A recent study has shown that scientists can create a network combining atoms and photons that might improve how artificial intelligence (AI) systems store and recall information. This network, known as a quantum-optical spin glass, functions like an associative memory—a type of AI that can retrieve complete memories from partial or incomplete information. Published in the journal Science, the research highlights that this new system can store and recall more information than traditional AI networks of the same size. Additionally, the network displayed short-term plasticity, a property similar to how connections between neurons in the human brain change when learning new information. "We can now make neural networks at the atomic level, and they adjust themselves in a way that is somewhat similar to how we believe our brains learn," said Benjamin Lev, the study's senior author and a professor of physics at Stanford University. A spin glass is a type of material where atoms are in a state of frustration, meaning their magnetic orientations are not aligned in a uniform way. In a typical magnet, all the atoms' magnetic orientations—called spins—point in the same direction, either up or down. In a spin glass, however, these spins are disordered, much like how atoms in everyday glass are arranged in a seemingly solid structure but have a fluid-like disorder at the atomic level. This concept was first explored in the 1980s by physicist John Hopfield, who used the properties of spin glasses to create a mathematical model for memory storage. Hopfield’s model showed that a network of spins could store and recall information, much like how the brain recalls memories. He was awarded the 2024 Nobel Prize in Physics for this work, which has influenced how AI systems like ChatGPT process language. However, Hopfield’s model has its limits. When the network contains too many memories, it becomes too complex and fails to recall information accurately. This is because the system transitions into a spin glass state, where the energy landscape becomes too cluttered. In this new study, Lev and his team found a way to use a spin glass state as an effective associative memory by constructing it from atoms and photons. Using quantum-optical effects—where atoms absorb and emit light—the network could still recall memories even in this frustrated state, surpassing the limitations of the Hopfield network. This quantum-optical spin glass was first experimentally realized in 2025, but this study took the research further by using it as an associative memory. The researchers used laser tweezers to create an array of atomic gases in an optical cavity, which acts as a trap for light between two curved mirrors. These gases, called Bose-Einstein condensates, consist of thousands of atoms in a special quantum state that behave like a single super atom. Photons were sent into the cavity, bouncing between the mirrors and connecting the atoms in a way similar to how synapses connect neurons in the brain. This process allowed the network to store and recall memories more efficiently. The quantum-optical spin glass demonstrated a memory capacity up to seven times greater than a traditional Hopfield network with the same number of spins. The study also showed that the photons could influence the atoms to create a type of plasticity, enabling the network to adapt and change, much like the human brain. The research is in its early stages, and the team is working on expanding the system to include more spins and quantum entanglement. "This teaches us a little bit more about how physical systems can compute, not just with the classical laws of physics, but also with quantum laws," Lev said. While the study opens the door to more efficient AI hardware, the researchers are equally interested in understanding how natural systems compute. "It's great to shoot for these applications, but we're also doing this because we want to know more about how nature works," Lev added.