Artificial intelligence (AI) is taking on a new role in one of science’s most ambitious pursuits: quantum computing. As reported in Nature, researchers in China have used an AI system to rapidly assemble grids of atoms that could one day function as the “brain” of a quantum computer — a leap forward in making these futuristic machines a reality.
The team, led by physicist Jian-Wei Pan of the University of Science and Technology of China, trained their AI model to optimize how rubidium atoms are arranged with laser light. This process, known as building neutral atom arrays, is considered a promising pathway for creating qubits — the quantum counterparts of classical bits. Qubits can exist in a “superposition” of 1 and 0 simultaneously, allowing them to perform calculations far beyond the reach of ordinary computers.
Traditionally, arranging atoms into stable, usable patterns is painstaking and time-consuming. But Pan’s team showed that AI can do it dramatically faster. Their system assembled an array of 2,024 rubidium atoms in just 60 milliseconds — more than ten times quicker than previous efforts without AI, which managed around 800 atoms in one second.
To showcase the system’s speed and precision, the researchers even used it to generate a high-speed atomic animation of Schrödinger’s cat — the famous thought experiment symbolizing quantum uncertainty. The AI-guided lasers shuffled atoms into intricate patterns that became visible when the atoms emitted light.
Experts in the field are impressed. “As the arrays get larger, it becomes increasingly difficult to calculate how to rearrange atoms efficiently,” said Joonhee Choi, a quantum researcher at Stanford University. “This AI-driven method is remarkable.” Mark Saffman, a physicist at the University of Wisconsin–Madison, added that the achievement drew attention across the quantum community.
Despite the progress, building a fully functional quantum computer remains a distant goal. Researchers estimate that such a machine would require about one million qubits operating with minimal errors — far beyond the couple of thousand demonstrated here. Still, Pan remains optimistic. The AI model, he says, can scale to tens of thousands or even hundreds of thousands of atoms without slowing down.
“AI for science is emerging as a powerful paradigm for addressing complex scientific problems,” Pan told Nature. For quantum computing, that could mean faster progress toward machines capable of solving problems that no classical supercomputer can touch.
