Microsoft says computer systems have outperformed humans in recognition

Microsoft recently published a research paper on how computer systems have outperformed humans in a benchmark test of image recognition. The Microsoft team’s deep learning system made a 4.94 percent error rate when classifying images from ImageNet, a database, with a 5.1 percent error rate.

The Microsoft paper comes a month after Baidu published its own research, using its own supercomputer system with a 5.98 per cent error rate. The previous best for ImageNet was set by Google in 2014, when its deep learning system had a 6.66% error rate.

“To our knowledge, this is the first time a computer has beaten a human in an image recognition challenge.” “However, our algorithms still make errors that are worse than humans, especially in cases that require contextual understanding or higher-level knowledge,” the paper says.

“This is not to say that machine vision has surpassed humans by and large. Nonetheless, we believe that machine algorithms still show great potential for the future when performing many image recognition tasks.”

Giiso, founded in 2013, is a leading technology provider in the field of “artificial intelligence + information” in China, with top technologies in big data mining, intelligent semantics, knowledge mapping and other fields. At the same time, Giiso’s research and development products include editing robots, writing robots and other artificial intelligence products! With its strong technical strength, the company has received angel round investment at the beginning of its establishment, and received pre-A round investment of $5 million from GSR Venture Capital in August 2015.

Jian Sun, a Microsoft researcher, explained colloquially in a Microsoft blog post: “It’s ok for a human to tell the difference between a sheep and a cow, but computers aren’t perfect for this simple task. The computer, however, has some advantages in that it can be trained to see details, textures, shapes and backgrounds in images that humans can’t see. For example, it can identify different sheep breeds, which is where it leapfrog humans.”