Procedural Generation Using Quantum Computation

Key insights

6 takeaways

01
The quantum blur method encodes grayscale images as quantum state amplitudes, with the probability of each bit string proportional to the brightness of the corresponding pixel.See it on page 3
02
The author reports that a Ludum Dare 44 game used one quantum process run to generate hundreds of textures, taking under 10 seconds on a laptop.See it on page 6
03
The author states that the quantum method cannot beat the complexity of the standard box blur.See it on page 7
04
The paper states that using a simulator is advantageous over a real quantum device, since estimating probabilities on real hardware requires sampling, whereas simulation allows probabilities to be accessed directly.See it on page 3
05
The author argues that quantum artefacts, usually seen as problematic in procedural generation, might prove useful for sci-fi aesthetics.See it on page 7
06
The paper says the method's artefacts come from quantum interference and from squeezing large images into few qubits, where non-neighbouring points can have neighbouring bit strings.See it on page 7

Summary

Wootton (IBM Quantum, IBM Research - Zurich) presents a proof-of-principle method for procedural generation that uses quantum computation. The method is a quantum generalization of a blurring process in which quantum interference produces a blur-like effect on grayscale height-map images. Images are encoded into quantum state amplitudes, with the probability of each bit string proportional to a pixel's brightness, using Qiskit code snippets and a mapping in which neighbouring coordinates map to neighbouring bit strings where possible. The method is aimed at simulators and prototype devices of up to around 20 qubits. The paper describes the method's use in game jams, including the PyWeek 27 art toy Quantograph, a GMTK Game Jam 2019 transition animation, and an island generator for Ludum Dare 44 that produced hundreds of textures from one quantum run taking under 10 seconds on a laptop. The author compares the method to the box blur, which has O(L^2) complexity, and notes that the quantum method cannot beat that complexity. The author also argues that interference artefacts may suit sci-fi aesthetics.