The researchers got a way to appeal to the brain through the image directly
Edition Today Lifestyle
Neurons showed 40% higher activity in response to the generated image
Neuroscientists at mit using neural networks and artificial intelligence were able to generate an image which directly affect the brain. One glance at this image, activate separate neurons in the brain that seem to be forced to “run”.
This image spot controls the individual neurons (it is safe)
Using the best available model of visual neural network of the brain, scientists have developed a new method of precise control of individual neurons and their populations in the middle of the network. The information obtained from computer models that allowed them to create images that affect certain neurons in the human brain.
Scientists, in fact, got a way to appeal to the brain through the image “directly”, bypassing the long way to the comprehension of images that reminds a little “brainwashing”.
Main findings — existing design versions of the visual neural networks quite similar to the real thing, so they can be used to monitor the state of the brain. How accurately these models mimic the functioning of the visual cortex — this issue is a very heated debate, says James DiCarlo, head of the Department of brain and cognitive Sciences at MIT, senior author of the study, published on the website of the Institute.
Stupid losers: artificial intelligence failed school tests
That is, no matter how does the computational model of the visual system of the brain — it is important that we are able to use it, that it is quite accurate and that on its basis it is possible to develop new experiments.
How it works
In this study, researchers wanted to test whether their model to perform certain tasks that were not previously shown. In particular, they were wondering whether it is possible to use these models for control of neural activity in the visual cortex of animals.
To achieve this, the researchers first created an accurate map one-to-one neurons in visual area V4 of the brain of the nodes in the computational model. They did this by showing pictures of animals and models and comparing their responses to the same pictures. In area V4 are millions of neurons, but for this study maps were made of subpopulations of 5 — 40 neurons simultaneously.
Artificial intelligence has improved the game DOOM 8 times: video
The researchers then decided to find out whether they can use these predictions to control the activity of individual neurons in the visual cortex. The first type of control, which they called “stretching” includes an image display which displays the activity of a particular neuron far beyond the activity usually called “natural” images, like those used for neural network learning.
The researchers found that the demonstration of such “synthetic” images, which are created by models and do not resemble natural objects, the target neurons reacted as expected. On average, the neurons showed approximately 40% more activity in response to these images than when shown natural images. This kind of control no one has ever reached before.
Robot uprising: smart house refused to let the owner
Scientists have also shown that we can use his model to predict how the neurons of area V4 respond to the synthesized image. Most previous tests of the model used the same type of naturalistic images, in which trained model. Scientists from MIT found that a model with an accuracy of 54% predict how the brain should respond to the synthesized image, and with 90% accuracy predict how the brain will respond to natural images.
Now scientists hope to improve the accuracy of models by allowing them to include the new information that they comprehend, looking at the synthesized image. In the course of this study, this was not used. Simply put, the model will learn on his own generated images.
Artificial intelligence had “vacation” for video games and watching TV
See also the story of how robots can take over the world:
Earlier, the neural network learned to create portraits of non-existent people. Also the artificial intelligence is able to turn a terrible drawings into masterpieces in seconds.
Read the most important and interesting news in our Telegram