Diff: ArtificialIntelligence

Differences between current version and predecessor to the previous major change of ArtificialIntelligence.

Other diffs: Previous Revision, Previous Author, or view the Annotated Edit History

Newer page: version 11 Last edited on Tuesday, December 30, 2003 6:56:50 am by SergeStinckwich
Older page: version 7 Last edited on Tuesday, September 23, 2003 6:26:19 pm by JohnMcPherson Revert
@@ -1,6 +1,10 @@
-Currently the term "[ ArtificialIntelligence]" is used almost exclusively as a marketing buzzword. In so far as it has any meaning it means either:  
+ArtificialIntelligence, very often abbreviated as [AI ], is defined as ''making a machine behave in ways that would be called intelligent if a human were so behaving''. The TuringTest is a famous criterion for whether a system is intelligent. Efforts in this area led to [Prolog], SymbolicManipulation, [NeuralNetwork]s, [ExpertSystem]s, and other techniques, which were very important contributions to the field of ComputerScience. As such, even if we never succeed in creating ArtificialIntelligence, the research is giving us valuable results.  
-# the development of programs and computers that act like people, a la the [TuringTest]. Work on this has been somewhat haphazard , but various attempts have given us [Prolog], [SymbolicManipulation], [NeuralNetwork]s and other techniques. After 30 years the easiest way to build a machine with human intelligence takes 9 months...  
-# anything a computer does for the first time. The first Accounts Receivable system was [ ArtificialIntelligence] , as was the first [Compiler], the first [WebBrowser] and the first [MTA]
+Outside ComputerScience the term ArtificialIntelligence is used almost exclusively as a marketing buzzword , meaning anything a computer does for the first time. The first Accounts Receivable system was called ArtificialIntelligence, as was the first [Compiler]. Many systems that just react to changes by following simple sets of rules are called "intelligent" in MarkeTroid lingo
-The definition of Artificial Intelligence changes as programs get more sophisticated - mostly because things that we think require intelligence eventually prove to be do-able with heuristics and/ or brute force and/or new algorithms . For example, machine vision ( recognising objects, driving cars, etc) used to be thought to require intelligence , but now computers can do these things fairly easily, without what we would call "intelligence"
+Computer scientists distinguish two goals in [AI] research.  
+* A __Strong__ ArtificialIntelligence would be able to truly reason and solve problems, and would be sentient, ie self-aware. No progress has been made in this department. After so many decades, the easiest way to create an entity with human intelligence takes 9 months. (Which maybe isn't so bad - it's more fun that way, too.) %%% %%%  
+* In contrast, a __Weak__ ArtificialIntelligence acts as if it were intelligent, but can't actually reason and solve problems. Such [AI]s employ heuristics and rule sets to make decisions, but act "intelligently" by trying to refine their rules and heuristics automatically based on their assessment of the success of their reactions. Modest progress has been achieved in this area. [ExpertSystem]s are the most major achievement.  
+ The scope of ArtificialIntelligence changes as computers get more powerful. Things that were assumed to require Strong ArtificialIntelligence turn out to be doable by a Weak ArtificialIntelligence, or even deterministically . An example is machine vision, ie recognising objects, driving cars, etc. As image processing methods have advanced , and with the explosion of processing power in current computers, simplistic heuristics have become viable and achieve fairly good success