During the July 14 Microsoft Research Faculty Summit in Washington Redmond, Microsoft research team claims that they have been passed Adam’s depth learning system and achieved a new record. Under the same load, compared with Google Brain, the training time is only 1/30 of the machine and the 1/2 of time .

We have entered into a new era of artificial intelligence.

Up to now, including Google, Facebook, Microsoft, Apple and other top technology companies are developing artificial intelligence through the “depth learning”, to seek to improve in all areas, their typical applications include speech recognition, language translation, and computer vision. In the new chapter in Artificial Intelligence, Google is generally considered to be a pioneer and it also recently hired a core researcher of depth learning movement – Geoff Hinton of University of Toronto. Previously, Google has also repeatedly demonstrated the enhancement of AI technology which impacted the real-world applications, including the improvements of deep study to the voice search and Android smart devices, moreover, left a number of records in the field of speech recognition and computer vision.

During the July 14 Microsoft Research Faculty Summit in Washington Redmond, Microsoft research team claims that they have brought machine learning into a new era by depth learning system. We learned via Microsoft that, contrasting with the previous system, Adam will twice the speed in image recognition (based on the training of dog or vegetation data set ), but only use less than 1/30 of the machines. Peter Lee, head of Microsoft Research said, Adam aims to build the largest brain. He also revealed that in a benchmark test called ImageNet 22K, the neural network performance of Adam was already much higher than the data has published by Google Brain. The test is based on 22,000 types of pictures, including Google Brain.

Microsoft said the purpose of the new Adam depth learning algorithm is not to beat Google, the key is to explore a better optimization which enable the machine to handle the data better.