Stop 1 of 12
The course in one picture
Three pillars hold the whole subject up: quantum computing — the computing of the future; machine learning — the technology already carrying today's world; and quantum machine learning where the two meet. The journey runs from qubits to quantum advantage to ML, in that order, because the intersection only makes sense once both sides are solid.
Read the full lesson → Stop 2 of 12
Bits, qubits, and the question of how
A classical bit is always 0 or 1. A qubit can exist in a blend of 0 and 1 at the same time. The natural question — how is this possible? — has a one-word answer: superposition, the strangest and most central idea in all of quantum computing.
Read the full lesson → Stop 3 of 12
Superposition and many qubits
Physicists write quantum states in Dirac's ket notation: |0⟩, |1⟩, and blends like |ψ⟩ = α|0⟩ + β|1⟩. Add a second qubit and you need 4 numbers to describe the pair; a third needs 8. The description doubles every time — 2ⁿ numbers for n qubits. Is it not very powerful?
Read the full lesson → Stop 4 of 12
Bits vs qubits: how scaling changes
Two punchlines capture the quantum edge. What takes linear effort classically can take only logarithmic effort quantumly — and what takes exponential effort classically can become merely linear. This is the root of quantum advantage: exponential compression of how much must be computed.
Read the full lesson → Stop 5 of 12
Measurement: the curse of the qubit
Measure a qubit and it collapses to a single state — superposition gone. And measurement means any intrusion into the qubit's privacy: a scientist observing it, a stray atom bumping it, a photon leaking information about it. One measurement, one classical answer. Algorithms must ask clever questions.
Read the full lesson → Stop 6 of 12
What can become a qubit?
A qubit is any controllable two-level quantum system. An electron's ground vs excited energy level. An electron's spin, up or down. A photon's polarization, horizontal or vertical. And many others — trapped ions, superconducting circuits, neutral atoms, quantum dots — each a different engineering trade-off.
Read the full lesson → Stop 7 of 12
Three fundamental properties
Superposition — the blend that makes quantum physics interesting for computing. Interference — steering a circuit's wave-like amplitudes so wrong answers cancel and the right one survives. Entanglement — the 'spooky' correlation between qubits that expands what a quantum system can represent. And no, it cannot send messages or travel through time.
Read the full lesson → Stop 8 of 12
Why quantum algorithms win
Plot computational steps against problem size and the classical curve for hard problems explodes exponentially while quantum curves stay low. The flagship: factoring numbers takes super-polynomial effort classically — the hardness all modern encryption rests on — while Shor's algorithm factors in polynomial time. Merit is measured in scaling, not clock speed.
Read the full lesson → Stop 9 of 12
Three circles of quantum algorithms
Picture three overlapping circles: practical utility, few number of qubits, and quantum advantage. The center — all three at once — is the desired area. Shor's algorithm has utility and advantage but needs many qubits; the 2019 supremacy experiment had advantage on few qubits but no utility yet. The center is the field's goal.
Read the full lesson → Stop 10 of 12
Quantum computers today: the NISQ era
We live in the NISQ era — Noisy Intermediate-Scale Quantum. Noisy: error correction is not yet in use. Intermediate: qubit counts are not huge. Noise drags qubits into decoherence, many designs need near-absolute-zero cooling, and raw qubit count says little — quantum volume and CLOPS measure quality and speed instead.
Read the full lesson → Stop 11 of 12
Machine learning in five minutes
Machine learning predicts from data — mathematically, function approximation plus optimization. Four problem types: classification (cat or not, spam or not), regression (tomorrow's weather, next quarter's sales), clustering (find the natural groups), and recommendation (what you'll watch next). Supervised when answers are labeled, unsupervised when they aren't.
Read the full lesson → Stop 12 of 12
Quantum machine learning: four flavors
Sort every QML idea into a 2×2 box — is the data classical or quantum, is the processor classical or quantum? CC: quantum-inspired classical algorithms (the Tang dequantization story). QC: ML improving quantum hardware. CQ: quantum speedups for ML, bottlenecked by loading data in. QQ: quantum data on quantum processors — no loading problem at all.
Read the full lesson →