Google Builds Machine with Digital Brain Capabilities

By CIOReview | Friday, March 27, 2015

FREMONT, CA: In this era of smart world, if computer vision, language understanding and speech recognition technology renders the human brain capabilities then definitely it will be easy to complete any task and save human engineering efforts. Leveraging the deep learning technology, Google has been tinkering on a similar project to build machines that can master classic computer games, reports Brian Caulfield of NVIDIA blog.

The company extends the new machine’s limit into new domains including speech, vision, language modeling, user prediction and translation. Deep learning algorithm allows the machine to work like digital brain where it can create higher levels of abstraction automatically in order to perform real world applications from the scratch.

Google includes Graphics Processor Units (GPUs) with parallel computing capabilities to its infrastructure that teaches neural networks to handle multiple tasks at the same time. With this inclusion, the company conducted a demo where the machine displayed superhuman capabilities by mastering 50 classic Atari video games after playing it few number of times.

In past, Google successfully implemented deep learning algorithm successfully in its Android smartphone software for predictive speech recognition. Now the company looks forward for more positive results with its more than 50 production applications integrated with similar technology, says Jeff Dean, Senior Research Member at Google.