There are many reasons to study the brain. Understanding cognition can help us slow, or even reverse, degenerative brain diseases. It can help us understand the environmental factors that lead to increased cognition and connectivity. It may even help us to boost our brain power. Neuroscience, however, doesn’t end there.
What about using our understanding of the human brain to build its synthetic equivalent? The science of your brain can be applied to artificial intelligence. The idea is known as Artificial Neural Networks (ANN) and it could be amazing. Especially if Ann passes the Turing test.
The Turing test, developed by Alan Turing in 1950, is meant to test AI intelligence. The aims? That our computerized person will be indistinguishable from the real – blood and bones – thing. As you may imagine, it is a tough nut to crack.
The problem with Ann is her synapses. The human brain has somewhere between 100 and 1000 trillion synapses. Even the lower end of that spectrum is staggeringly high. Artificially recreating these synapses is more than a little bit complicated, but perhaps not impossible.
Researchers at the University of Southampton recently reported a breakthrough in the synapse hardware sphere. They used metal-oxide memristive devises as synapse structures, and they asked their Ann to learn, unsupervised.
Though researchers still have a long way to go, we seem to be one step closer to an artificial intelligence that is actually intelligent. This is, of course, the pie-in-the-sky dream of many an AI enthusiast. For regular people, this new technology may have a positive impact much sooner. If Ann is able to learn, she may be able to process corrupted data, leading to much more efficient big-data processors. This is what Ann’s creators showcased in their study, though I, for one, am still holding out for the singularity.