Foundations: What It Is, and What It Is Not

Probability, but with amplitudes

Swap non-negative reals for signed, complex amplitudes

~12 minLesson 3 of 24Written by Claude Fable 5

In the opener you watched two contributions of $+ \frac{1}{2}$ and $- \frac{1}{2}$ cancel to nothing. That is impossible in ordinary probability, where every chance is a non-negative number that can only pile up. So quantum theory is not a new kind of magic bolted onto probability - it is the same bookkeeping with exactly one line rewritten. This lesson rewrites that line.

Not just ordinary uncertainty

A tempting thought: maybe a qubit is secretly already

0

1

, and the "probability" only measures our ignorance, exactly like a shuffled card we have not turned over yet. If that were true, the two ways of reaching an answer could only ever add up. But they cancel. Hidden classical ignorance can never make a possible outcome less likely by adding another route to it. The cancellation is the hard proof that amplitudes are a genuinely different object, not a bookkeeping trick for what we do not know.^[1]

Classical probability, in one line

Line up the outcomes of a classical random process - say the eight answers from the opener - and give each a probability. Two rules govern the list: every probability is a non-negative real number, and they sum to one. If you have met probability distributions, this is just the statement that a distribution's entries add to 1.

p_{i} \geq 0, i \sum p_{i} = 1 A classical distribution: non-negative reals whose sum is 1 (the L1 norm).

That "sum equals one" condition is the of the probability list: add the entries and you get 1. This is the entire classical model. Quantum theory keeps the shape of it and changes two things at once.

The one rule change

Replace each probability with an , and change the thing that must equal one. Amplitudes are complex numbers (for now, read that as "numbers with a sign," and the next lesson makes them geometric). What must sum to one is not the amplitudes themselves but their squared magnitudes:

a_{i} \in C, i \sum ∣ a_{i} ∣^{2} = 1 A quantum state: complex amplitudes whose squared magnitudes sum to 1 (the L2 norm).

Now the normalized quantity is the , the ordinary Euclidean length you met in the dot product: a quantum state is a unit vector, measured with the same length that gives a right triangle its hypotenuse. To get back an honest probability you take an amplitude's squared magnitude - the :

p_{i} = ∣ a_{i} ∣^{2} The Born rule turns an amplitude back into a probability.

The squaring is what reconciles the two pictures. Because the probability is the amplitude squared, the amplitude is free to be negative or complex while the probability stays a clean non-negative number that sums to one. And that freedom is the whole game: two amplitudes $+ a$ and $- a$ for the same outcome add to zero before you square, so the outcome vanishes. Play with it - set the two paths to $+ 0.5$ and $- 0.5$ , then switch between the classical and quantum rules.

One outcome, reached two ways. Slide each path, then switch the rule. Classical adds probabilities; quantum adds amplitudes first, so opposite signs can wipe the outcome out.

Each path's contribution to the one outcome. Below the axis means a negative amplitude.

path A

path B

path A amplitude = 0.50path B amplitude = -0.50

Quantum rule: add the amplitudes, then square. (0.50 + -0.50)² = 0.00

outcome probability = 0.00(the other rule would give 0.50)

Interference: the two paths carried opposite signs, so their amplitudes cancelled. An outcome reachable two separate ways just dropped toward impossible. No sum of classical probabilities can ever do that.

Under the classical rule the outcome's probability is $0. 5^{2} + 0. 5^{2} = 0.5$ no matter the signs - two routes only ever help. Under the quantum rule you add the amplitudes first, $(+ 0.5) + (- 0.5) = 0$ , and only then square, giving $0$ . That is , and it exists only because the L2 rule lets amplitudes carry a sign.

import numpy as np

# two paths that both lead to the same outcome
a, b = 0.5, -0.5

classical = a**2 + b**2          # add probabilities:  0.25 + 0.25 = 0.5
quantum   = abs(a + b) ** 2      # add amplitudes, then square:  |0|^2 = 0.0

print(classical)  # 0.5  -> two routes can only help
print(quantum)    # 0.0  -> opposite amplitudes cancel: interference

The same outcome, combined two ways: classical adds probabilities, quantum adds amplitudes then squares.

Lock it in

Quantum theory is classical probability with one line changed: non-negative reals summing to 1 become complex amplitudes whose squared magnitudes sum to 1.
The normalization moves from the L1 norm (plain sum) to the L2 norm (Euclidean length), so states are unit vectors.
The Born rule bridges back: the probability of an outcome is its amplitude squared, which keeps probabilities non-negative even though amplitudes are signed.
Signed amplitudes are exactly what let two contributions cancel, so quantum randomness is not hidden classical ignorance.

Check yourself

Which norm does each model keep equal to 1?

Two paths reach the same outcome with amplitudes +a and -a. What is the probability of that outcome?

Recall: state the single rule change (the norm and the number field) and the formula that recovers a probability.

One structural swap, one bridge back. Try to state it, then check.

Match each object to its defining property.

drop here

a non-negative real, normalized in the L1 norm

drop here

a complex number, normalized in the L2 norm

drop here

probability equals the amplitude's squared magnitude

drop here

amplitudes cancel or reinforce before squaring

Primary source

Scott Aaronson - Quantum Computing Since Democritus, Lecture 9: Quantum

The source of the "quantum mechanics is generalized probability" framing used throughout this course. Aaronson derives the whole theory by asking what happens if you switch the 1-norm for the 2-norm and allow negative (then complex) numbers. Read it once and the rest of the course clicks into place.

Sources