Google x: computers smarter, smarter than humans-Google's computer system, advanced learning system, Google-IT information Google x: computers smarter, cleverer than humans

Computer is now beyond human? Answers do not know, but Google's computer system (Computer System) is growing fast to learn new things, develop its Google engineer said, now can't keep up with the pace of the computer system, in particular their "deep learning" System. It was stated that the system without any human training and instruction, uses its own internal algorithms automatically extract information from the huge amounts of data.

Machine Learning Conference in San Francisco (machine learning Conference), Google software engineer in the Department of mysteries Quoc v. Le said, Google has been unable to accurately understand the "deep learning" system is how to improve their level of recognition in photos.

The current "deep learning" system has evolved into the Google cannot keep up with the pace of the monster, so surprise them. Quoc v. Le, developed a how to identify a particular thing is extremely complex, and he spent a long time on that project has not results, but the system to understand things without any problems.

It is understood that in order to study the machines "deep learning" system, Google's tens of thousands of computers to connect, including 16,000 computer processor, 1 billion nodes of neural networks, creating one of the largest neural network. Google makes the system on the Internet "self-taught". Faced with tens of thousands of different things, and machine learning system's ability to double growth.

As early as June this year, Google's "deep learning" system active learn to search for cat face in a creative way. Google randomly selected from 10 million videos on YouTube thumbnails. Over a period of 3 days, ignorant of the concept and appearance to a cat "deep learning" system to build a perfect cat picture, simultaneous use of different levels of storage units to distill the basic characteristics of the cat, and then can be identified from the tens of thousands of pictures pictures of cats, with an accuracy rate of up to 74.8%.

Google the "deep learning" system can be automated to process huge amounts of data and classification, such as pictures. Google is currently applying its technology to a number of their products, such as the Android voice search feature of image recognition, search engines, Google translate, and so on. While the decline in cost of computing resources, as well as large data centers growing Google plans to use the system to promote development in many areas, especially in machine vision and perception, speech recognition, language translation and other areas.

Google X：计算机更智能了，比人类还聪明 - 谷歌计算机系统,深度学习系统,谷歌 - IT资讯
Google X：计算机更智能了，比人类还聪明

计算机现在超越人类吗？答案也说不清楚，不过谷歌的计算机系统（Computer System）正在飞速学习新的东西，开发它的谷歌工程师表示，现在已经赶不上该计算机系统的步伐了，尤其是自家的“深度学习”系统。据称，该系统无需接受人类的任何培训和指令，自行利用内在算法从海量数据中自动提取信息学习。

在旧金山举行的Machine Learning Conference（机器学习大会）上，谷歌神秘部门的软件工程师Quoc V. Le称，谷歌已经无法准确了解“深度学习”系统究竟是如何提高自身在照片中识别事物的水平了。

目前“深度学习”系统已经演变成谷歌无法跟上步伐的怪物，令他们大吃一惊。Quoc V. Le表示，研发一个如何识别特定事物的功能非常复杂，他花了很长时间时间在该项目上始终没有结果，但是该系统对于理解事物没有任何问题。

据了解，为了研究机器“深度学习”系统，谷歌数万台电脑连接，包括1.6万个计算机处理器、10亿个节点的神经元网络，创造了全球最大的神经网络之一。谷歌让该系统在互联网中“自学成才”。面对上万种截然不同的事物，机器学习系统的辨识能力一直都在以翻倍增长。

早在今年6月份，谷歌的“深度学习”系统创造性地主动学会了搜索猫脸。谷歌从YouTube上随机抽取1000万张视频缩略图。在3天的时间里，对猫的长相和概念一无所知的“深度学习”系统构建出一张理想的猫咪图片，同时利用不同层级的存储单元成功提炼出猫的基本特性，然后能从数万张图片中识别出了猫的图片，正确率高达74.8%。

谷歌该“深度学习”系统可自动对海量数据进行处理和分类，比如图片。谷歌目前已经将其技术运用到了多项自家的产品当中，比如Android的语音搜索、搜索引擎的图像识别功能、谷歌翻译等等。虽然计算资源的成本下滑以及庞大数据中心的日益增多，谷歌计划利用该系统推动众多领域的发展，尤其在机器视觉与感知、语音识别、语言翻译等领域。