Answer: an advance in intuitive problem solving
Researchers recently announced that an AI system they had developed to win against humans at the game of Go — called AlphaGo — is advancing faster than they had anticipated. The system, which trained by playing millions of games against itself, beat the European Go champion Fan Hui in five out of five games, a feat they had not believed possible for another decade.
Go is difficult to for a computer to master because the game's simple rules and large board make for more configurations than there are atoms in the universe. Analyzing an enormous database of possible plays and outcomes is a less favorable technique than the human-style approach of using intuition, which is a large component of how humans succeed at the game, and how the system is able to consistently win.
"The search process itself is not based on brute force; it's based on something more akin to imagination," researcher David Silver told Phys.org.
Areas where AI systems are still classified as subhuman are those most closely associated with what's typically called "thinking," a process in humans that was tuned via the sensory apparatus through millions of years of biological evolution. They include areas like character recognition, handwriting recognition, object recognition, speech recognition, translation, word-sense disambiguation and natural-language processing.