MIT Scientists Design a Computer That Works Just Like a Human Brain

Monday, 22 January 2018 - 6:30PM
Artificial Intelligence
Monday, 22 January 2018 - 6:30PM
MIT Scientists Design a Computer That Works Just Like a Human Brain
< >
Flickr/Health Blog
As complex and impressive computers are becoming, they still don't hold a candle to the human brain. Sure, we can't instantly recall every piece of data that ever gets saved into our memory banks, and we're not as good at playing board games, but we have the ability to adapt and reason that a computer simply can't emulate.

Even advanced computer brains are unable to deal with concepts that don't fit into their existing programming, while organic brains (in most circumstances) are much better at the classic human trait of just making up plans as we go along. That said, humanity's time as the grand champions of winging it might be coming to a close.

A team of scientists at MIT have completed a successful initial test of a computer modeled after brain synapses rather than binary 1's and 0's, which could well lead to robot brains that are structured like our own - thereby giving our adaptability to computers so that we become truly obsolete.

Computers are limited by their dependency on binary inputs - if a code doesn't come in those 1's and 0's, it won't make any sense to a machine. Conversely, humans have a hundred billion neurons within our brains, along with a hundred trillion synapses that can fire out packets of analogue data, where a different strength of signal will affect the meaning of the message that's being sent.

According to Jeehwan Kim, who is working on the project:

Opening quote
"Once you apply some voltage to represent some data with your artificial neuron, you have to erase and be able to write it again in the exact same way. But in an amorphous solid, when you write again, the ions go in different directions because there are lots of defects. This stream is changing, and it's hard to control. That's the biggest problem - nonuniformity of the artificial synapse."
Closing quote

Computers literally aren't designed to put up with these kinds of complex messages, and as such, the scientists behind this new emerging technology have had to return to the drawing board, completely building hardware from scratch that's able to interpret a matrix of different inputs all controlled by the amount of electricity that's sent in each burst.

Essentially, this emulates the human brain, while also making allowances for the ways that computer technology can't quite manage to live up to our own processing power.

All of this has culminated in a relatively simple initial test: a computer that's able to correctly identify different people's handwriting at 95% accuracy. This is a small victory (traditional computers are already capable of managing 97% accuracy at present), but it's indicative of a major change that could revolutionize computer in the near future.

It's probably also worth noting that this test was conducted entirely within a simulation, so we're not able to see the actual results of such an experiment just yet. The initial successful computer model, though, suggests that this is worth exploring future in our quest to build bigger, better, more adaptable robots.

These efforts aren't designed to replace the way that current computers are designed, but rather, they'll work with emerging machine learning technology in order to produce a computer that can think like a human, and also process numbers like a calculator.

This best of both worlds approach will mean that the future of computing will make most human beings entirely obsolete. At least we'll always be better than computers when it comes to eating cake. It's not much, but it's something.
Science News
Artificial Intelligence