A game you cannot win unless you have a quantum computer
Google Quantum AI ran a series of tests on its superconducting chips showing that quantum systems can't be assigned fixed answers in advance, and that this property can be measured and used for computation. The team reports results from up to 105 qubits, and argues the same tests can serve as a benchmark for hardware errors.
- On a 3×3 "magic square" game, the qubits won 98.30% of rounds, beating the 88.89% ceiling any pre-written classical strategy can reach.
- A Bell–Kochen–Specker style test gave χ = 5.590 ± 0.008, above the classical limit of 4 and near the quantum maximum of 6.
- Shallow quantum circuits solved hidden-linear-function problems on grids up to 105 qubits, and the same tasks exposed subtle coherent errors that random-circuit benchmarks miss.
Google Quantum AI's researchers set out to test whether quantum contextuality, the property that measurement outcomes can depend on which other measurements are made alongside them, survives on real hardware with many qubits. Their starting point is a classic puzzle: a 3×3 grid of ±1 entries where rows must multiply to +1 and columns to −1 is impossible to fill. Two players who cannot communicate can at best win 8 of 9 rounds of the resulting "magic square" game using any pre-written table. On Google's Willow superconducting processor, using entangled Bell pairs, the team won 98.30% of 900,000 rounds (0.9830 ± 0.0001), a result they say is statistically significant and beyond what any classical strategy allows.
The team also measured contextuality directly with a Bell–Kochen–Specker inequality, χ = R1 + R2 + R3 − C1 − C2 − C3. Any predetermined-outcome model must satisfy χ ≤ 4, and quantum mechanics allows up to 6. They measured χ = 5.590 ± 0.008, using an ancilla qubit to read out each row and column without disturbing the two main qubits, over 540 measurements repeated 1,000 times. The test is state-independent, so it characterizes the machine itself rather than a particular input state.
To check whether the effect scales, they ran a GHZ parity game with up to 71 players, one per qubit. Classical strategies top out at 75% for three players and fall toward 50% as players are added, while a shared GHZ state wins every round in principle. Noise pulls the quantum win rate below 100% and erodes the lead as qubits are added, but it stays above the classical bound throughout. They also verified genuine entanglement across 45 qubits, more than many earlier GHZ experiments. The authors tie this back to a 2018 theorem showing a hidden-linear-function problem that constant-depth quantum circuits solve and classical constant-depth circuits cannot, and they argue contextuality is the underlying resource that makes the separation possible. The work is published in Science under the title "Quantum-classical separation in bounded-resource tasks arising from measurement contextuality."