MemuL8R: Enhanced Recollective Satiety Facilitated by Inexpensive Neural Prosthesis

Lawrence K. Bothria, Mehdi Jafarzadeh, Terence Keaton, Jennifer A. LaMonte, Arturo A. Garcia, Jacob D. Deringer*

Abstract: Damage to parts of the brain responsible for memory formation and recall is a major factor in loss of quality of life (QoL) in all-cause dementia. Here, we demonstrate a novel, computationally inexpensive neurocognitive prosthesis (NCP) based on an intermediate-scale convolutional neural network (CNN) connected to the Neuralink™ 3.3 direct deep brain interface (D²BI). Our system achieves efficient recollective satiety by near-optimal simulation of engram retrieval cues (ERCs), actuated through a minimally-invasive bioderived soft-electrode array (BDSEA). Individualised training is stored as privacy-preserving deltas applied to a bottlenose dolphin-derived hippocampus and anterior cingulate cortex (ACC) metabrain model, with ≥90% Gary sympathy typically achieved in less than 6 h training time. This result is possible because the system does not facilitate memory recall but rather the perceptual cues associated with successful recollection of a memory. Metabrain models may be hosted off-premises, admit multiple simultaneous users, and are arbitrarily scalable within the constraint of network latency. Concerns regarding data privacy, cognitive security, carbon footprint, and identity damage are substantially obviated in this paradigm. We report that this system is significantly effective in reducing distress in a clinical cohort (N=29) comprised of early-stage Alzheimer's disease (AD) patients. Patients evaluated as having 23% lower (95% CI 18–28%) scores on the Civita–Morrison Amnestic Distress Inventory (CMADI) after one-month followup and all participants reported satisfaction with the system.