Now recalling an interesting declaration of endeavor in Machine Teaching in an earlier blog entry. One component of the vision is the creation of useful, verifiable, efficient and generalized knowledge communication. Let’s, for the purpose of this fam-blog-based communication, call this Knowledge Science and Technology(KST, and MT-KST) representing a subset of all that is required conscience Machine Teaching. (In particular separating epistemological considerations from pedagogical considerations. Secondarily, it separates Scientific and Technological concerns from Political ones: policies, laws, monitoring and enforcement. Lastly, while MT-KST cannot be developed for computers completely independent of its KR, it is still separate from internal representations that are more geared towards internal reasoning and operations)
One puzzling question is the definition of knowledge, what is knowledge?
One stab in the dark, in the framework of MT-KST, it would seem logical to dictate that knowledge is transmittable information. Untranslatable and undescribable knowledge, while possibly worth while for the host system to organize and remember, is not worth anything to others and therefore not knowledge.
A related problem is the ROI of the effort to form such knowledge. AFAIK, nobody reads this blog as I write it. Is it worth it for me to create it? If I can derive the Fermat’s last theorem in 1ms, is it worth remembering it? If the work to create knowledge can be embedded within my own thinking and doing process, then it needn’t be externalized into knowledge as communication. For another example, it would appear to me that opening my mouth is a procedure that needs no teaching. The effort to fully teach it via information transfer is not efficient or effective. Even though it is context for eating, it is something babies do before they were born.
This far we have gathered two types of information that is surely not required in MT-KST:
- The knowledge that is not describable. A pathological example of this are statements that change meaning once interpreted: this sentence is untrue.
- The knowledge that is pre-built in the communicating systems.(Redundant idempotent communication)
- The knowledge that is trivially knowable through routine reasoning. (Performing cost-benefit analysis trading off communication costs with operating thinking costs)
What else? Can we describe knowledge further for MT-KST?