A correct definition of artificial intelligence

Artificial intelligence (AI) is intelligence in non-biological form. That seems easy enough, but intelligence itself is what is hard to define. Intelligence includes the ability to plan, problem solve, and apply pattern recognition. This form of intelligence is shared by many lifeforms on this planet. Even the common fruit fly.

We only see higher forms of intelligence in the most intelligent animals - us. In this case intelligence is the ability to to understand and use complex natural languages, to rationalize, to create complex works of art or invention, to be self-aware, to have metacognition and a theory of mind. Intelligence covers a wide aspect of concerns. AI today only can address some of these concerns.

So we can to some degree define intelligence. But, we can not easily test for it based on these definitions. This is because these descriptions define qualities. Qualities can not be easily quantified.

Alan Turing anticipated this in his foundational AI paper. He argued that testing for intelligence based on a definition is a fruitless endeavor. Instead, he proposed a test he called the Imitation Game (which we now call the Turing Test). To play the game you need a judge and two players. One of these players is a machine, the other a human. The judge (who is also human) doesn't know which player is a human and which is a machine. The judge is able to ask the players free form questions of any type. Through this question and answer format, the judge must be able to tell the players apart. If this game is repeated many times with different human players and judges, and yet the machine can not be told apart from the human, we can say that the machine passed the Turing Test, in other words, it is truly intelligent.