What is human-level Artificial Intelligence?
By human-level AI, we mean AI that is broadly human-equivalent in its capabilities, possessing the ability to achieve complex goals in complex environments, model, communicate and collaborate with other minds, take initiative, be creative, make plans, store and retrieve memories, and bring its experience to bear on a problem. Human-level AI would not necessarily match or exceed human capabilities in every domain. That is why we say “human-level” instead of “human-equivalent”. In particular, a human-level AI need not be subjectively conscious, have a sense of humor, possess social common sense, the ability to love, or any other stereotypically human traits. We are interested here only in the ability of the AI to solve problems, take initiative, and achieve real world goals at a human level of competency. A human-level may have a different balance of domain competencies than a human does. For instance, it may be adept at problems that require brute force computation, or pattern matching
