Imagine that your computer could think just like you can. It could carry on a conversation with you. You could tell it what you want to do, verbally ask it to find things or answer a question. You could dictate the essay you want to compose instead of typing anything down.
Such intelligent computer systems would be nice, but they would also bring up lots of ethical questions. If the computer can replicate human thinking, can it replace humans in some current jobs? In all current jobs? How does society decide how to handle such a powerful tool... or has it become something much greater? Are the protrayals of Hollywood movies like I. Robot or The Matrix something we should worry about?
While A.I. can bring up a lot of issues, you don't have to worry quite yet. It turns out that you - a human - happen to be a lot more advanced than you might have thought. The problems of using vision to see the world around you, using natural language to communicate with other people, or even the mundane task of driving to the store... it's all a lot, lot more complex than it seems at first glance. Consider the difficulty of producing the Java programs that we've created in this class. Given just a limited set of possibilities those programs had to contend with, it was still quite a challenge to design the atomic steps of the algorithm to handle the problem and then translate that into Java code. How then do you handle the enormous complexity of going out into the "real" world?
The textbook in Chapter 16 gives a nice introduction to some of the topics of A.I. For one, it gives a rough comparison between humans and A.I. on their performance on tasks of different complexity. Humans tend to do fairly evenly for all tasks. You can add up a set of numbers, read a paragraph of text, or develop a simple theory about something all with pretty good proficiency. Computers, on the other hand, are great at simple tasks (a calculator's got you beat when it comes to adding numbers) but have a lot of trouble with complex tasks (why do you think you have to enter "key words" into Google's search instead of just writing sentences?). That's not to say that the field of Computer Science hasn't tried to develop these abilities. Everyone thought these challenges would be conquered in a couple of decades... that was back in the 1960's.
A.I. in a lot of cases is just trying to duplicated the abilities that humans easily demonstrate on a daily basis. But the problem, just like when you first started writing Java programs, is that you might be able to solve the task yourself but you never analyzed exactly what you were doing at each step. It's hard to write a simple algorithm to do whatever it is you do on a daily basis. To start out, it's hard to simply define what "intelligence" is.
One measure that was developed early on by Alan Turing is the Turing Test. The test is basically that the computer must be indistinguishable from a human based upon written exchanges. So, if you were chatting on an Instant Messenger client with one human and one AI program and couldn't tell, then in this sense the computer program has duplicated some human abilities. This is just one measure, though. In Artificial Intelligence computer scientists study tasks that we would typically associate with human intelligence.
- performance comparison between humans and computers (humans reasonably good at all tasks vs. computers very good at easy tasks and very bad at complex tasks) - what is AI? The Turing Test here. - what does "intelligence" mean (Ideal Intell., Moving target) - comparison to human flight - the "all vs. many" approaches (first tried to get a single program to do everthing that a human does, now AI is divided into sections just trying to duplicate a single human-like ability) (Early Efforts...) - (Modern AI) - Topics: - can talk about chess/checker game playing of computers - can talk about computer vision (what can you see through a pin hole?) - motion planning (where are you in your environment, what path should you choose to go from point A to point B)