Artificial Intelligence Testbed
What does it mean to “understand” the origins of intelligence? In our view, understanding comes with the ability to build a system that mimics the system of interest. So if we truly understand the origins of intelligence, then we should be able to build machines that learn and think like newborn animals.
To this end, recent advances in artificial intelligence provide the opportunity to begin building “pixels-to-actions” models of newborn animals. These autonomous artificial agents can learn to interact successfully with their environment, receiving only rewards and pixels as input (akin to the rewards and retinal images received by newborn animals). Thus, these agents are formalized learning mechanisms in action. Each agent can be thought of as a quantitative model for predicting what a newborn animal will learn given a specific set of experiences. If an artificial agent learns like a newborn animal, then the agent and animal should develop the same abilities when reared in the same environment.