The ultimate goal of every artificial intelligence has been to emulate the human brain, but copying 86 billion neurons and their connecting synapses has always been one of the biggest hurdles facing AI design. This is because the synapses between neurons have the amazing ability to grow faster the more they are used, a process we experience as learning and something no machine has been able to fully simulate. However, researchers at Harvard School of Engineering and Applied Sciences (SEAS) may have found an answer.
Image courtesy of Michael Novelo on flickr CC
SEAS scientists have created a transistor that mimics the behavior of a synapse. It is able to control the flow of data that it receives and it can also physically react to changing signals. The chip could represent a new generation of artificial intelligence that is self-generated, rather than depending on complicated algorithms. In theory, a system integrating a large number of these synaptic transistors could emulate a human brain.
The device offers several major advantages over existing, most notably major energy efficiency due to higher electrical sensitivity in the materials. The transistor can also remember its current configuration even without power, meaning no loss of data, and is not restricted to a binary system because it can configure itself in an infinite number of ways.