Differences between version 9 and predecessor to the previous major change of ArtificialIntelligence.
Other diffs: Previous Revision, Previous Author, or view the Annotated Edit History
| Newer page: | version 9 | Last edited on Sunday, September 28, 2003 10:25:07 pm | by AristotlePagaltzis | Revert |
| Older page: | version 7 | Last edited on Tuesday, September 23, 2003 6:26:19 pm | by JohnMcPherson | Revert |
@@ -1,6 +1,9 @@
-Currently the term "
[ArtificialIntelligence]"
is used almost exclusively
as a marketing buzzword. In
so far as it
has any meaning it means either:
+[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.
-# 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]
.
+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.
-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
".
+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.
+
+
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.
+
+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
.
