Differences between version 10 and previous revision of ArtificialIntelligence.
Other diffs: Previous Major Revision, Previous Author, 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 9 | Last edited on Sunday, September 28, 2003 10:25:07 pm | by AristotlePagaltzis | Revert |
@@ -1,9 +1,10 @@
-[
ArtificialIntelligence]
, also known
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 this has been achieved
.
+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
.
-Strong
ArtificialIntelligence research
is an attempt to create
computer-based intelligence that can truly reason and solve problems
. Such an [AI] would sentient
, or self-aware. Initial efforts in this area led to
[Prolog
], SymbolicManipulation, [NeuralNetwork]s and other techniques
. However, no real progess has been made, and the easiest way
to create an entity with human intelligence currently takes 9 months
.
+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
.
-In contrast, a weak ArtificialIntelligence can't actually reason and solve problems, even if acts as if it were intelligent. Modest progress has been achieved
in this area
.
+Computer scientists distinguish two goals
in [AI] research
.
-The scope of
ArtificialIntelligence changes as computers get more powerful. Things that were assumed to require intelligence often turn out
to be doable with heuristics
, or even deterministically
. As an example, machine vision (recognising objects, driving cars
, etc) used
to be thought to require
intelligence. However, since image processing methods have advanced and current computers have orders of magnitude
more processing power than just
a few years ago
, simplistic heuristics
can be applied
to this problem with fairly good
success.
+* 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
.
-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 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
.
