Outlines and Papers

• Many outlines are difficult or impossible to find on the WEB.  These should be linked through your support group leader's page.  An example that works is Joyce Duckett's homepage which works from Geocities.  This links us to Lisa White's page.  Her diary entry gets us to her Project Outline.  Dennis Wilno's home page links to his diary and then to the Project Outline.  Jason de Charleroy's robotics page will be the basis of a presentation.  Any others interested in presenting their work to the class should let me know.
• An interesting project is a set of scenarious portraying the life of an American student in Paris in 2020.  I added the links to the students' pages

• Supercomputers vs. MainFrames vs. Personal Computers.  Essentially a matter of computing power, the number of calculations that can be done per second.  However, computing power is becoming cheaper and cheaper, so the distinctions are less meaningful
• Parallel vs. Serial architecture - does the computer do one many things at once or one thing at a time?  The brain is parallel, vonNeumann design computers are serial.  Early AI was done on parallel computers, but today almost everything is done on serial computers because they are cheap.
• Some of the supercomputers are parallel.  These vary between SIMD, "single-instruction, multiple dataset" and MIMD, "multiple instruction, multiple dataset" which is more brain-like.
• Analog vs. Digital - almost all computers today are digital but there are still some analog computers.
• Networking has made it possible to have a number of Serial computers working simultaneously and exchanging information. This seems to be the way more powerful artificial intelligence will evolve

• The earliest computer software was written in "machine language" which consisted of 0’s and 1’s, that gave specific instructions to registers in the Central Processing Unit. Programmers had to know exactly how the machine did everything it did.
• The next step was "assembly language" which uses alphabetic key words to write patterns of 0's and 1's.  This was still pretty close to the logic of the machine, but was a lot faster to write.
• Next came "high level" languages such as FORTRAN and BASIC, which used words and numbers similar to mathematical algorithms, e.g., one might write:
• FOR N = 1-190; GRADE(N) =  TESTS(N) + HOMEWORK(N) + ATTENDANCE(N); NEXT N
• With high level languages, programs were not specific to a computer - there was a program called a "compiler" which converted the high level program to a form that could be understood by the "operating system" of a given computer.   The operating system is software which enables standard programs to run on any compatible computer.  Programmers did not have to know as much about hardware.
• JAVA is another step away from the details of the machine.  In a JAVA program, instructions are sent to a program called the "JAVA virtual machine" which translates them into codes which can be read by different operating systems.
• Early computer programs were organized around procedures (e.g., SORT, ADD, PRINT).  Programmers called up the procedures they needed them to process data, which was stored in a separate place. As programs got larger and larger, and especially when many programmers worked on them, this became cumbersome.  "Object-oriented" programs were created.  In these programs, the procedures and data are bundled together.  If one programmer needs to modify procedures for his or her purposes, those modified procedures can go into his or her "objects" without messing everyone else up.
• But programmers cannot be constantly reinventing the wheel each time they create an object.  They have general models or patterns or archetypes they rely on - in programming language these are called "classes".
• This "object-oriented" approach analogous to the logic of the brain, where there are thousands of independent units which have many features in common.

• Webmind is written in JAVA - its basic units are "nodes" which are "objects" in JAVA
• The nodes are arranged in a hierarchy:  from lobes to pods to nodes.  Nodes relate more easily to others in the same place in the hierarchy. But they can also form form heterarchical links, crossing hierarchical lines.
• There are several types of links between nodes:
• similarity links, representing the belief that one actor is similar to another.
• inheritance links, representing the belief that one actor is a special case of another.
• spatiotemporal links, representing the belief that one actor represents something occurring near the other one in time or space.
• containment links, representing the belief that the entity represented by one actor is contained inside another one.
• associative links, representing simply the fact that Webmind's dynamics tend to associate one actor with another.
• Webmind is not created with a fixed body of knowledge, it learns by reading text, numerical data and (in the future) other kinds of information
• As a Webmind assimilates information, it organizes it into modules.  It also develops modules to record its memories of how it does things, such as the:
• The Categorization Module which creates new categories
• The Text-Numerical Interrelation Module, which remembers how text and numbers are interrelated
• The Query Module which remembers how to interact with humans and respond to their questions
• The Social Module which manages relationships between Webminds and between Webmind and people
• The Self Module, which maintains a simplified image of everything going on within the Webmind
• The really tricky thing about Webmind is how all these nodes and modules are coordinated to make a coherent whole.  This coordination is what makes true intelligence possible.  There are several methods of coordination:
• Each of Webmind's components learns how to take the needs of other components into account, giving them the information they can best use
• Sometimes, one module is defined as dominant, and others are forced to defer to it
• Sometimes, modules take turns being dominant, each of them adjusting to the others for a period of time.   New thoughts emerge from this process of mutual readjustment.
• Sometimes, the Self module intervenes and sorts things out