Neurosynaptic Chip

“Holistic Computing Intelligence” they call it at IBM, the combination of traditional linear, analytical processing, combined with Neurosynaptic chips adressing senses and pattern recognition. You see, old-fashioned computers use an insane amount of power for tasks that the human brain for instance can do almost effortlessly, like recognizing someone’s face, even when it’s not exactly like you’ve seen it before. Our brains can instantly ‘connect the dots’ if you will, spot the parts that are recognizable, throw away the parts that aren’t and then draw the conclusion. Emulating that process, which takes place in what’s called a “Neural Network” in software has shown great promise, but now we’re moving into an era in which the actual structure, the framework, will be built in hardware.

The SyNAPSE chip which was introduced 08_07_2014 is a major step in that direction; it contains the equivalent of 1 million neurons and 256 million synapses, or as some say, about the complexity of the brain of a bee! That doesn’t sound like much, but as we all know, technology moves fast and increases in complexity exponentially. The craziest part is that at full operational speed it only consumes 70 mW. That’s milli-Watts or 1000th of a Watt! A typical Intel i3 chip can draw as much as 60W, or a thousand times as much, without getting close to being able to simulate that kind of complexity. A human brain contains about 100 billion neurons, so we’re a factor 100,000 away, but when you realize that in 2011 IBM was making chips with only 256 you can imagine where we’re going. A mouse has 71 million and that’s a reasonably complex animal, isn’t it?

There’s so much talk about Artificial Intelligence these days, and a lot of ‘information’ and opinion being spread by people who really have very little if any understanding of the matter. One of the biggest holes in the arguments I often hear is that a machine can never be as complex and unpredictable as a human because everything it does is ‘programmed’; this almost always comes from people who have never even programmed anything, yet they feel very confident in their opinion! The truth seems to be that what we see as such typical biological traits comes from sufficient complexity of just about any system. Think about your PC: does it do always exactly what you tell it to? No way! Sometimes it behaves unexpectedly, due to the synergy of its components, in ways that nobody who designed the individual parts planned, or even could have foreseen. And that’s with traditional linear machines! Imagine what happens if we start building them in electrical versions of the way we’re built. Then comes the question: “will those machines believe in God?” Most asking it will immediately answer it themselves with a resolute “of course not!”, but the answer is actually “quite possibly so!” We’ve already seen complex enough simulations of neural networks exhibiting emotional behavior, and that’s no lie. In addition, we must remember that such systems are not programmed in the traditional sense of the word; a much better description would be that they are taught just like children, which means that sometimes they will not do what they were ‘instructed’ to do.

I don’t think we can overestimate the importance of this development, as I promise it will change our world even more than we can imagine. It’s not simple material, and it requires a really open mind. We MUST hypothesize what will happen if we build a machine that is as complex and capable as a human, and then what happens when it’s 1000 times as smart as the smartest human. We MUST speculate on the effects of a conscious machine.

Leave a Reply