Speaking from the Mind! Literally

Issue 8, Volume 113

By Angeline Song 

Cover Image

You knew it would be a death sentence, raising your hand for the first time in a couple of weeks. You had planned what you wanted to say but for whatever reason, your mouth chose not to work. You can’t help but wish your thoughts could be spoken through someone or something else. For decades, scientists around the world have pursued technology that would give people the ability to speak through machines. In 2021, the code was cracked: technology restored a man’s ability to speak after a major injury through a machine. This discovery is a fundamental and new approach to the engineering research done in neuroprosthesis.

In 2019, a study led by neurosurgeon Edward Chang at the University of California San Francisco was conducted to evaluate the potential of recording neural activity from the cerebral cortex. The study utilized electrodes and decoding techniques to assist communication and mobility. A 36-year-old man with anarthria—an inability to speak—due to his stroke, was chosen for his score of 26 out of 30 on the Mini-Mental State Examination for cognitive function. Unfortunately, while he was extremely capable of cognitive function, he was unable to physically express himself through speech, and could only produce vague sounds like grunts and groans due to his facial paralysis. The technology that assisted him with communication was a computer-based typing interface which clocked in at an average of five correct words per minute.

A medical procedure to install a neural implant in the subdural space was required to record and use brain signals in the cortical regions of the brain that control speech. This implant would then be hooked up to a connector that sent signals from the device to a computer. In February 2019, the operation was performed and ran smoothly over the span of three hours. With nothing but time to lose, the study was up for collecting data in April 2019.

Over a strenuous period of 81 weeks, 50 long sessions were conducted to develop and test the new technology. These sessions were separated into two groups: isolated and sentence tasks. For isolated tasks, the subject was visually presented with singular words derived from a list of basic and caregiving terms such as “good” and “water.” The subject would attempt to say the word on the screen and the device would collect data. Similarly, sentence tasks were conducted with the visual presentation; however, this time, the subject was presented with 50 sentences. The subject was told to attempt to speak these words as fast as they comfortably could. The data collected by both tasks were collected in real-time and inputted into an AI to help modify and concentrate what was said—essentially autocorrect for brainwaves. With the data collected, the AI would predict a set of words by processing the neural activity as well as a correlated probability that the subject “said” the words that the signals indicated.

The study proved to be successful. The technology developed bumped the subject’s previously stated average of five correct words per minute to a rate of 18 words per minute with up to an astonishing 93 percent accuracy.

This was a breakthrough and radical approach to a solution. The typical procedure for those who are in the industry of communication neuroprosthetics is through spelling-based approaches, to type out letters one-by-one in text. However, the method used in this study was critically different in that it translated signals that control the muscles of speech. This works to better translate how people normally communicate at a much faster pace. Writing and spelling on the other hand tend to be much slower and more ineffective compared to what is accomplished by mimicking speech.

Chang’s idea to cut out the middleman worked fantastically; the much higher word count and accuracy regarding the subject was both miraculous and a huge advancement regarding the field of neuroprosthesis. This study was the first successful demonstration of a direct translation of thoughts into words and it’s only going further in its ability to give those who need it a chance. Chang is currently supervising further studies working to refine the technology using a wider range of participants affected by major paralysis. While this may seem like a huge step toward a future where everyone will have the ability to speak their mind, the lengthy legal and ethical complications that await should be noted. With both the rise of AI as well as the rapid commercialization of modern technology, we have to be careful and think about the implication this could have on our society.

Nonetheless, a future where everyone on Earth will be able to say what’s in their heads is far away. But this successful study is a major leap and opens many doors for both the disabled and other creative thinkers in the field of neuroprosthesis.