The Slipnet can be roughly thought of as Copycat’s long term memory, where the knowledge of the concepts that are used in formulating solutions in the Workspace are stored. This knowledge is stored in a semantic network, where each concept is represented by a single node, and is linked to other concepts (i.e. other nodes) which are related in some way. Each node in the Slipnet has a corresponding activation value, which indicates how relevant Copycat deems the concept to be in the formation of a solution at each instance in time. These activations are used in determining what conceptual slippages are likely to occur within the Workspace (which led to the name "Slipnet"), as well as having an influence over the strength of structures built in the Workspace (which will be discussed in the "Coderack" section of this tutorial). This section describes the Slipnet interface, the concepts that it contains, the types of links that connect these concepts, how these concepts gain activation, and how these activations affect the types of structures built in the Workspace.
In the top right corner of the Slipnet window,there is a maximise button (which is represented by a triangle which is pointing up). You can press this button to see the Maximised Slipnet display. A network representation of the Slipnet should appear where the Workspace was, and the Workspace will become minimised and lose some of its detail. Now selecting the option "Show Slipnet Key" from the "Visible" pulldown menu, will display in the lower right window, a key for identifying the concepts in the slipnet :
Concepts in the Slipnet are connected by a link if they are semantically similar in some respect. Each of these links has an associated length, which describes how conceptually close they are (with a short link indicating that they are very similar). Slip links differ from the other forms of links in the Slipnet in that they connect concepts that are potential conceptual slippages.
Objects across domains can be mapped if they share a distinguishing descriptor, or distinguishing descriptors that are slippable concepts (according to the Slipndet).
Exercise 1. For the problem abc : abd, ijk : ?, write down the distinguishing descriptors for the letter "c", and from this, determine what objects in the target string it could be possibly mapped to.
The concepts in the Slipnet become fully active when structures which are described by these concepts are built in the Workspace (e.g. the concept successor becomes fully active when a successor bond is built). As stated earlier, these activations represent how relevant Copycat deems each concept to be in the formation of a solution at that point in time.
In the current implementation, the level of activation of a concept is shown by the brightness of the background colour of the node, with black indicating a low level of activation, and white indicating a high level. When the node is fully active, the border of the node is drawn in red.
Exercise 2. Set up the problem abc : abd, ijk : ? using a random seed of 3. Now run the problem until Codelets Run = 70. Press Single Step, and describe the kind of bond that is formed at this time.
Continue pressing Single Step from the Main Control Window until the concept corresponding to this bond (i.e. either Successor or Predecessor) becomes fully active. Note that the direction of this bond also becomes active. What is the value of Codelets Run at this point?
This delay in the activation jump of the Slipnet concept node corresponding to the built structure is due to the fact that the Slipnet is updated every 15 Codelets that are run.
So, once a structure is built, the nodes corresponding to the structure in the Slipnet become fully active a short time later. This places top down pressure on the System to try to build more structures of this type.
Exercise 3. To illustrate the influence of top-down pressures, run the previous problem (i.e using a random seed of 3) and note down the type of bonds that are used to describe the strings in the final solution.
Now press Stop to reset the problem, and Clamp the predeccessor concept in the minimised Slipnet. This is achieved by selecting the Clamp from the Slipnet Actions Window, and clicking on the node that you wish to clamp. A red circle should appear in the concept in the minimised Slipnet. Now fully activate this concept by selecting the eyedropper from the Slipnet Actions Window, and clicking on the node. Now set the random seed to 3, and run the problem. Note down the types of bonds that are now formed in the final solution.
One thing to note about the activations of concepts, is that they decay over time if they are not reactivated by the building of corresponding structures in the Workspace. This is important, because initially the structures that are built are quite random and are not globally consistent (e.g. both a successor and predecessor bond can be built in the string abc at the same time). If it were the case that both these concepts placed equal top-down pressures on the situation for the duration of the run, it would be difficult to built a structure that is globally consistent (i.e. with all strings being described by the same types of bonds and groups).
Exercise 4. Remove the clamp from the predecessor concept from the previous exercise (by selecting the Clamp from the Slipnet Actions Window, and clicking on the node). Now select the Arrow from the Slipnet Actions Window, and drag the concepts successor and predecessor to the Graph Area (Note: Nothing will appear in the Graph window until the Copycat is running a problem). Run the problem abc : abd, ijk : ?, using a Random seed of 4. You should note that the activations of predecessor and successor are related to the structures present in the Workspace.
In the following box, draw the graph of the activations of the successor and predecessor concepts.
By the end of this section of the tutorial you should: