interests. Caregiving is the solution to this.”