Lecture7

Subtle intermittent bug hard to detect other bugs

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: t setup positions came up “1” if incorrect, otherwise zero. However, it was stored in ONE BYTE. Every 256th time the routine was called, the flag overflowed and showed a value of zero anyhow. (Subtle, intermittent bug; hard to detect.) ❚ Other bugs sometimes allowed the machine to ignore changes or corrections made by the operator (edits not passed in real time). CIS 350 Lecture 7 © Roxanne Hiltz and Eli Rohn 20 10 Spring 2000 - Lecture 7 Therac 25: What went wrong? ❚ Manufacturer was informed of the incidents (several times) but responded that the machine could not possibly be responsible for the massive overdoses. ❚ Later, some machines removed from service and modified by manufacturer; but did not catch ALL the bugs before returned to service. CIS 350 Lecture 7 21 Therac 25: What went wrong? ❚ This case illustrates many of the things that a responsible, ethical company would not do. CIS 350 Lecture 7 © Roxanne Hiltz and Eli Rohn 22 11 Spring 2000 - Lecture 7 Expert Systems In Medicine ❚ Expert systems are an applied type of Artificial Intelligence ❚ To develop an expert system, AI researchers spend years picking the brains of human experts, extracting their knowledge and trying to understand the way it is organized CIS 350 Lecture 7 23 An Expert System Includes 1. A problem statement with a goal: limited domain-2. A knowledge base 3. A problem solving strategy that determines which knowledge to apply and constructs the lines of reasoning. ("heuristics on next slide) CIS 350 Lecture 7 © Roxanne Hiltz and Eli Rohn 24 12 Spring 2000 - Lecture 7 An Expert System Includes Heuristics are intellectual rules of thumb, or the method by which human experts make "good guesses." 4. A friendly interface that is "natural" to the user CIS 350 Lecture 7 25 Examples Of Expert Systems: ❚ MYCIN is the classic example of an expert system. ❙ It was created around 1972 to diagnose and propose treatments for bacterial infections. infections. ❙ It uses a probabilistic form of backward rule chaining, where assertions are tagged with certainty factors of –1 (false), through 0 (balanced) to 1 (true) ❚ PUFF helps diagnose lung pr...
View Full Document

This document was uploaded on 03/31/2014 for the course CS 350 at NJIT.

Ask a homework question - tutors are online