Skip to content
Mark Janssen edited this page Jan 6, 2020 · 4 revisions

Droids come pre-trained with a sophisticated "world view": a basic knowledge of the world. This is essentially cloned in all units with a version number for any additions. The goal is for the buyer to be able to say "Buy a 1/4 block of cheddar cheese." and it would know that this is available at a grocery store, that grocery stores are indexed in the public business database, and that it needs a digital cash device with appropriate amount available to purchase it, as well as the idea that purchases must be made before leaving the premises with any merchandize.

Rules are added through predicates. Trained much like Lenat's CYC or Mindpixel's database. "To purchase something requires money." Each keyword can be looked up in its own database of preprogrammed predicates until it reaches the leaves at the bottom of it`s language tree where no more data exist and the robot can query for clarification of the word used when either a word-conflict arises or an absence of knowledge.


Different trees of knowledge:
  • language: holding an organized tree of linguistic knowledge, downto definitions of word elements.
  • world map: holding place names and location data, perhaps access obstacles and pathway experiences.
  • family affiliations: holding trust values and derived therefrom as well as hierarchy of control.
  • and specially: it`s task list: a prioritized list of things it has yet to accomplish (a tree once subtasks to accomplish are made)

If a robot is caught breaking a law by a human, the human can take over ownership of the unit (or be given to scavengers for reclaimation and re-use). This new owner can re-program the units objectives. The owners task list is wiped or reviewed beforehand to see what caused the malfunction (misprogramming by the owner).
See also:
Clone this wiki locally