Neuralink has accumulated over 50,000 hours of neural activity data from clinical trial participants. This dataset serves as the foundation for a new base model designed for brain-computer interfaces, utilizing self-supervised neural encoders to process noisy brain signals into stable representations.
The company reports significant improvements in decoder stability, allowing systems to remain functional for weeks rather than days. Calibration time for some users has decreased from 10 minutes daily to 10 minutes per week. Additionally, Neuralink introduced a cursor control feature that allows the pointer to jump directly to a target.
Neuralink achieved a new cursor control speed record of 11.32 bits per second, which the company states surpasses the performance of most individuals without motor impairments. Future development goals include aggregating data across participants, enabling zero-shot calibration, and launching an API for motor cortex integration.
