We all know the situation where we literally have our hands full and cannot use our mobile phone to send an important message or check something. Language assistants like Siri or Alexa make life easier, but they are still not perfect in understanding what is meant. With a Brain-Machine-Interface (BMI) these problems can soon be a thing of the past.
While Elon Musk is working on his neuralink and the control of machines with pure thought power is still very much reminiscent of science fiction, there has now been a breakthrough in the field of brain-machine interferences. At the University of California, the team of Joseph Makin has now found a way to translate human brain activity into text.
This is made possible by an algorithm that is able to translate an entire sentence about the brain activities of the test person into a written sentence. Even though there have been such experiments in the past, this has been successful by shifting the focus of attention. The four test subjects already had brain implants for the treatment of seizures. In 30-minute training sessions, these persons read sentences out loud and the AI-supported algorithm recorded and interpreted the brain activities via the implants.
In this way, the artificial intelligence was able to learn which brain region was activated in which order and with which intensity to create a complete sentence. According to the study, the brain-machine interface was able to distinguish vowels and consonants and translate 30-50 sentences, with an error rate equivalent to a professional speech transcription.
The special feature of the California research series was also that they did not focus on individual word fragments, as was the case in the past, but on whole words. This allows the AI to interpret more broadly and make better conclusions about which words occur within a sentence.
The developments of the team around Joseph Makin are still very much in their infancy and a lot of work will be necessary before the machines can interpret the thoughts of their users without errors. But this step in the brain-machine interface area is a big step towards the future.