Scientists have developed a system of artificial intelligence that can transform a person's thoughts into text by analyzing their brain activity.
Researchers from the University of California, San Francisco developed AI to decode up to 250 words in real time from a set of 30 to 50 sentences.
The algorithm was trained using the nerve signals of four women with electrodes implanted in their brains, which were already able to monitor seizures.
Volunteers repeatedly read suggestions out loud, while researchers fed data of the brain in AI artificial intelligence system to remove patterns that could be associated with individual words. The average word error rate in a repeated set was 3%.
"A decade after the first expression of speech by human brain signals, accuracy and speed "remain well below that of natural speech," says a paper detailing the research published this week in the journal Nature Neuroscience.
"Taking into account recent developments in machine translation, we have trained an iterative neural network to encode each sequence along the neural activity in an abstract representation and then to decode this representation, word for word, into an English sentence ”.
The average active vocabulary of an English speaker is estimated to be around 20.000 words, which means the system is far from understanding the normal word.
Researchers aren't sure how well it will work, as artificial intelligence decoders rely on learning the structure of a phrase and use to improve her forecasts. This means that each new word increases the number of possible sentences, thereby reducing overall accuracy.
"Although we would like the decoder to learn and take advantage of its regularity language, it remains to show how many data we would need to expand to a more general form of English, "the document says.
One possibility might be to combine it with other brain-computer interconnection technologies that use different types of implants and algorithms.
The report cited technologies developed by Elon Musk's Neuralink startup and Facebook, who describe cyborg telepathy as the "next big wave in human orientation".
The Royal Society estimates that such interfaces will be an "established option" for the treatment of diseases such as Alzheimer's disease within two decades.
"People could become telepathic to some degree, able to converse not only without speaking but without words," the report said, adding that it extends to more futuristic applications, such as the ability to taste and smell, without the physical experience of feeling.