The importance of knowledge representation

Posted on by Alan White

How will we find enough people to design, engineer and build the spaceships of the future — the mile-long versions that will course through the universe? How will we reengineer the human genome and get rid of the errors that lead to disease and suffering? These achievements will require vast legions of engineers, scientists and project managers. There will never be enough people to fill those roles. We’ll depend on machines to supplement our ranks.

So how will we get those machines to mimic us? Will we use Python? R? Neural networks? AI algorithms? How will those technologies scale and provide the flexibility needed to store the knowledge and produce the behaviors of our electronic helpers?

The solution starts from the bottom up, from the foundation. Knowledge representation is that foundation.  A knowledge representation system has to be flexible enough to represent all human knowledge, and scalable enough to handle both the volume and complexity of such knowledge. If you use the right system, you can build up and branch out ad infinitum. If you don’t, you’ll quickly reach a point of diminishing returns as your solutions fill up with kludges and epicycles. You have to pick the right building blocks to represent knowledge. You have to get your first principles right.

 

Is or Has

Posted on by Alan White

Facts are connected in one of two ways — “is” or “has”. On the is side, things are grouped and categorized according to their features. On the has side, those features are spelled out. Has captures details; Is describes patterns derived from those details. Is holds only groups or categories; everything else is Has. In fact, a better way to label these two axes is Is and Everything else. If we were to diagram our facts, we could put Is on the vertical axis and Has (“Everything else”) on the horizontal axis.

A collection of connected facts

Posted on by Alan White

Knowledge, at its core, is simply a description of one thing in terms of another. Jane is a person, Jane lives in Boston. We call the verb “transitive” in these cases because there is an object (“person”) to go along with the subject. But Jane can participate in activities on her own — Jane works, Jane sleeps. We call the verb “intransitive” for those statements, but they belie the fact that a lot is going on underneath — Jane works at a company, Jane performs a task at that company, Jane’s shift begins at a certain time and ends at a later time, etc. Ultimately, all knowledge involves linking one thing to another.

Factango demonstrates how you can step from one thing to another through a succession of interlinked facts. Go there and enter something in the Find field at top. Then click on any of the words below and move from one thing to another, from one fact to another. One thing is connected to another.