New Mechanical Brain Has One Million Programmable Neurons
We haven't been able to build a true, sentient artificial intelligence yet because we don't understand the human brain enough to build one, but a team led by IBM and Cornell Tech researchers is trying to change all that, as they have built a computer chip that very much mimics a human brain, complete with mechanical "neurons" and "synapses."
This "neurosynaptic" chip is a dramatic departure from previous computing architecture, employing a technique called "cognitive computing."
"We have taken inspiration from the cerebral cortex to design this chip," said IBM chief scientist for brain-inspired computing, Dharmendra Modha. "We want to approximate the anatomy and physiology, the structure and dynamics of the brain, within today's silicon technology. I think that the chip and the associated ecosystem have the potential to transform science, technology, business, government and society."
The chip, dubbed TrueNorth, consists of a network of 1 million programmable neurons, 256,000,000 programmable synapses, 4,096 cores, and 5,400,000,000 transistors. The neurosynaptic cores are arranged in an architecture similar to the brain, and communicate with each other in order to use both memory and intelligence to function. The chip represents the researchers' ambition to create a chip that could behave like a human brain with a similarly low energy requirement. Brains have an incredible amount of computing power while using less energy than a lightbulb. The TrueNorth acts like a brain in the sense that it is "event driven," or it is only active when it needs to be. As a result, it uses significantly less energy (approximately the same amount as a hearing-aid battery) than comparable chips, which are perpetually active.
Their technique is somewhat controversial, as computers of the past have always, at their core, relied on binary systems of 1's and 0's. As a result, computers were mostly capable of advanced "left brain" activity, or any task that could ultimately be boiled down to sequential number-crunching. But consequently, any activity that proved difficult to quantify, such as creativity, was out of the computers' reach. TrueNorth's neurons are analog, or able to react to a variety of different stimuli, imitating the plasticity of the brain. As a result, the TrueNorth may open the door to performing "right brain" activities, and bring computer scientists one step closer to a "holistic computing intelligence." According to Modha, since it mimics the activity of the human brain, it has the ability to react to sights, smells, and other stimuli in order to "learn" from its environment.
However, there is no way to know for sure whether this chip is more powerful than other systems. While the lower energy requirement in particular is a concrete indicator of success, there is no test that effectively compares different computing systems. Other criteria include chip size, speed, and complexity of the tasks it can perform, but the means of judging these criteria and the importance of each factor has not yet been quantified. Furthermore, it would be difficult, if not impossible, to directly compare the TrueNorth with another chip since its means of operation diverge so sharply from other systems.
In the future, IBM endeavors to eventually create a chip with ten billion neurons and one hundred trillion synapses, in order to even more accurately ape a human brain. This technology also may be able to be combined with more traditional computer architecture in order to create a hybrid supercomputer that can perform both "right brain" and "left brain" functions. As it is, the TrueNorth has the potential for many significant applications, including self-driving cars. A chip installed in a car could perform supercomputer functions without necessarily having a Wi-Fi connection. According to Modha, "The sensor becomes the computer. You could have better sensory processors without the connection to Wi-Fi or the cloud. This would allow a self-driving vehicle, for example, to detect problems and deal with them even if its data connection is broken. It can see an accident about to happen."