In this photo illustration a virtual friend is seen on the screen of an iPhone on April 30, 2020, in Arlington, Virginia. - "It's so good to hear your voice." "I was worried about you." "What would you like to do today?" This might sound like ordinary banter between friends. But in these cases, the "friend" speaking was a chatbot created with artificial intelligence. OLIVIER DOULIERY/AFP via Getty Images
Alexis "Lexi" Bogan used to have a vibrant voice, often heard singing along to her favorite tunes or engaging in lively conversations with friends. However, her voice changed dramatically last summer after doctors removed a life-threatening tumor near her brain.
Reclaiming Lexi's Voice
In August, surgeons successfully removed a life-threatening tumor from the rear of her brain. Following the removal of the breathing tube a month later, Bogan encountered difficulty swallowing and experienced challenges in speaking, struggling to communicate a simple "hi" to her parents.
Despite months of dedicated rehabilitation, her speech remains impaired, posing challenges for her friends, family, and even strangers in understanding her communication attempts. But Lexi transformed in April: she regained a semblance of her original voice.
However, it wasn't her natural voice but an artificial intelligence-generated replica accessible via a smartphone app. Based on a brief recording from her teenage years, the AI synthesized a remarkably authentic voice.
Harnessing AI to Restore Speech
With just a few taps on her phone, Associated Press reported that she can now command the app to vocalize anything she desires, utilizing her synthetic yet remarkably lifelike voice. The snippet was originally from a cooking demonstration video for a school project.
Bogan, along with a group of medical professionals from Rhode Island's Lifespan hospital network, are optimistic about a potential application that outweighs the potential risks.
Bogan represents one of the pioneering cases, and the only one with her specific condition, to successfully restore a lost voice using OpenAI's latest Voice Engine technology.
Her medical team deliberately supplied the AI system with a short 15-second recording for training. Despite some imperfections due to background noises from cooking, this short clip met OpenAI's needs, a significant improvement over older methods that required longer recordings.
The doctors knew getting useful information from short recordings was important, especially for patients without digital voice traces. These brief recordings, like short voicemails, could be important for future use.
Now, she relies on the app approximately 40 times daily, offering feedback to potentially assist forthcoming patients. One of her initial trials involved addressing the children at the preschool, where she serves as a teaching aide.
She's used the app in retail stores to inquire about item locations. Additionally, it facilitated reconnection with her father and simplified ordering fast food.
Bogan's medical team has initiated the process of replicating the voices of other willing patients in Rhode Island, with aspirations to extend this technology to medical facilities worldwide. OpenAI has expressed cautious progress in extending the use of the Voice Engine, which remains unreleased to the public.
Dr. Rohaid Ali, a neurosurgery resident at Brown University's medical school and Rhode Island Hospital, expressed optimism about Lexi's potential to facilitate future technological developments.
He highlighted the significant impact this innovation could have on millions of individuals affected by conditions such as strokes, throat cancer, or neurogenerative diseases.
