In knowledge-based AI, which of the following statements about expert systems is NOT true? Choose the option that least aligns with how expert systems are typically developed, deployed, and supported.

Introduction / Context:
Expert systems are AI applications that capture and apply domain expertise using rules, facts, and inference engines. The question asks which statement is not true of such systems, focusing on development cost, deployment environments, knowledge capture, and long-term support.

Given Data / Assumptions:

• The term “expert system” refers to a knowledge base plus an inference mechanism and an explanation facility.
• Cost, platform requirements, and maintenance are evaluated in a general, historical sense.
• We are selecting the statement that is least accurate as a generalization.

Concept / Approach:
Historically, expert systems demanded intensive knowledge engineering and specialized shells or platforms. While they can run on general-purpose hardware, they were not “usually designed” with the constraint of small general-purpose computers as a defining trait. The other statements reflect widely reported realities: knowledge capture forms the core, projects can be costly, and maintenance (knowledge updates, rule consistency) can be challenging.

Step-by-Step Solution:

Verification / Alternative check:
Texts on knowledge engineering emphasize platform independence and suitability to the problem domain rather than a bias toward small machines. Early deployments even used specialized workstations and later common servers or cloud instances.

Why Other Options Are Wrong:

• Collections of human knowledge: Correct; knowledge acquisition is central.
• Expensive to design: Often correct due to elicitation and validation effort.
• Maintenance difficult: Correct; rule drift and updates are nontrivial.
• None of the above: Incorrect because one statement is indeed not true.

Common Pitfalls:
Confusing “can run on” with “usually designed for”; underestimating ongoing knowledge maintenance as domains evolve.

Final Answer:
Expert systems are usually designed to run on small general-purpose computers