The notion of memory-as-a-place raises many "how" questions about representation. "The next few sections explain a theory of memory that tries to answer all these questions at once by suggesting that we keep each thing we learn close to the agents that learn it in the first place."
Minsky proposes knowledge-lines or K-lines. A K-line associates itself with agents; when the K-line is activated, this re-activates these agents. "In other words, we `memorize' what we're thinking about by making a list of the agents involved in that activity." More than one K-line can be attached to an agent.
"The basic idea is simple: For each familiar kind of mental job, your K-lines can refill your mind with fragments of ideas you've used before on similar jobs. In such a moment, you become in those respects more like an earlier version of yourself." In other words, memory is viewed as the re-activation of knowledge that was useful for solving similar problems previously. This related directly to a point Minsky made in Section 7.4, that the most efficient way to solve a problem is to already know how to solve it.
(NB: This gives a nice functional account of a K-line. But what is it with respect to other accounts (e.g., implementational, or at least technological). K-lines can be activated, and can activate agents. Should a K-line thus be viewed as a special (or even *not* special) kind of agent?)
Pearl Street | Society of Mind Home Page | Dawson Home Page |