20 May Cybernetic Modeling for Smarty-Pants
Model railroads come in several scales: O, HO and N gauge enable hobbyists to model real-world objects in miniature using successively smaller standards. In N gauge it is possible to build an entire city in the basement. A good model photographed with still or motion pictures may be so realistic that viewers believe they are looking at the real world rather than a masterfully crafted fake.
The Turing test asks this classic question: Is it real, or is it Memorex (sorry for the old reference)? Artificial intelligence does not attempt to shrink the brain to 1/20th of its actual size and craft a look-alike; rather, it seeks to analyze the brain’s functions and attempts to mimic them. Here is an important question embodied in the Turing Test: What constitutes intelligent behavior? Does a pocket calculator exhibit artificial intelligence? Some people struggle with math.
|Understanding Context Cross-Reference
|Click on these Links to other posts and glossary/bibliography references
|Deixis and Context
|Modeling after a Fashion
|modeling artificial intelligence
|Barr 1982 Winston 1984
|Bobrow 1975 Barr 1989
|Tanimoto 1987 Minsky 1975
In this section of the Understanding Context blog, we look at some cybernetic modeling using a variety of computational paradigms, including different approaches to neural networks and some non-neural AI approaches. This section also expresses and defends the belief that it is often wise, efficient, and productive to combine diverse paradigms or techniques in a single system to produce a robust and intelligent processing mechanism. Then, in the last section, we turn that idea on its head and say we should break apart systems and have medium- to fine-grained services that are invoked based on the goals of the user and the facts as they exist in the moment.
the “logic” below
to see the Author’s per-
spective on the tired old
question, “Is there such a
thing as artificial intelligence?”
Whether you agree with it
or not, the objectives and the
results of many ongoing AI
efforts yield very useful
results, even if the best
AI technologies can
only serve as
Logic: A computer cannot think if we define “thinking” as something that only intelligent biological creatures can do. But machines can perform persuasively similar functions (brain tasks). How old were you when you learned linear algebra? Did it require intelligence on your part? Did it even tax your intelligence? Computers, given the same kind of instructions that you were given, can do linear algebra with a great deal more speed and accuracy than we can. There are already computers out there that are smarter than we are in certain brain tasks. Does this constitute intelligent behavior? Is it artificial? If so (and I think so), man has been successfully using artificial intelligence for many decades.
Cybernetics is a modeling science. If our objective is to make smarter computers, we need to begin with a model. So far in this blog, we have worked under the assumption that the human brain is worthy of emulation. If our only objective were to model the brain, however, perhaps a little gelatin and some gray food coloring would suffice.
Patrick Winston, a pioneer in artificial intelligence (AI) research, identifies two primary goals of AI:
- Make computers more useful.
- Help us understand the “principles that make intelligence possible” (Winston, 1984).
As we develop ideas about computer models of intelligence, we need to keep Winston’s pragmatic objectives in mind.
The illustration at right is a model of the raw material used for MIPUS’s brain. By the time we are through with this section, we should know what was poured into this lifeless fabrication to make it into a sentient machine. We should also know how to evaluate models for mechanical brains.
There is no limit to the number of commercial, political and private enterprises in which more intelligent and useful computers could be valuable. As you study modeling and compare the characteristics of different models, try to keep in mind the actual application of these models to real-world problems.
|Click below to look in each Understanding Context section
|Perception and Cognition
|Language and Dialog
|Apps and Processes
|The End of Code