FIT1061_MOC

📘 FIT1061: Introduction to Artificial Intelligence

Map of Content

Index for FIT1061 Introduction to Artificial Intelligence — the unit refuses “AI” as a definition and teaches algorithms in three blocks: how machines search (W2–5), decide under uncertainty (W6–9), and learn (W10–12). Every algorithm is drilled against three questions: how does it work, when does it fail, who does it affect.

📊 Assessment Map

• Format ➔ tiered portfolio, submitted via OnTrack; Pass tasks P1–P10 plus optional Credit / Distinction / HD extensions. You pick the tier and may raise or lower it any time before W13.
• The ladder ➔ Pass 50–59 implement each algorithm with scaffolding, engage with the milestone (~8 h/wk) · Credit 60–69 Pass + 4 Credit extensions, less scaffolding (~10–12 h) · Distinction 70–79 Credit + 4 tasks incl. A*, perceptron training, adversarial tournament (~12 h) · HD 80+ Distinction + one comprehensive project: define → implement → interview (~12–15 h).
• Task outcomes ➔ Complete (locked in) or Fix and Resubmit ➔ unlimited good-faith attempts until W13; a tutor may pause resubmissions that stop engaging with feedback.
• ⚠️ HURDLE — Week 10 test ➔ 30–40 min, paper, no code, no devices; short answer + hand-computation; covers Pass tasks 2–9 only (W1 excluded). Resit W12, supplementary W14 capped at Pass. The unit cannot be passed without it.
• What the hurdle actually tests ➔ hand-traces of search algorithms · hand-computed Bayes’ rule · confusion-matrix arithmetic · gradient-descent step calculations ➔ every by-hand sub-task already done in the portfolio, so hand-execution is the revision priority all semester.
• Kialo debates ➔ W2 (Deep Blue), W7 (COMPAS), W11 (face recognition); each is prompt → response → counter-argument → reply and feeds that week’s reflective response. Citing optional in W2, expected in W7 and W11.
• AI use ➔ cited at every tier (tool, prompt, date). Permission scales: Pass concept questions only · Credit debug your own code · Distinction pair-programming · HD critical evaluation expected. Never submit code or an argument you cannot trace in discussion.

📅 Knowledge Index

Week 1 — Milestone: ELIZA to ChatGPT

• Turing Test (Imitation Game) → Parent Framework: FIT1061_MOC (the criterion, and the three cases that test it)
• The AI Effect (Defining AI) → Parent Framework: FIT1061_MOC (why the label slips — the ten-system boundary spectrum)
• AI Algorithm Blocks (Search, Uncertainty, Learning) → Parent Framework: FIT1061_MOC (the unit map: question → recipe → flagship → weeks)

Week 2 — Milestone: Deep Blue vs Kasparov

• Search Problem Formulation → Parent Framework: AI Algorithm Blocks (Search, Uncertainty, Learning) (the four parts + why $b^d$ forbids exhaustion)
• Uninformed Search (BFS and DFS) → Parent Framework: Search Problem Formulation (hurdle hand-trace)
• Deep Blue vs Kasparov → Parent Framework: Turing Test (Imitation Game) (Kialo motion + reflective response)

🎯 Learning Outcomes

• W1 ➔
  • state Turing’s substitution: an unanswerable question ⟹ a text-only imitation game
  • reject a “Turing test passed” claim by naming its format conditions
  • explain the AI effect and why the label tracks novelty, not capability
  • classify a system by machinery — does it learn, does it plan — not by label
  • map any unit algorithm to its block, recipe and flagship system
  • apply the refrain: how does it work · when does it fail · who does it affect
• W2 ➔
  • formulate any problem as states, actions, successor function and goal test
  • compute states at depth $d$ from a branching factor $b$ and argue why exhaustion fails
  • hand-trace BFS, writing the frontier, visited set and came_from at every step
  • reconstruct a path backwards from came_from and report nodes_expanded
  • convert BFS to DFS by the single frontier swap (Queue (ADT) ➔ Stack (ADT))
  • state BFS’s shortest-path guarantee with both its preconditions
  • argue the Kialo motion by naming the machinery, not the scoreboard