• Researchers develop AI that converts brain activity to text

    Date:1 April 2020 Author: ilhaam Bardien Tags:,

    If you’ve ever wished that reading minds was a possibility, you may be in luck. Scientists have just taken a step forward in making this sci-fi skill a reality.

    A new paper, published in Nature Neuroscience, suggests that it is possible to interpret brain activity using artificial intelligence (AI). “We are not there yet but we think this could be the basis of a speech prothesis,” said co-author of the research, Joseph Makin of the University of California to The Guardian.

    Essentially, this study will assist those who cannot speak or write and to communicate.

    The study used 4 participants who all have a history with epileptic seizures and who already had electrodes implanted in their brains to monitor the activity. Each participant was required to read 50 different sentences aloud. Simple phrases such as “Tina Turner is a pop singer” were used.

    As they read the sentences, the researchers monitored their neural activity. The data was inserted into a machine-learning algorithm (an AI system). This system converted the brain activity for each spoken sentence into a string of numbers. The number sequences are compared to the recorded sounds and are converted them to words.

    At first, the sentencing being produced by the system were gibberish, but as it learned which words generally follow each other, it became much more accurate. The scientists were able to produce text with only 3% error.

    Unfortunately, the technology requires a person to hear spoken word, not only to have thoughts. This is not the only complication, however. “If you try to go outside of the [50 sentences used] the decoding gets much worse,” said Makin.

    It’s to be expected that the technology would not be perfect immediately, but this is a significant step towards the end goal.

    Dr Mahnaz Arvaneh, a brain machine interfaces expert at Shefflied University stressed the importance of ethics in this study. “We are still very, very far away from the point that machines can read our minds. But it doesn’t mean that we should not think about it and we should not plan about it, she told The Guardian.

    Image: Pixabay