Scientists One Step Closer to Building Human Brain

John Lister's picture

Scientists have designed a simple computer circuit to successfully recognize three different letters. It sounds unimpressive, but it's actually a huge step towards replicating and learning more about the human brain.

One of the curious things about a computer is that although it is considerably faster than a human at basic calculations (which are what its operations boil down to), there are certain tasks at which it is far inferior.

Most of these tasks involve 'human' abilities, such as judgment and making snap decisions. A good example is situations which involve a large number of possible solutions, such as working out the best route to visit several locations as quickly as possible. Here, humans are better because they can quickly spot an acceptable solution rather than explore every possible combination in detail.

Another such task is in image recognition. Humans are very skilled at recognizing pictures and faces, as demonstrated by the way we can quickly spot a friend on the street, even if they have had a haircut or are wearing new clothes. Computers tend to struggle with this because they can't spot patterns instinctively and instead rely on breaking images down into individual pixels and comparing them one by one. Oftentimes computers become 'confused' by small variations if the pattern is distorted or changed. (Source:

Computer Circuit Simulates Brain Workings

Researchers at University of California in Santa Barbara have been working on creating computer circuits that work in the same manner as a neural circuit in the brain. Each circuit is made up of millions of different interconnected neuron cells, which then pass on messages through chemical and electrical signals. (Source:

The big advantage of a brain is that it can route signals to process multiple pieces of information at a single time, allowing it to consider different possibilities and compare them. In contrast a computer processor, no matter how fast, can only usually work on one calculation at a time, meaning it must solve problems by running through a sequence.

The researchers created the computing equivalent of a network of 100 neurons, which compares to around 100 billion neurons in the human brain. They then programmed the computer to look at images of letters, each of which was a Z, and N or a V, distorted by "noise" in the picture, with a variety of different typefaces.

Memristors: Transistors with a Memory

The computer successfully recognized the images thanks to the use of "memristors," a word combining memory and resistor, to simulate neurons. A memristor is similar to a transistor, but its resistance level changes as electricity flows through it in a particular direction.

In effect this allows the circuit to build up "memories" that remain accessible as the computer continues to explore different possibilities. In turn, that allows it to learn and store rules of thumb that let it categorize the letters increasingly quickly.

While there's still a huge amount of work to do, the key is that the computer compared the images against its data on what the various letters look like using the tactics of the human brain, rather than a traditional computing program; that is, it tried to quickly spot patterns rather than meticulously break down the image.

What's Your Opinion?

Do you think a computer can ever truly replicate the workings of the human brain? Is such research a gimmick without practical use? Or do you think it can teach us more about both biology and better computing?

Rate this article: 
Average: 4.4 (8 votes)


Dennis Faas's picture

I can see a project like this growing at a rapid pace as they introduce more memristors to help solve complex puzzles. Eventually they'll get it down to a science, and then be able to produce memristors on a mass scale using VLSI (very large scale integration). Before you know it, drones will be equipped with the memristor technology and then we'll be one step closer to RoboCop - or maybe even James Cameron's Terminator.

stooobeee's picture

After Richard Feynman received his Nobel Prize in 1965 for his discovery of Quantum Electro-Dynamics, he took a sabbatical for three months to study the brain. At the end of the three months he said, "I think I am going back to studying what was far easier." There is a huge difference in building a computer verses building a brain---the first makes statistical errors, but the second must make random errors---and the right ones. A machine must be impregnated with an ability to make not only a right number of mistakes, but the right ones. A computing machine is all together different than a thinking machine. If a computing machine could be built to think, it would no longer be a computing machine, but a thinking machine.

Stuart Berg's picture

What if spammers could eventually use this technology? They would be able to read those website "Are you a human?" images that are even sometimes tough for humans to read. That will bury the Internet with orders of magnitude more spam than ever! It could be the downfall of the Internet.

John Lister's picture

Funnily enough, Google was so successful in developing computer software to automatically read house numbers on its Street View images that it realised that same software could do a very good job of getting past its own CAPTCHA security tests that ask you to read an image of some text. It's now having to develop alternative tests.

stekcapofni's picture

I believe these artificial brains will teach us that we are not as smart as we think we are.