What Machine Learning Actually Is

Machine Learning Fundamentals

Chapter 1 · What Machine Learning Actually Is

ds1-1 drew a line through the data science workflow — Collect → Clean → Explore → Model → Communicate — and marked "Model" out of scope for that entire course. ds1-10's own closing scope note named exactly why: "that's ml1's entire job." This is that course, starting exactly at that boundary.

Learned Rules vs. Hand-Coded Rules

historyai2-6 covered the 1980s expert-systems boom — DENDRAL, MYCIN, XCON/R1 — real, commercially deployed systems that made genuinely useful decisions (MYCIN recommended antibiotic treatments) using large sets of IF...THEN rules that human domain experts wrote out by hand, one rule at a time, encoding their own explicit knowledge directly into code.

A machine learning model does the opposite. Instead of a human writing "if mileage is high and the car is old, then price is low," an algorithm is shown many real examples — actual cars, with their actual mileage, age, and price — and it discovers a pattern connecting them on its own, without anyone ever writing that rule down explicitly. The "knowledge" ends up encoded in a set of learned numbers (later chapters call these coefficients or weights), not in a rule a person could read and have written themselves.

	Expert Systems (historyai2-6)	Machine Learning (this course)
Where the rules come from	A human expert, written by hand	Discovered from data by an algorithm
What's stored	Explicit, readable IF/THEN rules	Learned numeric weights/coefficients
How it improves	An expert edits the rules directly	More/better data, retraining
Real example	MYCIN's antibiotic rules	ml1-3's own learned price model

Not a claim that one replaced the other

Expert systems didn't fail because hand-coded rules are a bad idea — historyai2-6 covers real, honest reasons for their 1980s decline (brittleness, the knowledge-acquisition bottleneck of interviewing experts one rule at a time). Machine learning solves a specific version of that bottleneck — getting rules from data instead of from a slow, manual interview process — not every problem hand-coded rules are good at. ml1-10's own Fuzzy Logic chapter revisits this exact era from a different angle.

Three Branches

This course's own focus

Supervised Learning

Learning from labeled examples — each one has a known, correct answer. A used car's real price; whether an employee actually left. ml1-3 through ml1-8.

This course's own focus

Unsupervised Learning

Learning from data with no labels at all — finding structure nobody told the algorithm to look for. ml1-9.

Named, not covered

Reinforcement Learning

Learning through trial and error, guided by rewards and penalties rather than labeled examples at all — genuinely out of scope for this course. Named here so its absence later isn't mistaken for an oversight.

This Course's Own Two Running Case Studies

Rather than invented, disconnected examples, this course deliberately reuses two real datasets ds1 already built and genuinely didn't finish with:

ds1-9's own used-car listings — its own Step 7 hypothesis, "does mileage predict price more strongly than year does?", was raised and explicitly left untested. ml1-3 tests it for real.
ds1-10's own employee-attrition table — its own Step 7 hypothesis about salary, and its own unresolved department/salary confounding question, were both raised and explicitly left untested. ml1-5 tests both for real.

This course's own throughline

ds1 spent two entire chapters teaching how to form good, well-motivated questions from data — and then deliberately stopped short of answering any of them, naming that boundary explicitly every time. This course exists specifically to cross that boundary, on the exact same two datasets, so the payoff is concrete rather than abstract.

The General Workflow — This Course's Own Roadmap

Step	Covered in
Split data honestly (before training anything)	ml1-2
Train a regression model (continuous prediction)	ml1-3, ml1-4
Train a classification model (category prediction)	ml1-5, ml1-6
Try a structurally different model family	ml1-7
Diagnose and fix a model that's learned the wrong thing	ml1-8
Find structure with no labels at all	ml1-9
A classical, rule-adjacent alternative approach	ml1-10
Put it all together on real data	ml1-11 (capstone)

Hands-On Exercises

Exercise 1

Using this chapter's own compare-table, explain the fundamental difference between how MYCIN's own rules came to exist and how a machine learning model's own "rules" come to exist, and explain why this chapter says ML didn't replace expert systems so much as solve a specific bottleneck.

📄 View solution

Exercise 2

Explain why this chapter names reinforcement learning explicitly even though the course doesn't cover it, and explain what would go wrong for a reader if it were simply left unmentioned.

📄 View solution

Exercise 3

Explain, using this chapter's own warn-box, what specifically makes this course's own two case studies different from a typical "invented example" — what did ds1 already do that this course is now building on?

📄 View solution

Chapter 1 Quick Reference

This course starts exactly at ds1-1's own Explore/Model boundary — ds1-10 named this course as "Model"'s own job
Machine learning: rules discovered from data, vs. expert systems (historyai2-6): rules hand-written by a human expert
Supervised (labeled data) and unsupervised (no labels) are this course's own two branches; reinforcement learning is named but out of scope
This course's own throughline: closing ds1-9's and ds1-10's own deliberately unanswered hypotheses, for real, on the same datasets
Next chapter: The Train/Test/Validation Split & Why It Exists