To determine how accurate (or not) the AI-enabled vocoder had been, the researchers asked participants to listen to the recording and say what they heard.
"We found that people could understand and repeat the sounds about 75 percent of the time, which is well above and beyond any previous attempts," Mesgarani said of the result in a statement.
"The sensitive vocoder and powerful neural networks represented the sounds the patients had originally listened to with surprising accuracy."
The next steps will be to try to work on more complicated sequences – for example, actual sentences such as "I need a glass of water". But while there is clearly some way to go and there are limitations to the technology as it stands at the moment, the implications could be ground-breaking.
"Our voices help connect us to our friends, family and the world around us, which is why losing the power of one's voice due to injury or disease is so devastating," Mesgarani added.
"With today's study, we have a potential way to restore that power. We've shown that, with the right technology, these people's thoughts could be decoded and understood by any listener."