This preview has intentionally blurred sections. Sign up to view the full version.View Full Document
Unformatted text preview: CS221 Lecture notes Introduction and history What is AI? We begin with an almost content-free definition: “AI is the endeavor of building intelligent artifacts or systems.” It’s very hard to make this more precise. People (even within AI) sometimes disagree about pre- cisely what falls within the borderlines of AI. 1 History AI was born in 1956, at a workshop in Dartmouth organized by John Mc- Carthy. Those gathered agreed to adopt McCarthy’s name for the new field: Artificial Intelligence. At that point, there was lots of enthusiasm. Things seemed to work out really well. Only a few years before, computers were viewed as large calcula- tors, and now truly intelligent systems seemed within reach. Early programs did amazing things by simply representing knowledge about a domain and searching for a solution. For example, Newell & Simon’s Logic Theorist proved qualitative mathematical theorems, and even found a shorter proof for one of the theorems in Russell and Whitehead’s “Principia Mathematica.” In 1958, McCarthy suggested how the same paradigm could be used for commonsense reasoning: represent knowledge about the everyday world as logical axioms, and use that knowledge to figure out how to act. Amazingly, a general-purpose logical theorem prover was able (for instance) to generate a plan for driving to the airport. Arguably the first convincing machine learning program, Arthur Samuel’s Checkers playing program started out playing poorly, but learned to play better by playing many games against itself. Growing to play better than Samuel, this program disproved the (still-made) argument that computers can only do what they are told to do. A particularly good example of how a simple set of rules can produce seemingly complex behavior was Joseph Weizenbaum’s Eliza program, which 1 2 simulates a Rogerian psychotherapist. Although Eliza’s algorithms are best described as simple pattern matching, and it was not intended as a serious attempt at machine intelligence, it still produced appropriate responses to a variety of statements. Because of programs like Eliza, there was also a hope of building systems in the near future that would pass the Turing Test for machine intelligence....
View Full Document
- Winter '09
- Artificial Intelligence