A magazine where the digital world meets the real world.
On the web
- Home
- Browse by date
- Browse by topic
- Enter the maze
- Follow our blog
- Follow us on Twitter
- Resources for teachers
- Subscribe
In print
What is cs4fn?
- About us
- Contact us
- Partners
- Privacy and cookies
- Copyright and contributions
- Links to other fun sites
- Complete our questionnaire, give us feedback
Search:
A True Hardware Brain
by Paul Curzon, Queen Mary University of London
Computer scientists have long been working to build computers that work like human brains. Computers can do amazing things, but our brains are amazing too. While computer are good at crunching numbers, we are still better at spotting visual patterns, for example. Researchers aren't building computer brains just to make better computers though. We only understand a little of how the brain works. Building a computer brain is a way to check that understanding and to learn more. What is going on when we have those flashes of creativity out of nowhere, for example?
This is where hardware brains come in. Biologists can experiment on brain cells - neurons - both in living brains and by isolating individual neurons and experimenting on them in the lab. They can only explore the detailed working of a few neurons at a time that way though. They can't explore what the results mean for large numbers of neurons working together. By building artificial brains, we can explore what happens when billions of neurons work together.
The main way this has been done so far has been by creating software that simulates the way the brain works. A neuron can send chemical messages to others neurons. It does this only when it has received enough messages from ones that connect to it. It then 'fires', sending messages to the neurons it connects to. In the software version, the artificial neurons typically just keep a count of the messages. This is an example of a computational thinking skill called abstraction - ignoring the details of something computer scientists are modelling in software. Abstraction is used to keep things simple by leaving out aspects that don't affect the thing they are really interested in. That has led to complex 'neural network' systems that work very much like our brains and has led to new ways of creating software to do complex things humans are good at like decision making.
There is another way for a computer scientist to create a model of a brain though. Rather than building this kind of simulation, you could build hardware that works the same way the brain does - just out of different materials. It's called neuromorphic hardware, and the big difference to modern computer hardware is that it is analogue rather than digital.
What does that mean? A large part of the computer revolution was the 'digital' revolution - the idea that we can represent and process all information as 0s and 1s. We used to have analogue TVs, now we have digital ones. Music used to be stored in analogue form as the bumpy grooves on a record. Now it is digital. Music is a good example to explain the difference. The peaks and troughs of the groove on an old fashioned record encoded the music. The record player's needle sat in the groove, moving up and down. Those movements were translated into electrical signals smoothly rising and falling, mirroring the ups and downs of the groove. Those changing signals were then translated back in to the changes in sound originally recorded. With digital music everything is stored as 1s and 0s - highs or lows of voltage. Grouped together they stand for numbers which are approximations of the sound levels. Rather than gradually changing, the numbers jump from one to the next. With sound the difference doesn't matter since as long as you use enough numbers our brains can't hear the difference. Perhaps the differences between analogue and digital do matter when it comes to building brains though.
Digital took over for lots of reasons, not least the flexibility. A digital computer can process those 1s and 0s whether they are representing music or your bank account. Change the program and the computer can become a music player, a spreadsheet, a web browser, a TV (or even a brain). Analogue systems tend to be much less flexible. My old vinyl record player was just a record player, for example. It couldn't be used for anything else.
With analogue electronics, rather than forcing electronic signals to always be high or low you let the signals do the more natural thing of gradually change value. That is what neuromorphic hardware does. The artificial neurons don't just count the signals coming in, the levels change gradually just like in our brains. The resulting model will be much more accurate and that may make a difference in helping us understand how brains work. Our brains aren't digital. Perhaps things like those gradually varying levels do matter for the things we are interested in understanding, like flashes of inspiration. By building hardware brains we might just find out. Of course we might also end up with a really creative machine to rival ourselves.