MA/CS 109 Computer Science Lectures. Wayne Snyder Computer Science Department Boston University

Transcription

1 MA/CS 109 Lectures Wayne Snyder Department Boston University Today Artiificial Intelligence: Pro and Con Friday 12/9 AI Pro and Con continued The future of AI

2 Artificial Intelligence

3 Artificial Intelligence

4 Artificial Intelligence

5 Prospects for Artificial Intelligence: Pro The boldest statement of the prospects for true Artificial Intelligence is the STRONG AI THESIS: Strong AI (artificial intelligence) thesis: the Mind is assumed, or postulated, to be a consistent algorithm, and therefore if properly programmed, a digital computer can, in principle, mimick the mind, provided the basic assumptionabout the Mind is correct. A proper computer program can be intelligent, understand, perceive, have beliefs, learn, adapt, and exhibit other cognitive states normally ascribed to human mind. Thus: Brain = Hardware Mind = Software Informally: A computer can be an alive, intelligent, conscious being in the same sense that a human is.

6 Prospects for Artificial Intelligence: Pro What evidence is there for the Strong AI Thesis? Most arguments depends on extrapolating from our current situation: Computers can do many things that have in the past been attributed to life and intelligence: Properties of Life: 1. Organization as cells 2. Hierarchical organization 3. Use energy 4. Respond to environment? 5. Growth? 6. Reproduction? 7. Adapt to environment through evolution? Properties of Intelligence: 1. Reason, use strategy, solve puzzles, and make judgments under uncertainty; 2. Represent knowledge, including commonsense knowledge; 3. Plan and learn; 4. Communicate in natural language;? 5. And integrate all these skills towards common goals.? [NOTE: this list does not mention consciousness!]

7 Could Computers Ever Be Alive? What progress could be made towards the rest of these goals in say, 1,000,000 years? If it does not seem reasonable to you that a machine could be truly intelligent and even conscious, consider that in biological systems (such as your brain), intelligence is an Emergent Property: An Emergent Property is a property which a collection or complex system has, but which the individual members do not have.

8 Prospects for Artificial Intelligence: Pro There are many biological examples of emergent behavior: Ant Bridge Slime Mold Termite Mound

9 Could Computers Ever Be Alive? An Emergent Property is a property which a collection or complex system has, but which the individual members do not have. Example 1: One Neuron is not intelligent, but your brain (100 billion neurons) is! Example 2: Boids follow simple rules and exhibit very lifelike flocking behavior. As an example of how Emergent Properties work, let us look at an attempt to create artificial life

10 The Game of Life History: Mathematician John Conway wanted to answer an open problem in mathematics: can we define a machine that can make a copy of itself? He unwittingly created the concept of Artificial Life or ALife, in which researchers try to understand growth, evolution, and reproduction using computer models. Conway s Game of Life: The board is a 2D grid of squares infinite in four directions; Cells are alive or dead, and are each touched by 8 neighbors:

11 Could Computers Ever Be Alive? Rules of the Game of Life:: A cell which is alive will die if it has v Less than 2 neighbors (as if of loneliness); v More than 3 neighbors (as if of overcrowding); v Otherwise (if it has 2-3 neighbors) it remains alive. A cell which is dead will come to life if it has v Exactly 3 neighbors; v Otherwise it remains dead. Let s take a look..

12 Final Lecture Strong AI Thesis: A properly programmed digital computer could have a mind in the same sense that a human has a mind. This is not (yet) something we can test scientifically, but we can discuss the current thinking pro and con. PRO: Perhaps intelligence is an Emergent Property, and so current computers are not intelligent because they are not complex enough. If a single neuron is not intelligent, but 100 trillion neurons can be, perhaps the same can be said of computer circuitry. And the complexity of computers is growing exponentially...

13 Final Lecture

14 Final Lecture What arguments have been raised against the Strong AI Thesis? The most famous such argument is John Searle s Chinese Room Thought Experiment:

15 Final Lecture This has sparked a lively debate since it was introduced in 1980: The Systems Reply: The man in the room doesn t understand Chinese, but the room as a whole does. [ Basically the Emergent Properties argument. ] Searle s Answer to the Systems Reply: Simulation is not the same as real intentionality. The simulation of a rainstorm does not leave you wet. The Robot Reply: The room is not connected to the world by senses, there is no causal relationship to its environment. Put the Chinese Room inside a robot and the argument is unchanged. Searle s Response: Well, then, I ll memorize the rule book and follow it when I answer the Chinese questions I still don t understand. The Non-Classical Computer Reply: The problem is with the design; perhaps electrical neural networks or quantum computers would be different. Searle: Nope! These can be simulated by digital computers, so we re back where we started. Where will we end up?...