New interface from Facebook reads the thoughts almost without delay

At the moment the technology is already grown to the point when computer algorithms, along with special sensors that can read brain activity and use it as a control. Simply put, is the ability to create neural interfaces to control electronic devices by thought. Sounds fantastic, but if we’re talking about complex devices, they have until today, there was one problem: big delay in the operation of the devices. Why was that? Because scientists were able to develop a neural interface that works with almost no delay.

Who created the neural interface

The development of neural interface are engineers division Facebook called Reality Labs and scientists at the University of California in San Francisco. Reality Labs is still 2017 is engaged in the design and creation of interfaces of type “brain-computer” (brain–computer interface — BCI). Potential applications of the technology is huge. From typing (even without saying it aloud) to control machines and various electronics. And, according to the publication Nature Communications, the transcript and translation of human thoughts into the text and managed the team of experts.

How does the neural interface

The first prototype appeared in early 2019. At the same time and have passed the first stages of testing. In the beginning of the year tool from scientists at the University of California was able to translate human thoughts into speech. And it looked it had great detail and competently. Was he just one “little” drawback: decoding and signal processing took several weeks. Here they made good use of the experience of the guys from Facebook.

They were just the algorithms to work with neural interfaces, but no suitable models in order to optimize your code. The interface looks very simple: on the inner surface of the helmet special is a series of highly sensitive electrodes that read brain activity. Further, in the case involving decryption algorithms.

Of course, for proper operation of the algorithm it is necessary to someone to test. Scientists recruited a group of people who have read brain activity by asking them simple questions like “How are you feeling?” or “What music do you prefer?”. The answer cannot be “invented”. Each question was procured certain answers (total 24 pieces).


Electrodes to read brain activity

In the end, the algorithm has learned to identify people’s answers, the accuracy of 61-76% (depending on response), which is a very good result. Moreover, the delay in recognition was not significant and was measured in several milliseconds. Thanks to new technology, many people will be able to manage with a mind almost anything of the cursor of a computer mouse to a wheelchair and even military aircraft.

But the algorithm at the moment, and disadvantages. Now it is limited to the entered data in advance. That is, an improvised response, the neural network simply do not understand. However, according to scientists, it’s fixable, because to fix this disadvantage, it is sufficient to integrate in the algorithm is self-learning and speech recognition.

Advertisements