谷歌的新电子邮件功能将使你的收件箱不那么烦

2019-06-14 14:38:53 围观 : 188

  谷歌的新电子邮件功能将使你的收件箱不那么烦人

  有时这是快速简便的电子邮件,无法回答。也许朋友问你是否可以吃晚餐,即使你那天晚上有空,你也忘了回复。

  

                  谷歌去年10月推出的电子邮件应用程序Inbox的最新功能试图通过建议对电子邮件的回复来更轻松地处理这类情况。在按下回复按钮之前,Google会根据收到的电子邮件中的内容建议三个回复。

                  该功能称为智能回复并于11月5日发布,它使用机器学习来理解消息的上下文并撰写有意义的回复。

                  机器学习是指特定类型的计算机算法,它们可以学习如何完成特定程序的操作—例如完成任务或进行预测。

                  Google在其多个应用和服务中使用机器学习,包括其照片应用,可以智能地解密照片中的主题,以便用户可以执行真正有针对性的搜索。例如,搜索“狗”应该会提取您图书馆中包含狗的所有图像。

                    

                      

                  

                    

                      

                  

                  

                  

                    

                      

                        

                      

                  

                  谷歌承诺,随着用户更频繁地从其建议中选择回复,这种新的智能回复功能将得到改善。这是有道理的 - mdash;使用的智能回复越多,Google就越了解该特定用户通常如何回复电子邮件。因此,它可以使预测更准确。

                  智能回复不是革命性的 - mdash;智能手表上存在类似的功能,因为在小屏幕上打字或对手表说话通常不太理想。尽管如此,这似乎是在移动中快速响应电子邮件的一个方便的补充。

                  

                      看看奇怪的怪异图像谷歌的自我进化软件

                  

                    

                      

                            

                              

                              

                                

                                        

                                    

                                

                                

                                    这些照片并非来自迷幻药物的人的心灵。相反,它们代表了Google的AI软件能够重新解释不同图像的方式

                  

                                

                              

                            

                            

                              

                              

                                

                                        

                                        

                                          

                                        

                                    

                                

                                

                                    谷歌的软件,称为人工神经网络,能够像人类大脑一样学习。

                                

                              

                            

                            

                              

                              

                                

                                        

                                        

                                          

                                        

                                    

                                

                                

                                    Google通过向网络提供数以百万计的图像来训练网络,以便教会它如何解释不同的对象。

                                

                              

                            

                            

                              

                              

                                

                                        

                                        

                                          

                                        

                                    

                                

                                

                                    该网络由10到30层人工神经元构成,每个人工神经元解释图像中不同级别的复杂性。

                                

                              

                            

                            

                              

                              

                                

                                        

                                        

                                          

                                        

                                    

                                

                                

                                    这些图片是通过向网络提供任意图像并允许其增强其认为最重要的图像而制作的

                                

                              

                            

                            

                              

                              

                                

                                        

                                        

                                          

                                        

                                    

                                

                                

                                    这些图像可以作为测试网络在培训过程中学习情况的一种方式

                                

                              

                            

                            

                              

                              

                                

                                        

                                        

                                          

                                        

                                    

                                

                                

                                    结果丰富多彩,刺耳而美丽。网络的每一层都可以提供不同的解释

                                    谷歌

                                

                              

                            

                            

                              

                              

                                

                                        

                                        

                                          

                                        

                                    

                                

                                

                                    谷歌将这一过程称为抽象重新解读图像“初始主义”

                  

                                

                              

                            

                            

                              

                              

                                

                                        

                                        

                                          

                                        

                                    

                                

                                

                                    某些对象通常以类似的方式重新解释。地平线通常充满塔和宝塔,岩石变成建筑物,鸟类或昆虫出现在叶子的图片中

                                

                              

                            

                            

                              

                              

                                

                                        

                                        

                                          

                                        

                                    

                                

                                

                                    除了重新解释图像之外,网络还可以通过在旧图像上不断建立新的印象,从随机噪声图像中创建“梦想”

                  

                                    谷歌

                                

                              

                            

                            

                              

                              

                                

                                        

                                        

                                          

                                        

                                    

                                

                                

                                    在未来,神经网络可以被艺术家用作视觉表达的新形式

                                

                              

                            

                            

                              

                              

                                

                                        

                                        

                                          

                                        

                                    

                                

                                

                                    谷歌将继续使用神经网络,从语音识别到识别照片应用中的人物

                                

                              

                            

                  

                      

                      

                        

                      

                    

                    

                      

                        

                          

                          1的

                                  12

                          

                        

                        

                          

                        

                        

                          

                          

                        

                        

                      

                    

                      

                        广告