A group of researchers from the Massachusetts Institute of Technology (MIT) recently published a paper describing a new type of chip that can simulate synapses (neural connections) in our brain. This impressive project has immense implications for our understanding of computing performance and how humans and computers may be able to communicate in the future.
The research team at MIT experimented with unique devices called ‘memristors’ that are created using silicon and silver/copper alloy. These memristors are capable of simulating brain synapses in such a way that they can store and recall data with very high accuracy indeed. While your standard transistor will only provide you with binary results (TRUE or FALSE), a memristors provides a whole range of values which makes it similar in function to a human brain.
In the paper, the researchers state that their ultimate goal is to recreate complex artificial neural networks (ANNs) comprising of memristors. In the status quo, such ANNs use up massive amounts of GPU computing power. However, with memristors, the team claims that these complex neural nets can be localized in very small devices as well, including smartphones and cameras.
Essentially, what this means is that the application of memristors will be able to bring supercomputer-level intelligence to your smartphone in the future. Imagine being able to process tremendous chunks of data for weather predictions from a device that you can literally retrieve from your pocket!
It’s still a long way off, but the team behind this project suggests that eventually, this could lead to portable, artificial brain computers that can perform very complex tasks on the scale of today’s supercomputers, all with minimal power requirements and without any network connection required.