close
AI

In the face of Twitter, AI said “really with no heart”

4bba53f4-a391-11e9-8c6d-b7fde4f9c494.jpg

On the social platform Twitter, the vagaries of “public opinion and wind” often make us stunned. It used to be the darling of the technology industry, the birthplace of several political and economic entertainment events, analysts and Wall Street’s “social future”; after the high-profile listing, the stock price continued to dive, and was labeled “Twitter is dead”, ” The next Yahoo! label, I want to sell it, I can’t sell it. After the first profit in Q4 in 17 years, Twitter’s revenue began to grow and continue to the present, so there have been countless counts of Twitter’s second rise in the market. Successful learning: Focus on “news” positioning, video, management of spam, organizational adjustments… In short, these are hidden dangers when performance declines. These are all advantages when performance is rising, and it is very popular for passers-by. This is probably the legend. “Things have two sides”. Of course, no one in Twitter’s “Jesus Survival” denies the need to introduce AI technology. Twitter also did a lot of work on AI. Despite this, we are also aware that the process of reversing the results does not make sense after the phenomenon. Therefore, as a media that insists on AI endorsement, we are not willing to simply regard AI as Twitter. “The straw of life.” On the contrary, it is far from enough for us to plant and harvest Twitter on AI. So, what are the lessons and revelations brought about by looking back at Twitter’s evolutionary history of AI? What are the pits in front of Twitter for the survival of the Jedi? When Facebook was in the IPO in 2012, many critics were not optimistic and thought that Twitter was the model of this generation of Internet companies. But the winds and turns are turning, Facebook is in the limelight in 2016, and Twitter is going to the top, leaving the company, and having to sell the company’s collapse. After several acquisitions, Twitter started a difficult journey of self-help. The process is naturally painful, and the pits and minefields are also full of dazzling. External factors such as brain drain, growth stagnation, and social diversion are not mentioned at all. From the product level, there are many problems with Twitter. The first to bear the brunt is the content environment that has always been criticized by people. On Twitter, “Keyboard Man” is like a skeleton, spreading malicious comments and personal attacks that trample on human rights. It is also full of spam that is spread by fake news, robots, etc. In this regard, CEO Jack Dorsey believes that “Twitter is a communication tool, not good at judging content disputes between individuals.”, so Twitter’s practice after receiving a report is often a title. This neutral attitude has clearly accelerated the decline of Twitter users, especially the loss of celebrities and KOLs with the most serious language violence, which directly led to the shrinking of Twitter content. Another point is that after insisting on the positioning of the media platform, Twitter has no obvious upgrade in product experience and content value. Taking the live broadcast product that emphasizes “on the spot” as an example, it is undoubtedly the best shortcut to show Twitter “news power”, but Twitter still adopts the product thinking of the PC era, just moving the scene from the traditional TV station to the mobile phone, as a pair The simple addition of text and image content, as for the event process, other people’s chain reaction, etc., are collectively absent. In the words of technology analyst Ben Thompson, it just proves that “the success of this company (Twitter) relies more on luck than on insight.” Of course, there are still many dilemmas in Twitter. Stir-fried rice. What’s more important is what kind of action it has done in technology, and it has made a difficult road for itself in the minefield where there is a “surprise” everywhere. The Twitter AITwitter that flies after the stupid bird obviously feels that he has to make some changes. In 2017, US billionaire Mark Cuban began buying shares in Twitter. In an interview with CNBC, he said: “I think they have finally started the action of artificial intelligence.” At the time, Twitter began using artificial intelligence to filter and filter tweets recommended to readers.此后,Twitter开始了一系列的AI改造计划:1.用机器提高内容整治效率Twitter终于意识到,唯有健康有序的平台环境,才能留住核心用户,并为广告商提供价值,最终形成业绩增长的良性循环。因此,机器学习算法的引入就变得至关重要。Twitter收购了伦敦创业公司Fabula AI,这家公司的专长是图形深度学习,使用机器学习来分析大型数据集并找出关系和交互。这能够帮助Twitter的机器学习团队通过关系映射来更好地发现社交网络上的假新闻,垃圾邮件和其他问题,进而改善平台对话的健康状况。2018年,Twitter发布了一个名为Exploring Online Hate的面板工具,能够通过1000个Twitter账户的样本来寻找仇恨活动的规律、趋势,然后利用算法生成更大的数据集,从而实时掌握网络上仇恨言论,控制相关的热门话题和言论来源。目前,Twitter有38%的有害内容是由算法自动筛选出的。2.在信息流中引入了算法推荐信息流是Twitter首创的信息分发方式,2012年,Twitter就收购了新闻聚合推送网站Summify,用于理解和挖掘Twitter的信息流数据,追踪新闻的传播效益。但Twitter在信息流中接入算法推荐的时间,却要晚于Facebook。2016年,Twitter推出了一项新功能:通过算法自行筛选它认为对用户重要的内容。用户的信息流不再按照时间轴显示,而是通过推荐机制显示Twitter认为对用户重要的信息,以及向他们推荐未关注账号的内容。2017年,又上线了“探索”选项卡,通过算法将热门话题、重大新闻以及搜索功能集中起来,放在一个功能分区,以便用户快速追踪到真正感兴趣的故事和品牌。3.靠AI提升交互体验2016年,Twitter收购了伦敦的机器学习创业公司Magic Pony Technology,对方的算法能理解“图像的特征”,这使得Twitter能重新在视频流、直播等内容上再次出发。比如推出互动视频功能(conversational video),根据上一步的操作提前录制好多份剧情,按照用户行为(选择Yes or No)来推进到下一步。为直播增加了一个名为Moments的页面、将推文整理成前后连贯的“故事”。互动性的增强和新的内容形态也重新吸引了广告主的青睐,2018年第二三季度,来自视频广告的收入为Twitter贡献了超过一半的广告收入。再比如机器翻译之类的措施,来帮助各个区域的跨文化交流等等,Twitter在广告转化(ad cvr)和反垃圾信息上(anti spam)做了很多技术优化,用户的参与度也有所增强。目前看来,Twitter发力AI有几个明显的特点:1.起步晚,令人眼前一亮的案例和技术十分有限;2.技术布局围绕营收展开,尽可能让流量最大化,助攻广告效果;3.几乎所有功能创新都是在补短板,缩小与其他社交平台的体验差距。然而,Twitter真的把握住了AI的命脉了吗?似乎也未必。AI到底能不能挽救推特某种意义上来说,Twitter的AI战略更像是在“救火”,哪里有险情点哪里。对于极度渴望重回聚光灯下的Twitter来说,这么做似乎也无可厚非。但却造成了不少困境:一方面,Twitter无法凭借AI生长出属于自己独一无二的“长板”。与Facebook、谷歌、苹果,甚至它的“中国门徒”微博等动辄AR、云服务的能力相比,Twitter的技术创新,给人的感觉就很索然无味。比如Twitter在2018年11月推出的“探索”选项卡,Facebook在2016年就尝试过类似的做法。信息流算法的表现也并不好, 很多Twitter用户表示被智能算法控制的timeline变得支离破碎,在最新资讯里刷到二三小时前的内容。Twitter不得不在2018年9月,宣布重新为手机客户端用户提供了切换正常时间流的功能。另一方面,对广告变现的过度追求,正在透支媒体平台的公信力。作为一个广告收入占到90%以上的社交媒体,Twitter最大的矛盾就在于,用户与广告客户的利益最大化是有出入的。比如说,为了实现更高效率的广告变现,就需要精准定位、用户画像、精准推送,不可避免地要“售卖”用户数据。在一些极端情况下,社交媒体还会暴露用户的安全信息,比如确切地址。有研究显示,带有地址信息的推文占每天总推文的 2%,也就是 440 万条。另外,AI基础能力和人才的不足,或许会让Twitter面对更深的责难。面对上述问题,其他在AI技术上发力较早的科技企业,就可以通过技术手段最大程度地规避数据隐患。比如Facebook就宣布将端到端的加密服务贯穿到旗下包含的所有即时通讯工具。今年的谷歌I/O大会,也发布了一系列与隐私管理有关的新产品,比如语音助理谷歌assistant能够离线使用,无需将用户数据传输至云端处理。而同样的境况,Twitter超3200万Twitter账户密码泄露,官方除了甩锅黑客并重置密码,似乎就再也无所作为。(Twitter首席执行官杰克·多西,和Facebook首席运营官雪莉·桑德伯格,共同出席美国国会听证会)内容上也同样,谷歌近年来不断更新和改进搜索引擎和算法,满足用户对高质量新闻的需求。比如,Google Search将新闻内容整合到精选摘要中,形成 “热门新闻”的轮播。YouTube也在积极开发新闻内容,在信息流界面添加了“突发新闻”的功能板块。2018年,还投入了2500万美元扶持视频新闻。由此可见,尽管商业模式大同小异,Twitter也做了许多方面的努力,这并不能帮助它解决未来的问题。技术地基的匮乏,将让Twitter这座社交大厦继续风雨飘摇,勿谓言之不预。或许,在商业模式这个根本问题转变之前,AI注定只能是一个奢侈而空洞的梦。。

Tags : Start a business互联网创业互联网创业项目