Artificial intelligence specialists have been searching for a way in which the human brain can be accurately replicated for some time. The brain is still viewed as the most powerful computer currently known to humans owing to the immense number of operations that it can perform in a single second.
By copying nature, researchers believe that computers could achieve the same thing, creating a more powerful computer that requires less energy and time to operate. This was achieved for the first time by researchers at UC Santa Barbara earlier this week.
Engadget reports that the team successfully created a neural circuit of 100 artificial synapses. When viewed in comparison with the size of the brain, estimated at having one quadrillion synapses, the number may seem inconsequential but the team is viewing it as a significant and hugely important step towards creating a true “brain”.
The simple circuit is based on new memristor technology that is slated to replace resistors as quantum computing becomes more mainstream. Memristors are very new circuit components. Envisioned in 1971 but only made in 2008, a memristor’s resistance depends on how much charge has passed through it in the past, meaning that it remembers its own history. More detail is available from the memristor information site.
This unique characteristic of state preservation allowed the researchers, led by Dmitri Strukov, professor of electrical and computer engineering, to create a form of analogue memory based around charged ions instead of the usual electrons employed in digital memory.
The analogue memory is what ultimately led to the production of the hand-held circuit board. An equivalent digital device could be as large as a super-computer once scaled up, requiring much more energy and time to operate.
The “brain” was tested with an image classification task. This is an area which computers are known to struggle at, unable to recognise shapes and text in images with anything like the speed and precision of the human brain.
The memristor neural circuit identified the letters “z”, “v” and “n” in various different styles amongst large amounts of visual noise in a way that a digital computer could never have done. Although the technology must be scaled before any kind of “real” brain can be made, it is an exciting look into the future of computers and circuitry in which your devices may become genuinely “smart”.
