Learning by Asking Questions and Learning by Aligning Stories: How a Story-Grounded Problem Solver can Acquire Knowledge
Name
CMHI-Report-3.pdf
Size
7.49 MB
Format
Adobe PDF
Checksum (MD5)
343cd7aabff83178e427021d9023349e
Author(s)
Yang, Zhutian
Winston, Patrick Henry
Date Issued
December 17, 2018
Series/Report no.
3
Abstract
We describe how a problem solver, grounded in the Genesis Story Understanding System, acquires knowledge by asking questions and by aligning successful problem-solving stories.
To illustrate learning by asking questions, we demonstrate how the Genesis problem solver learns the steps to mix a blocks-world martini from another problem solver and how it learns to make a real-world fruit salad from a human.
To illustrate learning by aligning stories, we demonstrate how our Genesis problem solver learns to replace a phone battery from two 80-word stories that have much irrelevant detail and nothing expressed in exactly the same way.
We conclude that the Genesis problem solver learns much like humans do. It asks questions and exploits precedents. It learns something specific from each experience. It tells itself its own story as it solves problems, exhibiting a kind of self-aware behavior.
Description
This report has related videos that can be viewed online (see "has part" links below).
Subjects
computational models of human intelligence
story understanding
self-aware problem solving
learning by asking questions
learning by aligning stories
Persistent DSpace Link
