Scientists develop AI that can read mind and convert thoughts into Text
A computer that can read your mind and turn your mind into speech has been developed and tested by scientists. Researchers say they’ve developed a device that can convert brain messages directly into the text—a positive development toward a “speech prosthesis” that will essentially allow you to think text directly into a machine. Researchers at the University of California, San Francisco, created the AI to decode up to 250 words in real-time from a range of 30 to 50 sentences. “We’re not there yet,” the University of California researcher Joseph Makin said to The Guardian, “but we think this could be the basis of a speech prosthesis.”
Scientists have developed an artificial intelligence system that can translate a person’s thoughts into text by analyzing their brain activity. The creation is being heralded as ‘life-changing’ for those unable to communicate, such as stroke victims.
The algorithm was trained using the neuronal impulses of four women with electrodes embedded in their brains that were still in a position to control epileptic seizures. Volunteers regularly read sentences aloud as researchers fed brain data to the AI to unpack patterns that could be correlated with individual words. The average word error rate in the replicated range was as low as 3%.
Electrical engineering professor Dr. Nima Mesgarani at Columbia University, New York, hopes that the lives of thousands of people affected will improve. He said, ‘Our voices help bind us with our friends, families and the world around us, which is why losing the control of one’s voice due to accident or illness is so devastating.
AI Power
Makin and his colleagues identified the new method in a paper published in Nature Neuroscience. A computer that can read your mind and turn your mind into speech has been developed and tested by scientists. Creation is being heralded as ‘life-changing’ for those unable to speak, such as victims of stroke.
A computer that can read the mind and turn THOUGHTS into speech is created by scientists and brings new hope to the victims of stroke. Essentially, they hired four epilepsy patients who already previously had electrode arrays implanted in their brains for monitoring purposes. They instructed the participants to read sentences again and again when gathering neural information—a dataset they used to teach an algorithm to interpret signals even though the subjects were not reading out loud. The average active vocabulary of an English speaker is estimated to be around 20,000 words, meaning the system is a long way from being able to understand regular speech.
Locked In
The method does have weaknesses—most importantly that it performs well on sentences on which it has already been educated. But its accuracy is impressive, the Guardian notes, with an error rate of just around 3%, significantly lower than that of human transcribers.
The groundbreaking technology incorporates the power of voice synthesizers and AI (artificial intelligence) to transform brain activity into intelligible words. It’s based on the same technologies used by Amazon Echo and Apple Siri. It brings hope to people paralyzed following a stroke or an injury who interact with friends and loved ones-despite being unable to speak.
According to the newspaper, the developers behind the machine expect that it will be used as the basis for a communication device for individuals who are unable to type or communicate due to a syndrome or related disorders.
Professor Mesgarani said: ‘The responsive vocoder and strong neural networks reflected the sounds that patients had initially listened to with surprising precision.’ Researchers are now preparing to test more complex words and phrases next, and they want to perform the same experiments on brain impulses produced when a person talks or imagines speaking.
