Mind the Electrode
Mind reading is a concept rarely discussed outside the realm of science fiction and fantasy, however recent discoveries in machine learning are beginning to challenge that notion.
Five engineers from Columbia university recently conducted a study where they managed to convert intracranial signals into sound, meaning, they could more or less listen to someone’s thoughts.
The team accomplished this by gathering a group of patients and inserted depth electrodes into the auditory cortexes of their brains, and then proceeded to play sound clips which their patients would repeat in their minds, such as counting to 10 or relaying a story.
Then those neurological signals would be stored and decoded by a vocoder; which is basically a synthesizer that analyzes audio and turns it into code. However, they couldn’t just use any vocoder, they needed to use a form of machine learning called “deep learning” which would train the vocoder to recognize specific neural patterns as words. Their results far surpassed any prior studies ability to convert thoughts into sound, with over 75% of the words recorded being intelligible.
Though I’m sure this study conjures fanciful images of this technology being used for interrogations and criminal trials, it was actually designed as a means to aid vocally impaired people, like those who suffer from ALS.
Sources:
https://zuckermaninstitute.columbia.edu/columbia-engineers-translate-brain-signals-directly-speech
