Medicine has made enormous progress in recent years and today, we are able to do things that were unthinkable only a few decades ago and the work on artificial intelligence has contributed greatly to this. We remember that only a few months ago, an experiment led by the BrainGate consortium, allowed paralyzed individuals to operate a tablet with a brain implant and earlier last year, MIT researchers developed a device called AlterEgo that could transcribe words that the user verbalizes internally.
All these advances give hope to people who are paralyzed and unable to speak. It turns out that the signals of what these people would like to say are hidden in their brains and the direct decoding of these signals has not yet been done, but significant progress has been made. Indeed, three research teams have recently made progress in converting data generated by electrodes surgically placed on the brain into computer-generated speech. In these experiments, researchers monitored parts of patients’ brains as they read aloud, spoke quietly or listened to recordings.
Nima Mesgarani, a computer scientist at Columbia University, said: “We are trying to define the pattern of neurons that turn on and off at different times and infer the sound of speech. »
The way these signals translate speech sounds varies from person to person, so computer models must be adapted for each person. These models work best with extremely accurate data, which requires an opening of the skull.
Result: 40% of the words generated by computer were understandable.