Google develops the quantum computer Sycamore./Courtesy of Google Quantum AI

Quantum science technology seems like a superhero, as if anything is possible. It possesses information processing speeds that defy common sense and a level of security that makes existing security technologies appear ineffective. However, just as superheroes in movies have sidekicks to assist them, quantum science technology also needs helpers.

Controlling ultra-cold atoms or correcting complex computational errors, and efficiently generating quantum entanglement can sometimes be like solving puzzles that quantum science technology struggles to tackle alone. The crucial helper in solving these complex problems is artificial intelligence (AI).

AI simplifies the complex issues surrounding quantum technology and proposes optimal solutions. Chinese researchers implemented the world's largest ultra-cold quantum lattice in just 60 milliseconds (ms) using AI, and Google DeepMind succeeded in reducing errors in quantum computers by 6% through its AI model 'AlphaQubit'. German researchers opened new possibilities by developing an AI algorithm that simplifies the process of generating quantum entanglement.

◇AI-created world's largest quantum lattice

The performance of a quantum computer is determined by the number of quantum bits (qubits), the smallest unit of information used in quantum information systems. These qubits are made from electrically neutral atoms at extremely low temperatures. However, to develop a useful quantum computer, it is vital not only to increase the number of qubits but also to consider how the qubits are arranged. The quantum computer made from ultra-cold atoms must have each atom in a precise location within a highly uniform lattice for accurate calculations.

Professor Jian-Wei Pan and his research team from the University of Science and Technology of China (USTC) successfully arranged 2,024 atoms in an ultra-cold lattice using AI. The team used laser light to cool and control rubidium atoms. In this process, they created a lattice of laser beams called 'optical tweezers' to trap the atoms at specific points. The laser light applies electromagnetic force to the atoms, and the collisions between light and atoms reduce energy, fixing the atoms at lattice points.

The research team utilized AI algorithms to find the optimal sequence for moving the tweezers to rearrange the atoms. Following the order suggested by AI, the team was able to complete the lattice in just 60 ms with 2,024 arranged atoms. The researchers believe that since AI divides the lattice into several regions and performs tasks in parallel to maximize efficiency, the time needed for assembly will not significantly increase even if the size of the lattice grows larger.

Professor An Jae-wook from the Korea Advanced Institute of Science and Technology (KAIST) noted, 'This research is based on previous small-scale atom arrangement experiments, but AI has enabled us to expand to a record scale.' He added, 'Although we have not yet used this massive atomic lattice for actual calculations, this technology will make computations easier.'

The image represents a quantum bit (qubit)./Courtesy of DeepMind

◇Quantum computers 6% more accurate with AI assistance

Quantum computers are geniuses at ultra-fast calculations but are also error-prone 'klutzes.' Google DeepMind developed an AI model, 'AlphaQubit,' that can correct errors in quantum computers more effectively than before.

Quantum computers perform calculations using qubits. However, qubits are highly sensitive to external factors such as heat or radiation, making them prone to errors. To address this, researchers bundle multiple qubits together to form a single logical qubit. In this process, some qubits are used for calculations, while others serve the role of detecting errors. The task of interpreting the information from error-detecting qubits and determining correction methods is referred to as 'decoding,' which is a critical step that determines the error correction capability of quantum computers and their practical utility.

AlphaQubit was developed based on transformer neural networks. This technique is used in large language models like the protein structure prediction AI 'AlphaFold,' which won the Nobel Prize last year, and ChatGPT. The research team first trained the AI model using simulation data, and then refined the model with actual data collected from Google's Sycamore quantum computing chip. The experimental results showed that AlphaQubit, tested on small-scale qubits from the Sycamore chip, produced 6% fewer errors than the existing optimal algorithm, tensor networks.

Scott Aaronson, a professor at the University of Texas at Austin, commented, 'This is a very exciting development,' adding, 'Quickly correcting errors in fault-tolerant quantum computation has tested the limits of classical computers. It is becoming increasingly clear that utilizing machine learning for problems involving optimization or uncertainty can yield better results.'

◇Quantum entanglement made easy with AI

Quantum entanglement refers to a state where two particles have an 'invisible connection.' When in this entangled state, measuring the state of one particle automatically determines the same state for the other particle. In quantum communication, such entanglement is used to securely transmit information. However, the process of creating entanglement is generally very complex and time-consuming.

Researchers at the Max Planck Institute for the Science of Light in Germany discovered a simpler way to create quantum entanglement with the help of AI. The AI algorithm they developed, 'PyTheus,' proposes a new method for entangling two particles.

PyTheus is like an 'quantum manual' created by AI. When you input the desired quantum state, it provides instructions on how to achieve it. The research team experimented with the simple procedure suggested by PyTheus instead of existing complex methods, successfully entangling the particles. This new method does not require complicated additional equipment, making the experiment relatively straightforward.

This method can make the process of entangling particles that are far apart easier and faster, which could significantly help in building communication networks like quantum internet that ensure high security. The research team continues to study how this method can be best utilized in various quantum communication technologies.

References

arXiv(2024), DOI: https://doi.org/10.48550/arXiv.2412.14647

Nature(2024), DOI: https://doi.org/10.1038/s41586-024-08148-8

Physical Review Letters(2024), DOI: https://doi.org/10.1103/PhysRevLett.133.233601