
Research
Harnessing Chaos: How the Brain Turns Randomness into Robust Memory
AI systems reveal how random fluctuations in brain activity may help stabilize memories
Random noise, such as background hubbub on a phone call, is usually thought of as unwanted interference. Now researchers at Columbia Engineering find the brain may harness unavoidable random fluctuations of its activity to perform useful computations, particularly in tasks relying on memory. These findings not only deepen our understanding of how the brain works, but also may provide a blueprint for building smarter, more resilient technologies, the research team says.
They detailed their findings Jan. 16 in the Proceedings of the National Academy of Sciences.
Brain activity is filled with random, unpredictable variability. "Despite this, the brain excels at cognitive processes like memory and decision-making," says Nuttida Rungratsameetaweemana, an assistant professor of biomedical engineering at Columbia Engineering and a lead researcher on the project. "How does it manage this? Does noise hinder or help?"

Insights from computing
Previous work on brain-imitating artificial intelligence systems known as neural networks suggested that injecting random fluctuations into their activity could actually improve their performance as they learned to perform a task. However, this prior research was conducted on relatively simple neural networks, calling into question how much this effect might actually play a role in real life.
In the new study, Rungratsameetaweemana and her colleagues developed more biologically plausible neural networks that included modules that mimicked excitatory and inhibitory neurons, which respectively increase and suppress overall activity in the brain.
"This research relied on high-computing resources provided by the Systems Intelligence Laboratory, supported by Columbia Engineering," she says. "These computational capabilities allowed us to explore complex neural dynamics in unprecedented detail."
Bring on the noise
When the researchers incorporated random noise into these systems during their training, "we discovered that random noise, often thought of as a challenge to overcome, might actually be essential for how the brain performs critical computations like memory," she says.
Specifically, noise appears to increase the amount of time it takes for inhibitory neuron connections with other neurons to weaken. This slowing effect in turn stabilizes neural patterns of activity related to memories, helping them persist over time.
"Your brain doesn’t block out the noise — it works with it, integrating the randomness to stabilize your memory," Rungratsameetaweemana says. "For me, this project highlights the beauty in complexity and shows how even what seems random has a meaningful role in nature."
"Contrary to the traditional view that random noise in the brain weakens memory, this study reveals that such fluctuations can actually enhance memory stability," says Alex Williams, assistant professor at the Center for Neural Science at New York University and the Center for Computational Neuroscience at the Flatiron Institute in New York, who did not take part in this research. "This study is a great example of how computational models can provide us with counterintuitive insights into long-standing scientific questions such as, 'How do we hang on to memories for many years or even decades? How reliable can our memories be? Can we harness natural properties of neurons, like randomness, to enhance memory performance?'"

Next steps
These new findings may help inspire the design of artificial intelligence systems that are more adaptable and better equipped to handle real-world uncertainty. These may include robots operating in dynamic environments, or health care tools that track subtle, long-term changes in patient data, Rungratsameetaweemana says.
In addition, this work may help scientists understand the role that real-world brain variability may play in human cognition both in health and in diseases such as Alzheimer’s, schizophrenia, and attention-deficit/hyperactivity disorder (ADHD), Rungratsameetaweemana says. She and her colleagues also want to explore the role that noise may play in cognitive tasks beyond memory, such as attention, decision-making, or integrating complex sensory inputs. "These functions, much like memory, rely on the brain’s ability to stabilize and process information amidst inherent variability," she explains.
Furthermore, these findings may also help improve neurostimulation and neuromodulation techniques that use carefully controlled signals to interact with brain activity. "We’re especially excited to explore how such signals might work in harmony with the brain’s natural variability to improve specific computations, such as strengthening memory in patients with cognitive impairments," she adds.
About The Study
Journal: Proceedings of the National Academy of Sciences
Authors: Nuttida Rungratsameetaweemana1,2, Robert Kim2,3, Thiparat Chotibut4 , and Terrence J. Sejnowski2,5,6
Funding/Acknowledgments:
- Department of Biomedical Engineering, Columbia University, New York, NY 10027
- Computational Neurobiology Laboratory, Salk Institute for Biological Studies, La Jolla, CA 92037
- Department of Neurology, Cedars-Sinai Medical Center, Los Angeles, CA 90048
- Department of Physics, Chula Intelligent and Complex Systems, Chulalongkorn University, Bangkok 10330, Thailand
- Institute for Neural Computation, University of California San Diego, La Jolla, CA 92093
- Division of Biological Sciences, University of California San Diego, La Jolla, CA 92093
Acknowledgements: This work was supported by the Swartz Foundation (N.R.), the Kavli Institute for Brain and Mind at the University of California San Diego (N.R.), the National Institute of Biomedical Imaging and Bioengineering R01EB026899-01(T.J.S.), National Institute of Neurological Disorders and Stroke R01NS104368 (T.J.S.), mission funding from the cooperative agreement under the United States Army Research Laboratory W911NF-17-S-0003-03 (N.R.), the funding from the National Research Council of Thailand #1187111 on the fiscal year 2023 (T.C.), and from Thailand Science Research and Innovation Fund Chulalongkorn University (IND66230005) (T.C.).
The authors declare no financial or other conflicts of interest.