A groundbreaking AI device designed to assist stroke survivors in regaining their ability to speak has been developed at the University of Cambridge. The technology features a dual-sensor system: one sensor detects vibrations in the neck, while the other tracks emotional responses through pulse signals. Researchers observed that many patients experienced significant fatigue even with minimal vocal effort, prompting the creation of two AI "assistants" utilizing the GPT-4o mini framework. The first assistant effectively translates neck vibrations into basic, pre-defined phrases, while the second one enhances these phrases into full sentences by considering context, emotional state, and surrounding factors. For instance, a simple phrase like "We go to hospital" can be expanded to "Let's go to the hospital for a doctor's appointment." The integration of this AI technology has led to a remarkable reduction in the word error rate to 4.2% and a sentence error rate of 2.9%. Following the implementation of the second interpretive agent, there was a notable 55% increase in patient satisfaction. Project leader Luigi Occhipinti noted that many individuals gradually regain their speech abilities, indicating that they may not need invasive brain implants, but rather straightforward and accessible rehabilitation tools.
Informational material. 18+.
