Tutorial 7: Artificial Intelligence and Generative AI

COMP200 · Unit 4 · Section 7

Objectives

AI systems perform tasks associated with perception, prediction, planning, or language. Machine learning infers patterns from data; generative models produce new outputs from learned distributions. An output can be fluent and still be wrong, biased, private, or unsupported.

data -> training -> model -> prompt or input -> inference -> evaluation

Exercises

  1. Compare classification with text generation.
  2. Design a verification workflow for generated content.
  3. Identify data and bias risks.

Self-check

  1. What is inference?
  2. Why evaluate training data?
  3. Why verify generated output?

Self-Check Quiz

1. Does fluent output guarantee accuracy?

AnswerNo. Generative systems can produce plausible but false or unsupported content.

2. What is training data used for?

AnswerIt provides examples from which a model learns patterns or parameters.

Homework

  1. Evaluate an AI-assisted workflow.
  2. Specify human verification, privacy, and attribution controls.
  3. Discuss one appropriate and inappropriate use.
Sample answerUse AI for drafts or alternatives, verify factual claims against authoritative sources, avoid confidential inputs, disclose material assistance, check bias and accessibility, and keep a human accountable for the final decision.