I remember watching a Star Trak episode, where Riker mentions that he has been up all night programming the holodeck; programming the holodeck? Now what would that be like?
I think we are getting close to know the answer to that question. “ I want to be in an environment like Hawaii, where there are lots of nearly naked…” you get the picture; programming the holodeck means he has been up all night talking to his computer, describing his fantasy. Sergey is right LOP is the next step in the inevitable evolution of intelligent human-machine interactions.
As seem to be the case in fast pace environments, communities doing “analogous” activity don’t always see the connections, but as I read Sergey vision my mouth was hanging open in astonishment.
After it got embarrassing for people in the Artificial Intelligence field to call themselves AI’er, many of them started calling it the knowledge representation community. With the advent of the Semantic Web, and even before, there have been differing ways to capture, and persist knowledge, as apposed to information. One of the things that all most all of the knowledge representation formats require is an “ontology” of the domain concepts.
Since a domain ontology is itself a “knowledge base” it can be also be represented and manipulated in a structured knowledge store.
One such representation is the “conceptual graph.” Concept graphs were conceived of by John Sowa, one of the elders of the AI field. Sowa has recently been in the forefront of the development of Common Logic Controlled Language as a natural language domain ontology first order logic processing system. There is much work still to be done in this area; however there is a common logic standard group which is trying to solidify a conceptually sound yet easy to uses natural language way to represent problems and their solutions logically and unambiguiously.
What I was amazed at while reading Sergey’s description here was that company I work for has developed a multi-agent knowledge base system for aiding in decision making. I have been charged with developing an IDE for developing the configuration files or “knowledge modules” for the agents to use as marching orders. I have developed a interface that look a bit like the IntelliJ’s frameworks in terms of layout, though not nearly as spiffy. But the editors are a workflow editor, a conceptual graph editor, and now we are working on a rule editor, and so a decision tree, etc. Framed concepts show the relationships between concepts, and concepts are defined in an ontology and concepts have “slots” which are attributes.
Though modules may have different purposes they are in the end all knowledge models and they all use a common interface which emits saved files in our knowledge representation language. The next part of project is to build an ontology editor hopefully one that can populated a domain ontology using simple humanistic-interaction methods.
It seems to me that in the near future corporation won’t have “programmers” but there will be a need for people who can capture corporate and cultural domain ontologies for the enterprise, and “problem template” specialist. I think we are quickly approaching the time when programming a computer will amount to little more than “telling” in the computer what to do, e.g. “please go to the data warehouse and reorganize the dimensions so you can tell me what the soft drink sales volumes have been for the twenty-fifth of each month over the last ten years, and compare it to the volumes of the fifteen of each month for the same period.” Or, “Find in my image stores all of the pictures of children in dangerous situations.”
I applaud Sergey vision and look forward to hearing more about the field of LOP.
