IBM no longer sells computer chips, but the company has been investing a lot of money in order to develop game changing computer hardware. It is also focusing heavily on making substantial advances with regards to AI, in order to make the technology more energy efficient and easy to deploy on smaller devices like smartphones. Now it looks like the company has made a huge breakthrough.
According to MIT technology review, IBM seems to be working on creating a new chip that comes integrated with neural networking , making AI functions 100 times more efficient. Neural networking is generally integrated with software on a device. It is the feature that generally allows AI to consider the data and offer relevant responses and further learn from it. In a way it has been inspired by how the human brain works.
Researchers at IBM performed a test on the built-in neural networking chip, giving it two simple image recognition tasks with regards to handwriting and color image classification. Not only was the system accurate as the one on software, but was also able to perform the task using only 1 percent power in comparison.
Building the neural net on silicon has been thought of before. However, mimicking the synaptic that connect individual neurons in a brain onto a piece of hardware was a big challenge. But now IBM seems to have worked around the problem by taking inspiration from only two types of synapses which are related to short-term ones for computation and long-term ones for memory.
Watch: Amazon Echo Spot First Look
IBM still has a long way to go though. The design of its chips is still noted to be chunky. It carries five transistors in place of one at the moment. IBM also needs to perform more tests beyond lab simulations. However, it’s still a big breakthrough and if done right, can change the way AI is integrated onto our smartphones in the future.