It has been over 60 years since the term AI was first coined. Since then, the elusive human brain has fascinated us with its learning capabilities. The massively parallel network of synapses and neurons is practically impossible to replicate and that is why, there is no public record of a human brain being fully replicated with its full learning capabilities. The human brain has not gotten any better at improving itself in the last 60 years either (you can blame evolution for being slow), but we have created wonders using this brain, and “computational power” is undoubtedly one of those wonders.
But, what if we can build a machine using computational power that has the same learning capabilities as that of the brain? There are a number of limitations to that, though the biggest limitation is the remarkable power efficiency of the human brain. The cue is in massive parallelism. Currently, there is a well-known project by IBM known as the Systems of Neuromorphic Adaptive Plastic Scalable Electronics (SyNAPSE), which aims to achieve this goal in AI with its DARPA funding and IBM expertise. However, Google just sped ahead of IBM in AI, with its research at the Google X laboratories. Google’s neural network is built out of 16,000 computer processors, and is capable of performing complex tasks. One of those complex tasks is looking for cute cat pictures on the Internet, and the impressive fact is that the network has learned to search for these pictures on its own, without it being told to do so.
Daily Tech reports this, saying,
Thanks to the wealth of cat videos on YouTube, the cyber-brain eventually came to a single dream-like image representing the network’s knowledge of what a cat looks like. The network was able to then able to recognize its favorite thing — cat videos, no matter what subtle variations merry YouTubers come up with to their feline’s appearance.
In short, although rough, the network has successfully simulated the human visual cortex. David A. Bader, the Executive Director of High Performance Computing at Georgia Tech College of Computing claims that the visual cortex can be simulated fully within this decade.