ArtificialIntelligence By Caitlin Smith • “Artificial intelligence in one form or another is an idea that has pervaded Western intellectual history, a dream in urgent need of being realized.“ - Pamela McCorduck
Content • History • What is Artificial Intelligence? • Goals • Requirements • The Philosophy of AI • Problems • Deep Blue • Questions?
History • In the middle of the 20th century, a handful of scientists began a new approach to building intelligent machines, based on recent discoveries about the brain, new mathematical theories, cybernetics, and the invention of the digital computer. • The field of modern AI research was founded at conference at Dartmouth College in the summer of 1956. • Many research leaders attened, including John McCarthy, Marvin Minsky, Allen Newell and Herbert Simon. • They began writing programs that allowed computers to solve algebraic word problems, prove logical theorems and speak English In fact, things were going so well that in 1965, Herbert Simmon said: “Machines will be capable, within twenty years, of doing any work a man can do”
History • Sadly, AI researchers failed to recognize the difficulty of some of the problems they faced. In 1974, their main sponsor (DARPA) cut off all exploratory funds due to pressure from Congress and taunting from England. • This is commonly called the first ‘AI Winter’ • In the early 80s, the field was revived, and by 1985 the market for AI had reached more than a billion dollars. Minsky and others warned the community that enthusiasm for AI had spiraled out of control and that disappointment was sure to follow, which it did in 1987, starting the second AI Winter. • In the 90s and early 21st century AI achieved its greatest successes, although it’s more behind the scenes now. Artificial intelligence was adopted throughout the technology industry. The success was due to several factors: • The power of modern supercomputers today • The creation of new ties between AI and other fields working on similar problems • A new commitment by researchers to solid mathematical methods and rigorous scientific standards
What exactly is Artificial Intelligence? • The modern definition of artificial intelligence (or AI) is the study and design of artificial intelligent agents, where an ‘intelligent agent’ is a system that has the ability to perceive and evaluate its environment and take actions which maximize its chances of success. • The term was coined by John McCarthy in 1956, who defines artificial intelligence as "the science and engineering of making intelligent machines.“ • The term is also used to describe a property of machines or programs: the intelligence that the system demonstrates. • Other names for this field have been propsed, such as: • Computational Intelligence • Synthetic Intelligence • Computational rationality
Goals • Among the traits that researchers hope machines will eventually exhibit are: • Reasoning • Knowledge • Planning • Learning • Communication • Perception • General Intelligence (or, Strong AI) is the ultimate long-term goal of researchers, although it has yet to be achieved.
Requirements • For a potential AI unit to be considered “alive” or “intelligent” it would have to exceed its original programming. • This statement is a reflection of the sentiment that “the whole is greater than the sum of the parts”, and of humankind’s ability to overcome instinct (our initial ‘programming’) • The AI questioning its original programming without provocation. • AI was programmed to go “beep” every minute and then left alone to do so, eventually it would wonder if it was necessary to beep every minute. • The AI being able to solve problems it was not originally programmed to solve. • This requirement shows an ability to apply deductive reasoning without connections that have been specifically laid beforehand. It further requires the AI to draw upon all of its “knowledge” and “skills” • This idea further extends to the AI making connections that it was not specifically given; applying methods to situations where the methods weren’t originally intended to be applied, getting results, and either discarding the results as nonsensical or realizing that they are valid.
Philosophy of AI • The philosophy of artificial intelligence basically asks the question: “Can machines think?” • In the years since it was proposed, several answers have been given: • Newell and Simon’s physical symbol system hypothesis: • A physical symbol system has the necessary and sufficient means of general intelligent action • Claims the essense of intelligence is symbol manipulation • Hubert Dreyfus argued that, on the contrary, human expertise depends on unconscious instinct rather than conscious symbol manipulation. • Turing’s “polite convention”: • If a machine acts as intelligently as a human being, then it is as intelligent as a human being. • This ‘convention’ forms the basis of the Turing test.
Philosophy of AI • Turing Test: • A test in which a human judge has a normal conversation with one other person and one machine who are both trying to appear human. If the human judge cannot tell which one is which, the machine passes the test. To keep it fair, the conversation is usually text-based, as in through an instant messaging service.
Philosophy of AI • Gödel's incompleteness theorem: • There are statements that no physical symbol system can prove. • Many claim that Gödel's theorem limits what machines can do. • Searle’s “Strong AI” position: • A physical symbol system can have a mind and mental states. • Searle counters this assertion with his Chinese room argument.
Philosophy of AI • The Chinese Room: A thought experiment designed by John Searle to show that a symbol processing machine (like a computer) can’t be properly described as having a "mind" or "understanding", regardless of how intelligently it may behave. • Searle asks his audience to imagine that there is a computer which behaves as though it understood Chinese. The computer takes Chinese characters as input and, following a program, produces other Chinese characters, which it presents as output. This computer performs this task so well that it easily passes the Turing test. The conclusion that proponents of artificial intelligence would like to draw is that the computer understands Chinese, just as the person does.
Philosophy of AI • The Chinese Room: • Now, Searle asks the audience to suppose that he is in a room in which he receives Chinese characters, consults a book containing an English version of the computer program, and processes the Chinese characters according to the instructions in the book. Searle does not understand Chinese, but he simply manipulates the symbols, using the book version of the computer program. • After manipulating the symbols, Searle will produce the answer in Chinese. Since the computer passed the Turing test, so does Searle running its program by hand.
Philosophy of AI • The Chinese Room: • Searle argues that his lack of understanding goes to show that computers do not understand Chinese either, because they are in the same situation as he is. They are mindless manipulators of symbols, just as he is. They don't understand what they're "saying", just as he doesn't. • Since they do not have conscious mental states like "understanding", they can not properly be said to have minds.
Problems of Artificial Intelligence • Deduction, Reasoning, and Problem Solving • Early AI researchers developed algorithms that could imitate the process of conscious, step-by-step reasoning that human beings use when they solve puzzles or make logical deductions. • By the 80s and 90s, AI research had also developed successful methods for dealing with uncertain or incomplete information by using concepts from probability and economics. Problems: • For difficult scenarios, most of these algorithms require enormous computational resources. Most experience what’s called a "combinatorial explosion," where the amount of memory or computer time required becomes astronomical. Finding more efficient problem solving algorithms is a constant challenge for researchers. • In addition, humans do not normally solve abstract problems through step-by-step reasoning, using instead unconscious reasoning that AI researchers have been unable to duplicate.
Problems of Artificial Intelligence • Knowledge Representation • Most problems machines are expected to solve require extensive knowledge about the world. Some things that AI needs to have information on is: • Objects and their Properties • Categories and Relations between objects • Situations and Events • States • Time • Cause and effect • Knowledge about knowledge (what we know about what other people know) Problems • Default reasoning: Many of the things people know take the form of "working assumptions." • Qualification problem: for any commonsense rule, there is almost an incalcuable number of exceptions. AI research has explored a number of solutions to this problem. • Unconscious knowledge: Much of what we know isn't represented as “facts” or “statements” that we can actually say out loud. They take the form of tendencies and are unconsciously represented. • Common sense knowledge: The number of facts that the average person knows is astronomical. Research projects that attempt to build complete commonsense knowledge bases require enormous amounts of tedious step-by-step ontological engineering.
Problems of Artificial Intelligence • Planning: • Intelligent agents must be able to set goals and achieve them. • They need a way to visualize the future. They must have a representation of the state of the world and be able to make predictions about how their actions will change it. • They must also attempt to determine the utility or "value" of the choices available to it. • Problem: • Often, the agent assumes that it is the only thing acting on the world and it can be certain what the consequences of it's actions may be. However, if this is not true, it must periodically check if the world matches its predictions and it must change its plan as this becomes necessary, requiring the agent to reason under uncertainty.
Deep Blue • Deep Blue was a chess-playing computer developed by IBM. • On May 11th, 1997, the machine won a six-game match by two wins to one with three draws to Garry Kasparov, the world chamption. • The system derived its playing strength mainly out of brute force computing power. It was a massively parallel, 30-node, RS/6000, SP-based computer system enhanced with 480 special purpose VLSI chess chips. • It was capable of evaluating 200 million positions per second, twice as fast as its earlier version. In June 1997, Deep Blue was the 259th most powerful supercomputer in the world, capable of calculating 11.38 gigaflops. • A gigaflop is billion floating-point operations per second (FLOPS). • 11380000000 operations per second