Differences between version 10 and revision by previous author of ArtificialIntelligence.
Other diffs: Previous Major Revision, Previous Revision, or view the Annotated Edit History
| Newer page: | version 10 | Last edited on Sunday, September 28, 2003 11:35:48 pm | by AristotlePagaltzis | Revert |
| 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 progess 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
.
