Wow, very cool: Scientists train an algorithm to translate brain activity into text with high accuracy and at natural-speech rates. It's early days but this kind of tech could one day be used to create speech prostheses for people unable to talk.
Machine translation of cortical activity to text with an encoder–decoder framework.
Damn - this is awesome | Machine translation of cortical activity to text with an encoder–decoder framework
Makin, Moses, and Chang at used an algorithm inspired by machine translation to decode speech from recorded neural activity, and obtained nearly perfect decoding accuracy for one participant.
¿Te da fiaca tipear? Mira esto. Paper de hoy: Leyendo señales eléctricas del cerebro se puede, en tiempo real, decodificar y tipear lo que está diciendo una persona. Le pegan al ~97% de las palabras. Si tan solo se lograra hacer de manera no invasiva..
brain-activity-to-text is getting better. the failures are particularly entertaining
Machine translation of cortical activity to text with an encoder–decoder framework #deeplearning #neuroscience #brain