John Koza
<p>Feb 18, 2009 7:01 PM
 
My Big Reaction:
 
This is Generate and Test.
 
USE Consciousness and Conscious Knowledgge Application to
Improve Generation and to Imrpove Screening prior to each TEST cycle
(assuming that screening is less costly than just testing).
 
CHESS PROGRAMS improved because humans were able to improve  node evaluation which
led to improved pruning of search tree  (better evaluation heuristics for terminal nodes) => more efficient
search.
 
Do same thing with GP.
 
Finally:  suppose you could include automated human-level node evaluation  (included machine consciousness)
(or recognize solutions from "the book")  OR  work around a given solution to find
a variation.  Also,  may need human consciousness to screen candidates for human acceptability
(parsimony, elegance, incorp of human values)
*
He is showing the way... surf Moore's Law
automated design and evaluation.
 
 
 
 
Look at TURING 's ideas in first group of slides:
 
Turing wrote a 1948 Essay "Intelligent Machines"
He refers to Cultural Search  
Another approach is the "cultural search" which relies on
knowledge and expertise acquired over a period of years
from others (akin to present-day <b>knowledge-based systems</b>).
CRITERION FOR SUCCESS
"The aim [is] ... to get machines to exhibit behavior,
which if done by humans, would be assumed to
involve the use of intelligence."
⎯ Arthur Samuel (1983)
 
 
Unfort. I missed John Koza's talk... slept thru it.
 
I scanned his slides briefly.  (on-line at  ee380).
 
Annual Human-Competitive Awares  (the Humies)
 
www.human-competitive.org
 
Definition of HIGH RETURN based on the AI RATIO:
 
AI ratio is the ARTIFICIAL TO  (Human) Intelligence ratio is the proportion of the prog-solving method delived by the machine
in comparison to the amount supplied by the huamn.
 
● Genetic programming engineers around existing patents
and creates novel designs more frequently than it creates
infringing solutions
 
ADFs work... Automat Defined Functions work and help  (these are subroutines)
 
 
 
PETA-OPS<b>
• Human brain operates at 1012 neurons operating at 103
per second = 1015 ops per second</b>
• 1015 ops = 1 peta-op = 1 bs (brain second)
 
 
Programs are in LISP!
 
main site is www.genetic-programming.org
 
Other star in this field is  William LANGDON...  maintains website with  5000 papers as of Jan 2003 by
1000 authors.
 
Conferences on GP:  GECCO:  www.SigEvo   Genetic and Evolutionary Computation Conf.
 
MAIN POINTS OF GP-1,2,3,4 BOOKS
Book Main Points<b>
1992 • Virtually all problems in artificial intelligence, machine
learning, adaptive systems, and automated learning can be
recast as a search for a computer program.</b>
• Genetic programming provides a way to successfully conduct
the search for a computer program in the space of computer
programs.
<b>1994 • Scalability is essential for solving non-trivial problems in
artificial intelligence, machine learning, adaptive systems, and
automated learning.
• Scalability can be achieved by reuse.
• Genetic programming provides a way to automatically
discover and reuse subprograms in the course of automatically
creating computer programs to solve problems.</b>
1999 • Genetic programming possesses the attributes that can
reasonably be expected of a system for automatically creating
computer programs.
<b>2003 • Genetic programming now routinely delivers high-return
human-competitive machine intelligence.
• Genetic programming is an automated invention machine.
• Genetic programming can automatically create a general
solution to a problem in the form of a parameterized topology.
• Genetic programming has delivered a progression of
qualitatively more substantial results in synchrony with five
approximately order-of-magnitude increases in the expenditure
of computer time.</b>
 
PROGRESSION OF QUALITATIVELY
MORE SUBSTANTIAL RESULTS
PRODUCED BY GENETIC
PROGRAMMING IN RELATION TO FIVE
ORDER-OF-MAGNITUDE INCREASES IN
COMPUTATIONAL POWER
• toy problems
• human-competitive results not related to
patented inventions
• 20th
-century patented inventions
<b>
• 21st
-century patented inventions
</b>
<b>• patentable new inventions</b>
 
John ran his Beowulf cluster (1000 node Pnet II parallel machine) in 2002 and got two patentable new inventions
 
CRITERIA FOR “HUMAN-
COMPETITIVENESS”
<b>
● The result is equal or better than human-designed solution
to the same problem
</b>
●<b> The result was patented as an invention in the past, is an
improvement over a patented invention, or would qualify
today as a patentable new invention.
</b>
● The result is equal to or better than a result that was
accepted as a new scientific result at the time when it was
published in a peer-reviewed scientific journal.
● The result is equal to or better than a result that was
placed into a database or archive of results maintained by an
internationally recognized panel of scientific experts.
<b>● The result is publishable in its own right as a new scientific</b>
resultindependent of the fact that the result was
mechanically created.
● The result is equal to or better than the most recent
human-created solution to a long-standing problem for
which there has been a succession of increasingly better
human-created solutions.
● The result is equal to or better than a result that was
considered an achievement in its field at the time it was first
discovered.
 
● The result solves a problem of indisputable difficulty in its
field.
● The result holds its own or wins a regulated competition
involving human contestants (in the form of either live
human players or human-written computer programs).
</p>