The latest brainlike computer chip launched by IBM might not be able to beat a sixth grader when it comes to thinking like humans, but it still manages to stimulate millions of neurons and perform complex tasks without using a lot of energy. Researchers working for the computer giant have managed to develop a stamp sized chip that has been equipped with 5.4 billion transistors and is capable of stimulating more than 1 million neurons as well as 256 million neural connections. Moreover, apart from mimicking the brain’s processing functions, individual chips can also be linked to each other to mimic the way circuits are linked in the brain.
The chip was used to perform a task – identifying objects and people from an image – something that computers have found to be extremely challenging until now. According to Dharmendra Modha, the study leader and lead researcher, the chip is in no ways a human brain. What it does is that it allows us to learn about the physiology and anatomy of the human brain.
Modha also offered an analogy in order to explain how these chips differ from the typical chips found in computers. He states that computer chips can be termed as left brain machines. They are fast, good at crunching numbers and sequential. This chip developed by IBM is typically a right brain machine.
Researchers connected approximately 4000 cores on a chip and tested the performance on complex image recognition tasks. Modha states that the project was a major undertaking and a great insight into the future. He also stated that the work took a lot of years and a large team to complete as it was multi-institutional and multi-disciplinary in nature.
DARPA is believed to have provided about $53.5 million as funding for this project. Once the team had constructed the chip, work was work was halted because Modha had issued a challenge where he promised to offer a $1000 bottle of champagne to any team member who could find a bug in the chip. However, despite the incentive, none of the experts could find one. Furthermore, Modha states that the chip isn’t just efficient, it also produces lesser heat when compared to traditional computer chips.
Modha believes that as science develops and newer technologies emerge, this creation could act as the base to rectify one of the biggest technological challenges – visual and sensory impairment. However, if devices can be made to act like human brains, they might eventually be able to understand their environments in a better manner and manage to act accordingly.
This is not the first project of its kind. A group of researchers at the Stanford University also created a unique Neurogrid system that was successful in stimulating billions of synapses and 1 million neurons. However, what makes the IBM chip so unique is that it required a single chip to perform tasks that required 16 of these Neurogrid chips. One of the biggest reasons for this difference is that Neurogrid’s memory is stored off-chip, whereas the IBM chip integrates memory as well as computation onto the same chip.
Yes, science has come a long way, but there is a longer path ahead of us.