Artificial intelligence could play a role in assembling the next largest quantum computer.
The world鈥檚 largest known quantum computer has 1180 quantum bits, or qubits, made from extremely cold, electrically neutral atoms. One of the prerequisites for making truly useful quantum computers is to make them bigger 鈥� to build them with more qubits. But the make-up of those qubits also matters, and when they are made from ultracold atoms, quantum computers only run calculations precisely when each atom is at just the right place within a very uniform grid.
at the University of Science and Technology of China and his colleagues have now used AI to assemble 2024 atoms into such an ultracold grid, setting the stage for a new record-breaking quantum computer.
Advertisement
The researchers used atoms of the metal rubidium, which they could cool down and control with laser light. They created an array, or a lattice, of 鈥�optical tweezers鈥� consisting of laser beams, and watched atoms get stuck at some of its points as they cooled them 鈥� precisely tuned light can exert electromagnetic forces on the atoms, and collisions between particles of light and atoms can decrease their energy. However, some spots in the grid remained unoccupied and some atoms misplaced, like a car parked in between two designated parking spots.
One challenge of setting up an ultracold atom grid is knowing the best sequence for moving the tweezers and using them to rearrange the atoms 鈥� but AI can do this easily. The algorithm suggested a sequence that produced the largest such grid yet, and it only took about 60 milliseconds to make it. Remarkably, the researchers estimated that this time would not increase for larger grids.
When they tried to assemble more complicated shapes, like stacking three smaller grids on top of each other vertically or spelling out the letters 鈥淯STC鈥� with the atoms in a pointillist style, the AI-generated rearrangement sequence never took more than 60 milliseconds.
at the Korea Advanced Institute of Science and Technology says the new experiment builds on techniques that his team previously explored with fewer atoms, but it is the use of AI that has made it possible to scale it up to the new, record-breaking number of atoms. This is because the AI divides up the grid into sections and works on many of them in parallel.
The researchers have not used their unprecedentedly large atom array to perform any calculations so far, but Ahn says their new method may make doing that easier.
鈥淓stimating the exact size of atom arrays required for significant breakthroughs in quantum information and simulation remains challenging. However, predictions suggest that arrays containing 1000 to 10,000 atoms could start functioning as quantum processing units,鈥� he says.
Quantum computers made from atoms trapped and controlled by lasers have already demonstrated some of the best performance when it comes to correcting their own errors, and several research teams are working on advancing both the size and complexity of these devices.
arXiv
