Baidu is getting closer to finishing the world’s largest computer cluster for deep learning.
The system is reported to have 100 billion digitally simulated neural connections conceived by heavy graphics processing units (GPUs). To put things in perspective the human brain has around a hundred trillion neural connections.
Five years from now voice and picture based online searches will trump text based usership by more than half, owing to the need for easier modes of seeking information through mobile devices, explained Baidu Chief Executive Officer Robin Li.
Andrew Ng, the chief scientist at China’s biggest search engine company, Baidu, has previously worked on the “Google Brain” project as well as the Stanford Artificial Intelligence Laboratory project before he joined Baidu in May. He believes that the ‘Baidu system will be about 10 times more powerful than a 2013 computer cluster’ at the Stanford Lab, to serve provide more computing power and better image recognition. “The bigger you build these things, the better they perform,” said Ng. “Our initial task is to recognize images better, to create computer vision.”
In order to simulate brain function computers are structured with deep learning capabilities and Ng points out that such developments have in the recent past brought down speech recognition errors by 25 percent. The advent of smart mobile devices is ‘shifting demand for how users query for information;’ where it leads the industry might not remain a matter of conjecture.
Read more here.
(Image credit: Flickr)