Page 152 - CyberSecurityEssentialsEnglish

Chapter Glossary
TABLE 3-17: GLOSSARY OF ESSENTIAL TERMS

Review Questions
1. Why is the question “what is the best AI tool?” insufficient for making a decision?

2. What is the difference among generation, retrieval, extraction, computation, and
execution? Give one example of each.
3. How can two products that use the same model give different results?

4. Explain the difference among temperature, top-p, and top-k, and why none of them adds
new knowledge.
5. What must be recorded in order to repeat a generative experiment reproducibly?

6. When is retrieval more appropriate than fine-tuning, and when might you need both?
7. What is the difference between open-weight and open source AI?

8. Why is a general benchmark insufficient for evaluating a tool in your own environment?
9. What are two possible biases when a language model is used as a judge, and how do you
reduce them?

10. How do you build a fair local test for Arabic support?
11. What is the difference between a mandatory gate and a weighted factor in the matching
framework?

12. Briefly design a comparison experiment for two tools on a task you repeat at work, and
define the success criterion and the impact of failure.
13. What does the reader actually use when they open a tool: the model, the application, or

the whole system, and why does the distinction matter?
14. Which medium in your own work is most likely to be misread, and what would you test
to find out?

Practical Exercises

↻ Exercise 1: Dissecting a task

Goal: Turn a general request into a specification you can test.
Required: One task you want a tool for. Allow about 20 minutes.
Steps: Write the verb that names the operation. Then define the input, the output, the
user, the impact of failure, and the evidence that would show success. Remove every
word that cannot be checked.
Output: A half-page task card.

Assessment criterion: The description contains no vague word such as professional
without a checkable property behind it.

↻ Exercise 2: Testing variation
Goal: Understand the effect of probabilistic generation on repeated runs.
Required: One prompt and a tool you can run repeatedly. Allow about 25 minutes.

151