AI.pdf

Philosophy of Artificial Intelligence Learning objectives: In this lecture you will: Get introduced to some of the philosophical questions in AI Try to answer the question “Can a machine acquire a conscious mind?” Learn about Intentionality and Qualia Can a machine think / understand / ac...

Philosophy of Artificial Intelligence Learning objectives: In this lecture you will: Get introduced to some of the philosophical questions in AI Try to answer the question “Can a machine acquire a conscious mind?” Learn about Intentionality and Qualia Can a machine think / understand / acquire Intelligence / be conscious? in short, can it possess mental states?  WHAT BELONGS TO ARTIFICIAL INTELLIGENCE? Automated Reasoning, Theorem Proving and Game Playing Expert and Knowledge-Based Systems Natural Language Processing and Semantic Modelling Robotics Machine Learning Evolutionary Techniques Neural Networks Genetic Algorithms ….and more  COMPUTATION AND TURING MACHINE Computation is computation of Turing Machine functions The simplest tool to compute INTELLIGENCE; THE TURING TEST (1950) A behavioral approach “The imitation game” If A is mistaken for a human, it has passed the test. Turing’s prediction: A machine would pass the test in about 50 years. It hasn’t so far!! PROBLEMS TO AI: 1) INTENTIONALITY the power of minds to be about, to represent, or to stand for, things, properties and states of affairs A Mind can have an idea about a chair, but a chair is not about anything. It just exists. Your mind has ideas about the words on this slide, but the words just exist. Minds have the ability to understand meanings; or rather they create meanings about the world, but the world itself just exists. “Every sign by itself seems dead.What gives it life?” - Ludwig Wittgenstein INTELLIGENCE; INTENTIONALITY John Searle: Even if a machine passed the Turing test we cannot assume that it acquires a mind like humans. There’s nothing inside. It does not understand meanings.(semantics) Machines acting intelligently -> Weak AI Machines acquiring an understanding mind -> Strong AI All technology developed so far is considered instances of “Weak AI”. INTENTIONALITY The Chinese room thought experiment (1980) Correct manipulation of symbols does not mean that the man understands Chinese! A symbol represents nothing to a machine. It deals with the syntactics, but does not get the semantics/meanings Potential reply: The system reply (The whole system composed of the room, man, dictionary, …etc.., understands Chinese) (However, memorizing the dictionary then going out of the room would still leave the man incapable of understanding Chinese while still giving correct answers) 2) QUALIA ineffable subjective qualitative properties of experience The painess of pain, the redness of red "an unfamiliar term for something that could not be more familiar to each of us: the ways things seem to us" - Daniel Dennett QUALIA QUALIA Mary’s room thought experiment (Frank Jackson 1982) Does she learn anything new? The knowledge argument: There are non-physical, informal (ineffable; cannot be put in words) components of knowledge that can only be discovered through lived conscious experience. QUALIA A machine acts as if it is in pain, but the pain is not really experienced! Such creature being conceivable stands against Physicalism QUALIA All that Science can understand are the easy problems , like recalling memories, being awake, integrating perceptions, finding neural correlates…etc.. We can have all the information needed about the brain & its processes, and we can deduce the mental state of a person from such information, but we can never deduce how this state feels like. E.g: We can predict that a person is in pain David Chalmers from the firing of the C fibers and activation of certain brain areas, but we can never deduce how this pain feels like. Explaining Qualia is the hard problem of consciousness 3) Lady Lovelace objection to AI: Humans are autonomous; Machines are automata In other words, can a machine acquire Free Will? What we learned The Turing machine and the Turing test A behaviorist approach (weak AI) would consider AI possible A cognitive approach (strong AI) would on the other hand consider the path to AI full of obstacles Intentionality, Qualia & Free Will represent such possible obstacles