As a hacker, do the hacking.

Here I mean the [hacker](http://en.wikipedia.org/wiki/Hacker_(term) ) . If you don't understand, never again.

Think in the hacker way.

Now I tried to experience the google photos. It's a new product since the Google IO conference 2015.

The most i want to experience is that google photos can distinguish animals from all your pictures. If you upload your photos there, the google parallel computing servers in Dalas Texas will do some machine learning based on tones of datasets those are derived from enormous objects to return you some pictures which are correspondng to your keywords.

Here I assump a fact that Google photos will not recognize your photos at the very first time after you have uploaded. Google photos will take some time backend on machine learning your photos. At this moment, your searching will not be responded.

Here I downloaded 10 photos of dogs. And try to upload all there.

And here's my google photos. I do not use that often.

Now I am trying to upload these photos at once.

I will perform search **dog** instantly after uploading's finished.

Now upload them all

No limited

And after you've finished uploading.

Try see your photos.

Yes they were upload.

Now try search ASAP.

At a time of

Not. But clearly you have uploaded 10 dog pictures.

Now think out of box. The parrallel coputing servers in Dalas are so busy learning your pictures, comparing your pictures with a enormous dataset which google had ever collected.

When can your get the **dog** result?

Later. ( I try to search it later)

In another way, all your privacy are under the inspection of google. Google can detect a dog, and also google can detect what a human is doing. It's about who, where, when, how and what. All in a photo and things will be revealed on a consumption of machine learning.

What do you think?

At the time of

I got a piece of result about **dog**.

Now the servers in Dalas has just labeled one of the 10 photos which i have uploaded. The next 9 is on-going.

As what google said, they can detech a dog successful at 85%.

If this is for real, in this instance, I could get 7-10 photos if the bias is not too big.

It is still on-going. Try to wait.

Now it's 10:54 P.M.

And yet still only one photo is labeled related to 'dog'.

Is google photos is that accurate?

The accuracy by now is 10%.

You've just uploaded your photos to google. And trust them. But they do not give you what you want to feel.

Or you could take google photos as an image/video warehouse for you. But you pay google/(big brother) with YOUR PRIVACY.

SO HOW ABOUT THAT?

---updated at 1:29 A.M. June 2015 ---

Now we can see almost 5 out of 10. Rate at 50%.

Still machine learning...

Google photos -- reverse thinking的更多相关文章

  1. 使用纯 CSS 实现 Google Photos 照片列表布局

    文章太长,因为介绍了如何一步一步进化到最后接近完美的效果的,不想读的同学可以直接跳到最后一个大标题之后看代码.demo及原理就好,或者也可以直接看下面这个链接的源代码. 不过还是建议顺序读下去,因为后 ...

  2. Say goodbye to my photos&videos

    刚刚得知一个悲惨的消息:虽然2012已经过去了,但是世界末日并未过去.嗯,我不是来严肃的,我是来搞笑的.毕竟,我已经如此伤心了.中午结束考试,下午看了一半的电影然后躺室友的床上睡了一觉,醒来看到阿姨发 ...

  3. (转) Graph-powered Machine Learning at Google

        Graph-powered Machine Learning at Google     Thursday, October 06, 2016 Posted by Sujith Ravi, S ...

  4. 谷歌 google

    google Google是搜索引擎名,也是一家美国上市公司名称.Google公司于1998年9月7日以私有股份公司的形式创立,以设计并管理一个互联网的搜索引擎.Google公司的总部称作“Googl ...

  5. 2015 GOOGLE I/O大会看点总结:新的开发工具及云端测试工具

        http://ask.android-studio.org/?/article/48 尽管凌晨开始的主题演讲略显晦涩并继承了谷歌一贯的东一句西一句想起啥说啥的混乱风格,但期待用技术改变世界的技 ...

  6. Google推出iOS功能性UI测试框架EarlGrey

    经过了一段时间的酝酿后,Google很高兴地宣布了EarlGrey,一款针对于iOS的功能性UI测试框架.诸如YouTube.Google Calendar.Google Photos.Google ...

  7. 10 Free Image Hosting Sites for Your Photos

    https://www.lifewire.com/free-image-hosting-sites-3486329 Wondering if there are there any good site ...

  8. 通过AI自学习,Google让Pixel 3的人像模式更优秀

    通过AI自学习,Google让Pixel 3的人像模式更优秀 Link: https://news.cnblogs.com/n/613720/ 虽然双摄手机已经在市场上普及,其所带来的人像模式.多倍变 ...

  9. 这是一份很有诚意的2017 Google I/O大会的汇总 & 解析

    前言 在刚过去的凌晨(北京时间 5月18日 1.00-3.00),一年一度的2017年Google I/O大会在美国谷歌山景城海岸线圆形剧场如期举行 Google I/O 大会:Innovation ...

随机推荐

  1. 使用Webbrowser的一点心得体会

    原文:使用Webbrowser的一点心得体会 自从用上VS2005后,发现多了个WebBrowser控件(.net 2003中不带),为图方便吧,有好多小工具就用这个写的,慢慢也有点体会了,总结一下, ...

  2. Web API 2:Action的返回类型

    Web API 2:Action的返回类型 Web API控制器中的Action方法有如下几种返回类型: void HttpResponseMessage IHttpActionResult 其它类型 ...

  3. rpm安装FAQ

    1.error: cannot create %sourcedir /usr/src/redhat/SOURCES错误的解决方案 显现error: cannot create %sourcedir / ...

  4. inux平台的C与C++

    课堂里学不到的C与C++那些事(一) 首先,声明一下这是一个系列的文章.至于整个系列有多少篇,笔者也不知道,不知道有多少篇,也不知道多久会更新一篇.反正只有一个原则,写出来的文 章能见得人才会公布出来 ...

  5. java通过SVG导出图片

    import java.io.ByteArrayInputStream; import java.io.File; import java.io.FileOutputStream; import ja ...

  6. Oralce 处理字符串函数

    原文:Oralce 处理字符串函数 平常我们用Oracle主要有两种字符串类型1.char始终为固定的长度,如果设置了长度小于char列的值,则Oracle会自动用空格填充的.当比较char时,Ora ...

  7. POJ Big Christmas Tree(最短的基础)

    Big Christmas Tree 题目分析: 叫你构造一颗圣诞树,使得 (sum of weights of all descendant nodes) × (unit price of the ...

  8. CF:Problem 427C - Checkposts良好的沟通 Tarjan算法

    tarjan算法的第一个问题 喷我的脸....手写叠式开成BOOL,我一直在找错了... #include<cstdio> #include<cstring> #include ...

  9. jquery调用wcf案例

    ----------根据其他网友总结 1.在契约接口上添加:[WebInvoke(RequestFormat=WebMessageFormat.Json,ResponseFormat=WebMessa ...

  10. boost------ref的使用(Boost程序库完全开发指南)读书笔记

    STL和Boost中的算法和函数大量使用了函数对象作为判断式或谓词参数,而这些参数都是传值语义,算法或函数在内部保修函数对象的拷贝并使用,例如: #include "stdafx.h&quo ...