--- title: "Reading and rewriting memory: China's early bets on memory brain-computer interfaces" date: 2026-10-08 category: Brain–Computer Interface site: NeuroAI canonical: https://neuroai.site/a/na-brain-memory-cognition language: en --- # Reading and rewriting memory: China's early bets on memory brain-computer interfaces > From an Alzheimer's mouse study at the Chinese Academy of Sciences to invasive memory-decoding research, Chinese labs are moving from reading brain signals to actively modulating memory and cognition. A mouse with Alzheimer's disease cannot remember which arm of a maze holds the reward. After a few minutes of soft electrical pulses at 40 Hz, it starts to. The pulses did not add a memory — they appear to have helped the brain find one again. ## Memory is the next frontier after movement The first wave of brain-computer interface (脑机接口) stories was about motion: a paralyzed person moving a cursor, a robotic arm, a wheelchair. The second wave is quieter and harder — cognition itself. Can we read a memory as it forms, and can we nudge a failing one back online? Chinese labs are now publishing work on both halves of that question. ## The 40 Hz mouse In August 2025, the Aerospace Information Research Institute of the Chinese Academy of Sciences (中国科学院空天信息创新研究院) reported a bidirectional flexible brain-computer interface (脑机接口) microelectrode — the team calls it NeuroRevive-FlexChip — that does two jobs at once: it records neural electrical activity *and* senses dopamine, the chemical tied to reward and learning. The experiment targeted the hippocampus, the brain's memory hub. In Alzheimer's model mice, 40 Hz stimulation there triggered a brief rise in hippocampal dopamine and shifted neuron firing from burst to sustained patterns. The result, published in *Nature Communications*: treated mice performed better on a Y-maze task, showed less Aβ42 plaque buildup, and had remodeled immune cells in the brain. In plain terms, targeted stimulation seemed to improve cognition and ease a hallmark of the disease — at least in mice. The chip's real trick is doing two jobs at once: it records the electrical spikes *and* chemically senses dopamine on the same tiny implant, so measurement and stimulation share one device instead of fighting for space in the brain. ## Decoding memory from deep-brain signals Reading memory is a separate problem from stimulating it. At the Institute of Psychology of the Chinese Academy of Sciences (中国科学院心理研究所), researchers posted a 2025 pre-print on *invasive* brain-computer interface (脑机接口) applications for memory: decoding and modulation. Their focus is episodic memory — the spatial and emotional information of a moment — pulled from local field potentials in deep brain regions using machine learning. The claim is that movement, environment boundaries, spatial position, and emotional valence can all be parsed from those signals, and that the same neural features can be used to target interventions. The authors are blunt about the bottlenecks: individual differences between patients, unstable electrodes, and adaptive algorithms that still do not generalize well. Their proposed fix is a closed-loop paradigm that fuses dynamic network models with flexible electrodes. ## Sleep as a memory workshop Memory is not only about stimulation during waking hours. At Beijing Normal University (北京师范大学), professor Yunzhe Liu (柳昀哲) presented work showing that a 90-minute nap builds better logical reasoning than quiet video-watching — especially for connecting distant ideas. Brain magnetoencephalography (MEG) data showed the brain replaying key items during non-rapid-eye-movement sleep, and the strength of that replay predicted later performance. The thread connecting these labs: memory is a *process* spread across time and circuits, not a file you open. A brain-computer interface (脑机接口) that wants to help memory has to read it, time the intervention, and close the loop — whether that is a 40 Hz pulse, a deep-brain decode, or a sleep-stage trigger. ## Why China is leaning in Cognitive decline is a demographic time bomb. China's own researchers cite projections that Alzheimer's patients in the country could pass 30 million by 2050. Drugs alone have not solved it, and traditional screening tools are blunt. That pressure makes memory and cognition a natural target for brain-computer interface (脑机接口) research that couples diagnosis with modulation, not just observation. ## Honest limitations Every result above is early-stage. The 40 Hz finding is in *Alzheimer's model mice*, not humans; mouse memory recovery does not guarantee human therapy. The psychology-institute work is a pre-print on invasive decoding — conceptually powerful, clinically unproven, and it explicitly lists electrode stability and individual variation as unsolved. The sleep study is about healthy reasoning, not disease. None of these is an approved treatment, and the "enhancement versus therapy" line is ethically loaded: a tool that restores memory can drift toward one that augments it. We have not verified any human memory-modulation trial from these specific groups, and readers should not read these as available care. ## What readers can do now - Families worried about memory loss should rely on established clinical pathways (cognitive scales, imaging, specialist referral) — research like the 40 Hz mouse study is not a treatment you can request yet. - Students and lifelong learners can borrow the *non-invasive* lesson: consistent sleep, especially a midday nap, reliably helps reasoning and memory consolidation per the Beijing Normal sleep work. - Researchers should follow the *Nature Communications* paper and the psychology-institute pre-print as the two clearest Chinese anchors for where memory brain-computer interface (脑机接口) is heading: decode first, then close the loop. --- Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-brain-memory-cognition Free to quote with attribution and a link to the original.