Artificial Intelligence is now ready for Tracking Brain activity
The idea of reading and converting what someone is thinking has always been fascinating. Scientists are trying to convert this concept to reality for nearly 20 years. A recent study published in Nature Neuroscience shows how researchers have taken another step, using AI, towards decoding and interpreting the brain activity, while people listened to the sentences, into text.
All the participants utilized in the testing had some history of epileptic seizures and had electrodes placed in their brains. The electrodes helped monitor brain activity. The researchers read aloud 50 fixed sentences and used the electrodes to trace the activity. This data was then provided to the neural network to decode. However, the decoded sentences differed tons.
The data of the brain activity and therefore the voice of the spoken sentences were fed into a machine-learning algorithm, which learned to acknowledge speech. Initially, the results were highly flawed and had mistakes in them. However, because the program learned the errors within the translation decreased drastically. Though having the ability to translate a limited set of sentences may be a big step, it’s still an extended way from decoding the whole English.
Makin in an interview with The Guardian said, “If you are trying to travel outside the [50 sentences used] the decoding gets much worst”
This technology can definitely help people that experience the locked-in syndrome, a medical condition during which the body and most of the facial muscles of the person are paralyzed but the person remains conscious. Also, the program relies only upon brain activity from hearing spoken words and not random thoughts.
Nevertheless, it’s still an enormous achievement for a machine to interpret human speech from brain activity so well after only an hour with each of the four participants.