Module · What learning is✅Knowledge checkCheck: one word, and no model at all6 questions · Pass at 60%1. “Left” and “miss” are both vague. Why is only one of them a different KIND of problem?“Miss” is less accurate than “left”“Left” says which way to move; “miss” says nothing about direction, so nothing can be corrected“Miss” arrives more slowlyThey are the same kind — “miss” is simply the vaguest label2. Learning from hit-or-miss took 517 arrows against 12 for the exact miss. What changed between the two runs?Only how much each answer was allowed to containThe bow, which was harder to shoot in the second runThe target, which was smallerThe algorithm, which was less efficient3. With a 20 cm gold ring the robot never learned the habit at all, inside 4,000 arrows. Why?A bigger target is harder to aim atIt hits from almost anywhere, so every answer is “hit” and the word carries nothingThe wobble grows with the size of the ringIt ran out of arrows before it could explore4. Which of these is the same failure as that 20 cm ring?A test suite that fails intermittentlyA sales target that has been hit every quarter for three yearsA model that scores well in training and badly in productionA dataset with more columns than rows5. Measuring the habit once and subtracting it settled at 5.2 cm; the learner, given all sixty arrows, settled at 5.3. What is the honest conclusion?Machine learning is usually unnecessaryThe learner was badly configuredOn a problem that holds still, the simplest thing that could work usually doesSixty arrows was too few for the learner6. After the wind changed, the fixed correction averaged 11.6 cm for the rest of the run. What is the actual risk it illustrates?Rules are slower to compute than modelsIt is not slightly worse — it is wrong permanently, and nothing in it will ever noticeIt needed to be re-measured over more than ten arrowsThe wind was too strong for any method to handleCheck answers← Back to What learning isNext topic →
