Quiz: The Quantum Programming Career

DataField.Dev

Quiz: The Quantum Programming Career

20 questions — the last set. Answers with explanations follow.

1. Across the book's six head-to-head comparisons against tuned classical baselines, quantum methods won:

(a) all six
(b) four
(c) one
(d) zero

2. Which role does the field employ fewest people in?

(a) control software
(b) compilers and transpilers
(c) algorithm theory
(d) platform engineering

3. Chapter 39 found the dominant cost of a quantum computation is:

(a) the qubits
(b) the queue
(c) circuit depth
(d) error mitigation

4. ★ Chapter 15 measured a circuit with zero T gates at 450 physical qubits. With one T gate:

(a) 451
(b) 900
(c) 2,882
(d) unchanged

5. ★ Chapter 19 cut an 8-bit oracle from 26,978 T gates to:

(a) 13,489
(b) 55
(c) 0
(d) 26,978 — ancillas did not help

6. ★ Chapter 31 measured dynamical decoupling as:

(a) a large improvement
(b) no change
(c) significantly worse — XX at −0.0053 ± 0.0012
(d) unmeasurable

7. ★ Chapter 28 found approximation_degree=0.9 produced:

(a) a 10% speedup
(b) zero two-qubit gates — the optimization deleted the circuit
(c) an error
(d) identical output

8. ★★ Chapter 30 measured quoted two-qubit errors on a single chip spanning:

(a) 0.00750 to 0.07205 — a factor of 9.6
(b) all within 5%
(c) 0 to 1
(d) identical across qubits

9. ★★ How many times did this book publish a conclusion from too small a measurement?

(a) never
(b) once
(c) three times
(d) seven times

10. ★★★ Chapter 39's instance of that error was different because:

(b) it had 24 samples but the wrong system — a 4-qubit test circuit
(a) it had only one sample
(c) it was never corrected
(d) it was not an error

11. ★★ is_tuned_baseline("full CI") returns False because:

(a) full CI is not a real method
(b) nobody runs it past ~20 orbitals, so beating it proves nothing (Ch.36 §36.6)
(c) it is too slow to implement
(d) it is quantum, not classical

12. ★★ significant() raises on a standard error of exactly zero because:

(a) division by zero
(b) zero uncertainty means it was never estimated, not that it is absent
(c) the API requires it
(d) the difference must be zero too

13. ★★ Chapter 33 measured two gaps: +0.0202 ± 0.0170 and +0.0626 ± 0.0067. Respectively they are:

(a) both significant
(b) neither significant
(c) not significant, and significant at nine standard errors
(d) significant, and not significant

14. ★★★ audit_claim() returns failures rather than a score because:

(a) scores are hard to compute
(b) a claim that cannot name its baseline is unevaluable, not fractionally correct
(c) users prefer lists
(d) it is faster

15. ★★★ Applying the checklist to Chapter 36's own central result gives:

(a) all eight checks pass
(b) seven of eight — it fails on sample size, having measured one molecule at one geometry
(c) it fails all eight
(d) the checklist cannot be applied to it

16. ★★ The three areas §40.5 says survive the book's scrutiny are:

(a) finance, optimization, and near-term ML
(b) error correction, simulation of quantum systems, and learning from quantum data
(c) cryptography, search, and factoring
(d) none

17. ★★ Chapter 31 opened by noting that qiskit.pulse was:

(a) renamed
(b) removed in Qiskit 2.0
(c) deprecated but functional
(d) newly added

18. ★★★ Chapter 39 priced the same 31.2 seconds of device time at:

(a) $50, $7,432, or $185,542
(b) always $50
(c) $185,542 everywhere
(d) free

19. ★★★ The book's summary of the evidence is:

(a) quantum computing does not work
(b) quantum computing already beats classical methods
(c) the techniques work, the physics is real, and the comparisons lose — both halves are true
(d) the evidence is inconclusive

20. ★★★ The single lesson §40.7 says to keep:

(a) always use the latest SDK
(b) the number that is easy to get is not the number that answers the question, and the easy number is almost always the flattering one
(c) quantum computers are exponentially faster
(d) always measure fidelity

Answers

1. (d) zero. Five accuracy comparisons produced one exact tie and four classical wins; Chapter 37's ten MaxCut instances gave QAOA 0 wins, Goemans–Williamson 6, and 4 ties.

2. (c) algorithm theory — the role people imagine when they hear "quantum job," and the smallest by headcount. The largest categories are classical engineering roles where quantum literacy is the differentiator.

3. (b) the queue. A 4,096-shot Bell job occupies the device for 6.92 ms; at a five-minute queue that is 2.31 × 10⁻⁵ utilization. Improving it is a scheduling problem, not a physics problem.

4. (c) 2,882 — a 6.4× cliff from a single non-Clifford gate. This is the clearest single number in the book for why fault tolerance is the real problem.

5. (b) 55. Ancillas are the trade worth making, and one of the book's cleanest wins.

6. (c) significantly worse. A standard technique, applied correctly, measured at −0.0053 ± 0.0012 — 4.4 standard errors in the wrong direction. Negative results with uncertainty attached are the most trustworthy measurements in this book, because nobody wanted them.

7. (b) zero two-qubit gates. The "optimization" deleted the circuit. It is not a bug — the setting does what it says — and it is a perfect illustration of a number that looks like success.

8. (a) a factor of 9.6. Which is why Chapter 30 §30.5 refuses to quote a fidelity without naming its statistic.

9. (d) seven times — Chapters 27, 28, 33, 34, 37, 38, and 39. Each was corrected in print, and the list is in §40.4.

10. (b) 24 samples, wrong system. A 4-qubit test circuit fits the coupling map without routing, so every transpiler seed finds the same layout. It is a true measurement of a false general claim. A measurement can be too small in more than one dimension.

11. (b). Chapter 21 §21.7's lesson with a molecule attached: benchmarking against the method nobody uses measures nothing.

12. (b). Returning True there would be the book's single-sample error with extra steps.

13. (c). Same chapter, same experiment. One gap survives its uncertainty and one does not, and the difference decides what may be said.

14. (b). Chapter 36 passing seven of eight and a vendor claim passing one of eight are not points on a scale — Chapter 36 has a specific, nameable, fixable gap; the vendor claim cannot be evaluated at all. Averaging them would be the last instance of the error the book spent forty chapters on.

15. (b). The 9,870,104× ratio came from one molecule at one geometry. The finding is almost certainly robust, and almost certainly robust is a different claim from measured. A checklist that never fails your own work is a checklist you are not applying.

16. (b). And the applications with the loudest commercial claims — optimization, finance, near-term ML — are the ones the book measured most carefully and found furthest behind.

17. (b) removed in Qiskit 2.0, taking add_calibration, backend.defaults, and instruction_schedule_map with it. The API you learn will be deprecated; the reasoning will not.

18. (a). Fifty dollars, or one hundred and eighty-five thousand, for identical computation — depending only on how the provider counts. You are not paying for device time.

19. (c). Both halves are true, and a person who can hold both is more useful to this field than one who believes either alone. The techniques are real and elegant; the classical baseline is decades of optimization on hardware that costs nothing to run.

20. (b). And the reason the easy number is the flattering one is not a coincidence: the flattering number is the one that stops the search.

This is the last quiz in the book. If you got question 19 right and can explain why both halves matter, you have the thing the field actually needs.