http:^^www.cs.unm.edu^~high^future.html

This data set contains WWW-pages collected from computer science departments of various universities

Date: Wed, 15 Jan 1997 02:13:04 GMT
Server: NCSA/1.4.2
Content-type: text/html
Last-modified: Tue, 27 Aug 1996 20:14:27 GMT
Content-length: 2489

<h1>Future Directions</h1>
<a name="Contents"></a>
<ul>
<li><!WA0><a href="#EIEIO">Engines for Emergent Intelligence</a></li>
<li><!WA1><a href="#Genetic">Evolutionary Techniques</a></li>

</ul>

<HR> <!----------------------------------------------------------->
<h2>
<a name="EIEIO">Engines for Emergent Intelligence</a>
<!- (return to <!WA2><a href="#Contents"RBContents</aRB) >
</h2>

Attempts to create machines exhibiting intelligent behavior can be roughly
sorted along a continuum between symbolic artificial intelligence (SAI) and
connectionist artificial intelligence (CAI).  <p>
Symbolic AI is a
top-down approach to the engineering of intelligent behavior.  
The physical-symbol hypothesis of Newell and Simon
<!WA3><a href="http://www.cs.unm.edu/~high/et001.html">(1972)</a> suggests (correctly, I believe) that human
thought can be fully described in terms of the manipulation of abstract
symbols, embodied as states in the physical world.  The common 
(mis)interpretation
(incorrect, I believe) is that symbols are discrete entities like those
found in computer programming languages (e.g. OBJECT = BIRD, COLOR = BLUE).
This is a bias due to both the computer scientist's special 
perceptual system but also the basic human perceptual system.
That mind has a physical basis is apparent.
That this physical basis can be construed as a processing of physical
symbols seems acceptable, but only if we allow a loose
definition of the word "symbol."<p>

<!- The various forms of connectionism suggest a bottom-up approach.->
more...
<ul>
<li>the importance of learning</li>
<li>the brittle, unlearning nature of discrete symbols</li>
<li>the learning power of connectionism</li>
</ul><p>
<ul>
<li>the importance of structural composabiliy</li>
<li>the uncomposability of connectionism</li>
</ul><p>
<ul>
<li>the emergent synthetic alternative</li>
</ul>

<ol>
</ol>


<HR> <!----------------------------------------------------------->
<h2>
<a name="Genetic">Evolutionary Techniques</a>
<!- (return to <!WA4><a href="#Contents"RBContents</aRB) >
</h2>
<ul>

<li>
<!-a href="ga002.html"->
Genetic Algorithms versus Artificial Evolution</a>,
</li>

<li>
<!-a href="ga001.html"->
Evolving complex computational systems</a>,
  <ul>
  <li>Harvey's method for evolving neural networks (NN)</li>
  <li>Gruau's method for evolving NN construction programs</li>
  <li>Koza's Genetic "Programming" </li>
  <li>learning classifier systems</li>
  </ul>
</li>

<li>
<!-a href="ga003.html"->
Complex Genotype to Phenotype Mappings</a>
</li>

</ul>