Computer chips based on the human model
If we talk with our smartphones or let Facebook calculate which of our friends is in a photo, we’re using an enormous amount of processor power from the chips running in the background. The market leaders in the tech industry are therefore constantly looking for ways to make their systems more efficient. Ultimately, this performance also determines the speed at which AI research is advanced. A current and promising approach is taking its cues from biology. While our smartphones, tablets and computers depend on the performance of a central chip, the human brain monitors our nervous system and outsources individual tasks. Google’s servers look somewhat similar today – a processor optimised for neural networks called a Tensor Processing Unit (TPU) controls the work of a number of specialised chips. Competitors are advancing this method as well, while Google is already working on the next generation of TPUs. Customers of their cloud computing service are set to be able to use the new version at the end of this year.
We’d like to thank everyone involved for their fantastic work!
Read the full article here.