CogSci1 Top 10 List Chapter Summary
xex Home

The following Top Ten List of Big Ideas summarizes Chapter 4, "Symbolic AI", of Stan Franklin's [[Artificial Minds]]. The the words are generally taken straight from the text.

  1. Production Systems A production system is purely a syntactic device manipulating symbols without regard to their meaning. A production system contains three parts: the global database, the production rules, and a control structure.

  2. Turing Machines Turing machines exist as mathematical entities. A Turing Machines also consists of three parts like a production system. The three parts of a during machines are the tape, the read-write head, and a finite-state machine. The Turing Machine was created by Turing because he wanted to prove theorems about what could and could not be computed. A Turing Machine can be used to compute anything that is computable.

  3. Turing Machines v. Production Systems Turing Machines can be emulated by a production system and vice versa. Any Turing machine can be emulated by some suitably defined production system. Production systems, along with Turing machines are capable of computing anything that is computable.

  4. Parallel Production Systems Parallel production systems are production systems where more than one production rule can simultaneously be fired. It is believed that (at the time this book was written) they are approaching at a high speed. A parallel production system would have to utilize shared memory machines, a memory machine that the processors communicate with each other by writing to the memory.

  5. Production Rules Production rules are often called productions. These productions are condition/action rules. A condiction/action rule states that whenever a certain condition is satisfied, then a specified action is perfomed or may be performed.

  6. SOAR "The goal is to provide the underlying structure that would enable a system to perform the full range of cognitive tasks, employ the full range of problem solving methods and representations appropriate for the tasks, and learn about the aspects of the tasks and its performance on them. (Laird et al. 1987, abstract, p. 1)

  7. SOAR Doesn't Work There are several different reasons why SOAR doesn't work how it is meant to. These reasons are that SOAR has no deliberate planning facility, it has no automatic task acquisition, several important learning techniques have still not been incorporated into it, SOAR's single learning mechanism is montonic, and SOAR can not interact with the real world in real time.

  8. SOAR's Mechanisms SOAR's goals are represented symbolically and progress towards them is accomplished by symbolic processes, SOAR explicitly uses its knowledge to control its behavior, SOAR's problem-solving activity is based on searching its problem space for a goal state, SOAR uses a means-ends analysis, productions in SOAR serve only as memory which react to cues by recalling associated facts to working memory,and SOAR's decision procedure uses only preferences.

  9. SOAR's Applications One of SOAR's most successful commercial expert systems is R1. R1 is a knowledgeintensive, special-purpose expert system, carefully tuned to its specific task, that does as much direct recognition and as little search as possible.

  10. SOAR's Hypothesis SOAR has many different hypothesis, all which lean on the first. The first hypothesis of SOAR is the physical symbol system hypothesis. The physical symbol hypothesis states that every general intelligence must be realized by a symbolic system.
    11.