24 Mar School of Hard Knocks
Are you a graduate of the School of Hard Knocks, or like me, are you still trying to escape the gravity of freshman year? Lessons about how altitude affects physical objects may be learned by slipping and falling down a few stairs. These lessons become ingrained early. Burned fingers have a profound impact. Notions of hot and cold, large and small, near and far, accessible and inaccessible, safe and dangerous quickly become part of a person’s common sense. How do we survive in this messy, dangerous world? What would you need to teach general purpose robots to enable them to effectively navigate the planet and the society: things we humans do all day every day without the proverbial “second thought”?
|Understanding Context Cross-Reference
|Click on these Links to other posts and glossary/bibliography references
|Common Sense and Thresholds
|Generalization and Inference
|Seely Brown 1989
Two words that embody much of the knowledge that constitutes our common sense are NATURE and HUMANS. In nature and in society we observe the multi-dimensional phenomena that bring about change and influence the cycle of action and reaction. Though we may take for granted the depth of these complex interactions between things in the world, the ebb and flow of cause and effect that we intuitively understand is not easily expressed in terms computers can use for automatic reasoning. Common sense is a reality for real brains, but it is still a dream for artificial ones.
If, however, we can capture the essence of common sense, and be able to represent the information and the processes in ways that computers can do something useful with, we will be on the way to a breakthrough in the information revolution.
What is common sense? Knowing the natural rules? Two primary kinds of knowledge are essential for common sense reasoning:
- existential knowledge of the attributes of objects; and
- causal knowledge of the consequences of actions.
Common sense knowledge also includes these two caveats:
- Objects in the real world have natural attributes.
- Actions are subject to and limited by natural laws.
Reasoning to Learn
Generalization is a reasoning process. In this process, when we encounter something new, we seek similarities with things we already know about; then, based on those similarities, we categorize the new thing. By categorizing, we place the thing in a hierarchy.
Generalization is a cognitive behavior exhibited even in infants. Very young children will respond to a person whose appearance resembles that of their closest kin (mom or dad). This is due to their ability to generalize. As children grow and begin to hear language, even without prompting, they will start to identify affixes that represent number (plural) or tense (past) and create words such as “mans” and “goed.” This kind of learning process involves experimenting (whether witting or not) with words they have never before heard that are constructed of familiar words and familiar affixes.
Because of the tense and gender richness of French, as an example, one could presume that French children have more fertile ground for experimenting. This would be even more pronounced in Bahasa Indonesia since affixes in that language include infixes as well as prefixes and suffixes.
Complex cognition requires basic knowledge and the circuitry for making associations. The circuitry alone, however, is not enough. We must be motivated to develop these higher cognitive skills and use them. Because this goes beyond the bounds of typical cybernetic research, we’ll leave motivation for another book. But it is important to note here that there is an innate curiosity or thirst for knowledge that prompts even the smallest of us to reach out with wonder to learn about this amazing world.
|Click below to look in each Understanding Context section
|Perception and Cognition
|Language and Dialog
|Apps and Processes
|The End of Code