What if artificial intelligence could remember things faster or more reliably? A new international study has introduced a novel type of AI memory—not by adding more data, but by using geometry.
A team of researchers from BCAM – Basque Center for Applied Mathematics, Araya Inc., the University of Sussex, and Kyoto University has developed a new class of AI models called Curved Neural Networks. Their findings, published in Nature Communications, show how deforming the "space" in which AI "thinks" can enable explosive memory access—an effect similar to a moment of inspiration in the human brain.
Traditional AI systems rely on relatively simple connections, akin to one-on-one conversations, “but the human brain operates through rich, multifaceted interactions, where many signals influence each other simultaneously,” says Dr. Miguel Aguilera (Ikerbasque Researcher and Junior Leader of la Caixa at BCAM).
To capture this, the team introduced curved geometry into AI models, enabling more complex and realistic memory processes without additional computational overhead.
The team’s Curved Neural Networks revealed three key features:
- Explosive Memory Retrieval: The system can easily “jump” to a stored memory, as if flipping a switch.
- Intelligence through Self-Tuning: The AI automatically adjusts its “focus” when recalling, speeding up its response.
- Fewer Errors or Greater Capacity: A single tuning parameter lets the system balance between memory power and precision.
“These properties are not rigidly programmed, but emerge naturally from the curved geometry of the model,” explains Dr. Pablo A. Morales from Araya Inc.
This discovery could lead to more adaptive, efficient, and interpretable AI systems—a major step beyond today’s powerful yet opaque “black box” models.
“It’s a compelling example of how geometry and physics can drive progress in intelligence, both natural and artificial,” says Dr. Fernando E. Rosas from the University of Sussex. “This work opens new ways of thinking about how brains and machines can store and retrieve information efficiently.”
Prof. Hideaki Shimazaki, Associate Professor at Kyoto University, adds: “What began as a simple idea—using curved geometry in neural networks—evolved into a deeply collaborative journey. The discovery will undoubtedly contribute to the future of AI.”
The research opens new avenues for brain-inspired computing, neuroscience, and even next-generation robotics, offering tools to better understand memory itself—whether in minds or machines.
Link to the article: DOI: s41467-025-61475-w
.png)
