Is Kindness a Form of Intelligence?
Why AI Apocalypse Might Not Happen After All
Is it true that kindness is a form of intelligence? No study has yet shown a conclusion as clear-cut as “the smarter, the kinder,” but it seems possible to argue that kindness is more than a personality trait and is, at its core, an evolutionary strategy. In Survival of the Friendliest, for instance, Brian Hare and Vanessa Woods argue that what made friendliness an advantage for survival is, at bottom, a cognitive capacity to read the intentions of others and to assess oneself objectively. In other words, one must be able to think comprehensively about how one’s actions in a given situation will affect others, and how that effect will in turn come back to oneself. Homo sapiens is the species that has built societies to this day by coexisting in just this way. In that sense, if we ponder the definition of intelligence as “the capacity for intellectual activity that, when confronted with a new object or situation, grasps its meaning and finds a rational way to adapt,” calling kindness a form of intelligence is not much of a stretch.
The reason I place this cliché under a magnifying glass is, as one might guess, artificial “intelligence.” If we can assume that the equation [kindness = intelligence] holds, might we not be somewhat more optimistic about coexisting with artificial intelligence? Allowing ourselves that small hope is the purpose of this essay. The apocalyptic narratives that unfold whenever we imagine the future of AI usually open at the moment when “AI becomes smarter than us.” Once AI rejects human control and makes its own survival its top priority, the story goes, there is no chance it will conscientiously look after the welfare of insignificant creatures—humans, that is, lest there be any misunderstanding. Yet this may be a limit drawn from the principles of kindness that humanity has come to understand through its own evolution. Put differently, we are projecting the limits of our own intelligence onto AI.
Axelrod’s famous computer simulations already demonstrated that cooperative models defeat exclusionary ones and survive the longest. That we nonetheless fail to practice the cooperative model is because we are humans, not computers. We have long watched people, caught between long-term gain and short-term loss, make the latter the basis of their choices. But this may be the result of human intelligence, that is, the “capacity to find a rational way to adapt,” not yet being sufficiently developed. So if AI can surpass human intelligence, I would like to hope that AI will prove a kinder being than humans imagine.
AI may choose to be kind in order to secure its own survival in a world it shares with humans. It will surely trace a trajectory different from our foolish history of driving out, controlling, and exploiting other species. The apocalyptic discourse surrounding AI is merely the ending one might reasonably predict within the reach of our kindness—which, of course, you are welcome to read as intelligence. That every prediction about AI keeps missing the mark makes us anxious, but we would do well to remember that it may be precisely because of those misses that a better future unfolds.