Imagine: you are standing in front of a pile of ripe melons in the grocery store or market. You grab one, get on the scales and take home, dreaming of how I will cut it and eat — completely without any effort. Now answer the question: how do you know which melon to take, I could not take a whole bunch? A simple question, right?
To answer this question, neuroscientists asked for tips in the sphere of… games. Just as the best computer games run on engines that simulate the physics of the world, probably, we have similar neural networks, which allow us to gracefully navigate in the real world. Instead of processing every detail, game engines offer shortcuts to simulate actions in the game space so that it looked nice and was not unreasonable, and the player could react on the fly. The brain can work this way.
In a paper published in the academic journal proceedings of the National Academy of Sciences (PNAS), a group from Johns Hopkins University in collaboration with Harvard and Massachusetts universities, provided a first description of what may be a separate physics simulator, located in our brain.
Let’s say you suddenly need to Dodge the ricocheted ball from the branches of a tree or off the road, seeing it on the cow. Let’s slow down these reactions: you see an object — color, shape, material, shadow — aware of your position and movement in three-dimensional space, plan further actions, and then allow his body to make an accurate movement, which will allow to avoid collision. Obviously, in your head sickly occur parallel processes.
“I see in the understanding of the physical scene is highly integrative process in which different sources of information from different areas of the brain work together,” says lead author Dr. Jason Fisher.
In other words, it is unlikely that the brain will be a separate piece of fabric, devoted exclusively to physical modeling that is waiting at rest to emergencies. Find such a region is unlikely.
Instead, the authors ask a more meaningful question is: can the background of the continuous calculations in the brain to show individual region, the activity of which increases when the brain collides with the physical world?
To separate the interfering background, the scientists carefully planned four experiments, each of which is based on the previous one.
In first trial, twelve of the participants lying in the fMRI machine and watched live as the towers of yellow and blue blocks rapidly overlap each other, ready to topple over at any moment. The volunteers had to predict when the blocks will start to fall or when the yellow or blue blocks will be more. Unlike the first task, for which it was necessary to apply the so-called physical intuition, the second task was based only on color, serving as a control.
After the first tests, scientists have narrowed the range of their searches to eleven brain regions that showed stronger activity when the volunteers were predicting the fall of the towers, unlike attempts to guess the color.
However, are these “area of interest” characteristic of physical intuition or to simulate future events in General?
And this should help the second experiment. Volunteers watched video clips of two interacting balls, physically — for example, colliding with each other — or social — when one chased the other, like two children playing. After a few seconds one of the balls disappeared and the volunteers had to guess where he is moving on. Obviously, this requires the modeling. But here’s the caveat: although the latter scenario also requires a prediction, it relies heavily on social assumptions, not physical.
By selecting only those areas of the brain that are mainly involved in this game of physics, and comparing the results with the first experiment, the scientists were able to reduce the number of candidates to five. Then they checked their guesses by the third experiment, when the volunteers passively watching a video with different kinds of physico-centered materials: stationary face, the vase falling, colliding cars. While their brain activity was scanned in fMRI.
The result: the more the video was physically active content, the more intensified the five regions, even when the volunteers were actively trying to predict what will happen next.
Finally, scientists have wondered how these regions of the brain suitable for physical modeling? As it turned out, not so much: some of the areas activated when the volunteers were faced with the difficult task of memorizing information, which have nothing to do with physics.
Perhaps it has something to do with the difficulty of predicting the behavior of the world around us. Previous studies have shown that when faced with the challenges of the brain usually connects a series of interconnected areas, asking them a variety of tasks. Activation of this network helps to solve complex problems, such as the planning of the next movement and the use of new sophisticated tools.
Does it make sense to talk about “physics engine”, if this area of the brain is also making other things?
Scientists point to the graphics card (GPU) as an analogy.
“The highly parallel architecture of the GPU was originally driven by the requirements of graphic-intensive computing applications,” explain the authors, “but since then, GPUs have become necessary for other applications like computer vision, deep learning, neural networks, and the approximate physical simulation in real-time in computer games”.
GPUs active in performing all of these tasks, that is, at the same time and physical engines, and graphics, and engines computer vision and so on. In a sense, the brain network is defined in this study as a biological GPU — it supports the physics processing, but embedded in a larger network that is responsible for the execution of other complex tasks, such as planning actions.
Why, then, the physical handling and planning are so closely linked in the brain?
“We believe that this is due to the fact that the children study the physical model of the world, honing their motor skills, playing with objects to learn about their behavior. In addition, to reach out and grab something in the right place with the right amount of force necessary physical insight in real time,” explains Fisher.
This kind of understanding is disrupted during disease called apraxia, when people have difficulty performing certain movements due to brain damage.
“Many cases of apraxia are the result of damage to the same brain regions that we identified as important to the movement,” said Fisher. Further studies should test whether breaks temporarily disable the physics engine of the brain processing the physical information and does not signs of apraxia. If so, a broken physics engine, you may be able to fix it.
Studying how areas of the brain involved in the physics engine interact with each other, we could even build robots, which will lie perfect understanding of physics.
“In this study we tested only a small subset of all possible types of processing physics,” the authors write. For example, whether our brains respond to interactive liquid two waves colliding with each other — just like two colliding solid object.
If the physical engine is the brain really like a game physics simulator, it can be specialized for a small number of liquids or solid materials, for example. Consequently, we could detect more types of physical engines buried in our giant neural networks.
We might be able to provide robots many physical engines running continuously as a video game. If the robots can quickly and efficiently simulate the outcomes of physical scenarios, as a people, they could anticipate what will happen before it happens. They could interact with the outside world is not the worst of people.