In fact, scientists from the Okinawa Institute of Technology tried to reproduce the activity of a second of brain life on a computer.
It took 83 thousand processors with the highest possible computing power and 40 minutes of work to get close to the efficiency of a single second of our neurons. Despite the elusive speed at which information flows through our neural network, an experience can take up to six hours to consolidate as a long-term memory.
Memory is much slower than thinking. A fundamental component of our conscious experience, the nervous system, is even faster. Nerves are made up of bundles of nerve fibers, which transmit action potentials to the nervous system, or from the nervous system to muscles. However, not all nerve fibers are the same. Some are larger in diameter, and these generally transmit the potential more quickly, the action potential in the thicker fibers is transmitted at the speed of meters per second, the speed of a Formula 1 car.
In the finest fibers, on the other hand, the transmission speed is half a meter per second, that of a man walking, in terms of sensitive experience, according to a study carried out by neurologist Mary Potter, of the Massachusetts Institute of Technology MIT , we can process an image in just 13 milliseconds.
Sight is the fastest of our senses, while taste is the slowest: it takes milliseconds to identify a flavor, since information related to smell and touch must be linked and linked to the data offered by thermoreceptors as in the case of spicy food.
So a petaflops means that a computer can perform 1,,,,, basic arithmetic operations per second. A petaflops is a unit used to measure the computational performance of floating-point operations, which are those operations that require arithmetic operations with extremely large and small real numbers.
Specifically, a petaflops is , although there are larger units all of them with names of sweets zettaflops or yottaflops Simulating the human brain is so complicated that even one of the most powerful supercomputers on the planet could barely do it after having processed data for a space of forty minutes, with a result equivalent to a single second of brain activity.
Silicon has made spectacular advances in recent years, but from various points of view, the best computer is still between our ears, hence the enormous interest associated with studying its operation. Although experts already know how to face a brain simulation project, the truth is that the processing power to do it is not available, a striking fact if we consider that there are supercomputers that already have the goal of overcoming the barrier of one hundred petaflops.
Take for example the supercomputer K, created by Fujitsu. At the time, K took first place on the TOP list, and thanks to his ten petaflops, he still remains in fourth place. However, when faced with this simulation project, the supercomputer K barely managed to reproduce the equivalent of one second of brain activity … after chewing numbers for forty minutes.
According to the researchers who participated in the simulation, a total of 82, processors with K at their disposal were used to create a network of 1, million nerve cells, connected to Computers equipped with AI programs often learn using so-called neural networks, which have several layers in which lower calculations feed into higher ones.
And this process requires the heavy use of matrices. In addition, Summit has loads of superfast memory RAM available on each of its nodes, where localized calculations can take place. So-called adaptive routing means Summit has some flexibility in how it runs calculations — sort of like networks of brain cells connected to synapses. For instance, just as artificial intelligence programs are being co-opted to learn to pick out cats from images, said Jack Wells, the director of science at ORNL, these AI programs running on Summit could learn to pick out and categorize all kinds of data, ranging from those in biological sciences to physics, such as detections of neutrinos and other particles.
Summit's placement as the "world's fastest" isn't exactly official yet, because the Top list for supercomputer rankings hasn't been updated yet, but according to the Times article, it should get the top slot when the list is updated later this month. Editor's Note: This article was updated to correct the speed of the former "world's fastest supercomputer.
Jeanna is the editor-in-chief of Live Science. China's Sunway TaihuLight supercomputer, until now the world's most powerful machine, has a processing power of 93 petaflops. Summit's initial uses will include areas of astrophysics, cancer research and systems biology.
Supercomputers are typically large, expensive systems featuring tens of thousands of processors designed to carry out specialised calculation-intensive tasks. Summit contains 4, compute servers and has more than 10 petabytes of memory.
0コメント