Scientists have learned to turn brain impulses into oral speech
The research department of IMD school published its digital competitiveness rating for 2018. The Russian Federation ranked 40 out of 63 states. It is noteworthy that these positions are held at the same level for several years: in 2016, the country took the same place, and in 2017 fell by 2 positions. Experts studied how digital services have penetrated into the daily lives of citizens from 63 countries. It came as no surprise that the leaders were America, several Scandinavian countries, Switzerland and Singapore.
The place of states in the list was determined by 50 indicators divided into 3 groups. Most of all, we scored points for the opportunities offered by the state science in terms of the development and launch of innovative technologies – on this point Russia took 24th place, leaving behind even Luxembourg and Belgium. The biggest problem of our country with the possibility of real use of digital products. In this indicator, Russia is on the 51st place, behind South Africa, Greece and Jordan.
EY conducted another survey, which revealed that the number of fintech users in Russia is 43% in million-plus cities. This is more than the average in 20 most economically stable countries in the world. Services penetrate deeper only in China and India. Russia was approximately on an equal footing with Brazil and Great Britain. But it should be noted that in these countries the indicators were calculated not only for the largest settlements, but in the whole country.
Brain impulses and speech: how is it connected?
According to the journal Science, scientists from the United States, Germany and the Netherlands, used special models on artificial intelligence to reconstruct conversational speech, based on signals from the nervous system. Professionals tracked the activities of individual parts of the brain when the test subjects read aloud, listened to finished recordings, or spoke.
Researcher Nima Mesgarani, representing the University of Colombia, explained that her working group intends to create a system of man-made neurons that can activate and deactivate during the right periods, reproducing words and sentences. These signals are transformed into speech in a special way from different people, which is why a computer should learn to “understand” how to perform such a task. To teach the system such algorithms, it is necessary to train it on the brain impulses received from patients with speech disorder.
In the course of the study, researchers used information obtained from 5 patients with a disorder such as epilepsy. Artificial intelligence analyzed how the crust behaves. It was possible to find out that she is in an active state and when a person is talking and when she is listening. After that, the neural network reproduced the voice from the brain impulses of the subjects. According to the results, the computer performed the task with an amazing accuracy of 75%.
Another group of researchers, including neuroscientists Miguel Angrik of the University of Bremen and Christian Herf of the University of Maastricht, studied the performance of 6 patients who survived brain surgery. Experts recorded through the microphone of their voices, at the time of pronouncing the words. The neural network has interconnected audio recordings and electrode readings. 40% of the information was recognized correctly.
Another team of scientists, this time from the University of California, restored whole sentences based on brain impulses from three patients with epilepsy who read phrases out loud. Some of the proposals neural network correctly identified in 80% of tests.
In the scientific world, the results are considered to be really good, but we understand that they are far from ideal. Developments are still at a very initial level, and require further development. It will take more than one year until the system is debugged so that it can be used on real patients. And then hundreds of thousands of people with paralysis throughout the world will again be able to talk, even if they are using synthetic helpers. It remains to be hoped that the developments will prove to be really interesting to scientific institutes and commercial enterprises all over the world, and the teams of researchers will receive sufficient funding to continue their work.