Japanese neural network is taught to "read minds"


2018-01-16 15:00:07




1Like 0Dislike


Japanese neural network is taught to

It Seems that computers can already read our minds. The Google autocomplete suggested friends in Facebook and targeted advertising that POPs up in your browser, sometimes make you think: how does it work? We are slowly but surely moving in the direction of computers that read our minds, and a new study from Kyoto, Japan, was clearly a step in that direction.

A Team of scientists from Kyoto University used a deep neural network to read and interpret people's thoughts. Sounds unbelievable? In fact, such do not for the first time. The difference is that previous methods — and results — were easier, they had deconstructed the image-based and pixel, and fundamental circuits. New technology, called "deep image reconstruction", is outside of the binary pixels and provides researchers with the ability to decode image with multiple layers of color and texture.

"Our brain processes visual information by extracting hierarchical traits of different levels or components of different complexity," says Yukiyasu Kamitani, one of the scientists involved in the study. "These neural networks or model AI can be used as a proxy for the hierarchical structure of the human brain".

The Study took place 10 months. In the study, three people studied images of three different categories: natural phenomena (animals or people), artificial geometric shapes and letters of the alphabet.

Brain Activity was measured by observers either while viewing the images, or after. To measure brain activity after viewing images, people are simply asked to think about the images that they were shown.

The Recorded activity is then fed to the neural network, which is "decrypted" data and used them to generate their own interpretations of the thoughts of people.

In humans (and, indeed, all mammals) visual cortex located at the back of the brain, the occipital lobe, which is above the cerebellum. Activity in visual cortex measured using functional magnetic resonance imaging (fMRI), which was transformed into the hierarchical features with deep neural networks.

Starting with a random image, the network is iteratively optimizes the pixel values of this image. Features of introduction to neural network image become similar to the features decoded from brain activity.

Importantly, the model scientists were trained using only natural images (people or nature), but I learned to remodel a man-made form. This means that the model is really "generated" image, starting from brain activity, and did not compare this activity with the existing examples.

Unsurprisingly, the model had difficulty with decoding the brain activity when people were asked to remember the image, and it is easier when they are directly viewed these images. Our brains can't remember all the details seen images, so the memories will always be vague.

The Reconstructed images from the study retain some resemblance to the original images seen by the participants, but for the most part appear to be as minimally detailed blots. However, the accuracy of the technology will only improve, and with it will expand and possible applications.

Imagine the "instant art", when you can create a work of art, just presenting it in my head. Or if the AI will record your brain activity while you sleep, and then re-create your dreams for analysis? Last year completely paralyzed patients to communicate with their families through brain-computer interface.

There are a lot of possible applications of the model used in the Kyoto study. But brain-computer interfaces can also recreate the horrific images, if we learn to properly interpret brain activity. The cost of mistakes in case of wrong reading the thoughts may be too high.

With all this Japanese company is not alone in its efforts to create a mind-reading AI. Elon Musk founded Neuralink to create brain-computer interface between people and computers. Kernel is working on creating chips that can read and write the neural code. Mind reading is slowly developing.


Comments (0)

This article has no comment, be the first!

Add comment

Related News

#news high tech | Issue 226

#news high tech | Issue 226

Every Monday in the new issue of «News high-tech» we summarize the previous week, talking about some of the most important events, the key discoveries and inventions.

#2018 CES | Samsung showed a prototype of bendable smartphone

#2018 CES | Samsung showed a prototype of bendable smartphone

at the close of the exhibition CES-2018, traditionally held in Las Vegas, the South Korean company Samsung has decided to do a very narrow number of invited participants covered by the cameras flashing, the prototype demonstration of his secret Galaxy X, rumors of which have long circulated in the Network.

French designers have developed a airplane bird

French designers have developed a airplane bird

As you know, aircraft use wing design, "borrowed" from the birds.