A quantum bit, or qubit, is like a coin spinning on a table, able to be both heads and tails at once. For the computer to know what came up, you have to stop the coin and look. Previously, this stopping was done with a delay: data was read out, taken to a regular computer for processing, and only then was the result known. During that time, the qubit could change its state, and the answer would be spoiled.
Engineers found a way to look at the coin without stopping it. They built a small board with artificial intelligence that sits right next to the quantum device. The neural network on this board is trained to recognize the qubit's state almost instantly, like an experienced cashier who can tell heads from tails just by the sound of a falling coin.
This approach allows checking qubits right in the middle of computations. That's important for quantum computers that correct their errors on the fly. Now experiments run faster, and results are more accurate.
A regular computer takes as long to recognize a qubit's state as it does to read a short sentence. The new board does it in the time of a single breath.