Zac | 人工智能什么时候才可全面影响搜索算法?
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><img src="https://mmbiz.qpic.cn/mmbiz_gif/P5rwNx8qkB2lxnoq0ZOjCE2picJk1uCB2znBc0C68dI1AGqxLvX3TM6pjmKzLyVbLichIXREmEtEOeublX2AqzpA/0?wx_fmt=gif&tp=webp&wxfrom=5&wx_lazy=1" style="width: 50%; margin-bottom: 20px;"></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">过去一两年,人工智能是最火的并且快速进入实用的技术。以前写过人工智能将彻底改变SEO,<span style="color: black;">亦</span>介绍过人工智能在搜索算法中的<span style="color: black;">实质</span>应用,但<span style="color: black;">必须</span>说明的是,到<span style="color: black;">日前</span>为止,人工智能在搜索算法中的应用并不广泛。影响人工智能在搜索算法中大范围<span style="color: black;">运用</span>的最<span style="color: black;">重点</span><span style="color: black;">原因</span>可能是,搜索引擎工程师不<span style="color: black;">晓得</span>人工智能系统到底是怎么做出判断的,然后又<span style="color: black;">引起</span>另一个<span style="color: black;">要紧</span>问题:很难debug。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">人工智能是个黑盒子</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">用不太严格但容易理解的方式说,深度学习<span style="color: black;">便是</span>给现有数据(<span style="color: black;">海量</span>数据)打标签,<span style="color: black;">而后</span>系统自己总结数据和结果(<span style="color: black;">亦</span><span style="color: black;">便是</span>所打的标签)之间的关系,面对新数据时,就能依据自己总结的规律给出判断。对围棋<span style="color: black;">来讲</span>,<span style="color: black;">没</span>论历史棋局还是自我对弈,AlphaGo<span style="color: black;">晓得</span>盘面,<span style="color: black;">亦</span>知道<span style="color: black;">结果</span>(<span style="color: black;">亦</span>是一种标签),系统就会总结规律,面对新盘面时判断赢棋的概率。但AI系统找到的是数据的<span style="color: black;">那些</span>特征,与结果之间是<span style="color: black;">怎么样</span>的关系,连创造AI的工程师<span style="color: black;">亦</span>不<span style="color: black;">晓得</span>。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">因此</span>,<span style="color: black;">此刻</span>的人工智能系统是个黑盒子。<span style="color: black;">咱们</span><span style="color: black;">晓得</span>AI判断的正确率高,但不<span style="color: black;">晓得</span><span style="color: black;">为何</span>,不<span style="color: black;">晓得</span>是怎么判断的。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">搜索算法中的AI<span style="color: black;">亦</span>是如此。百度搜索工程师的说法很少见到,只是<span style="color: black;">晓得</span>百度<span style="color: black;">此刻</span>All In AI了。Google工程师<span style="color: black;">知道</span><span style="color: black;">暗示</span>过,<span style="color: black;">她们</span>对RankBrain到底是怎么工作的<span style="color: black;">亦</span>不太清楚。在这种<span style="color: black;">状况</span>下,在算法中<span style="color: black;">海量</span><span style="color: black;">运用</span>人工智能就比较麻烦了,一旦<span style="color: black;">显现</span><span style="color: black;">反常</span>结果,不<span style="color: black;">晓得</span>是什么<span style="color: black;">原由</span>,<span style="color: black;">亦</span><span style="color: black;">没</span>法debug。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">写这篇帖子是<span style="color: black;">由于</span>前些天看到一篇纽约时报的<span style="color: black;">文案</span>“AI能学会解释它自己吗?”,非常有意思。一位心理学家Michal Kosinski把20万社交网络账号(是个约会网站)的照片及个人信息(<span style="color: black;">包含</span><span style="color: black;">非常多</span>内容,如性向)输入面部识别人工智能系统,<span style="color: black;">发掘</span>人工智能在只看到照片的<span style="color: black;">状况</span>下判断性向准确率很高。人工<span style="color: black;">经过</span>照片判断一个人<span style="color: black;">是不是</span>同性恋的准确率是60%,比扔硬币高一点,但人工智能判断男性<span style="color: black;">是不是</span>同性恋准确率高达91%,判断女性低<span style="color: black;">有些</span>,<span style="color: black;">亦</span>有83%。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">从照片里是看不到音色语调、体态、<span style="color: black;">平常</span><span style="color: black;">行径</span>、人际关系之类<span style="color: black;">帮忙</span>判断的信息的。同性恋有纯相貌方面的特征吗?我个人的经验是,靠相貌判断不大<span style="color: black;">可靠</span>。我以前认识一对男同,<span style="color: black;">她们</span>都是很man的那种,常年健身,待人彬彬有礼但绝<span style="color: black;">无</span>女气,从外表是看不出来的。<span style="color: black;">亦</span>可能是依靠某种服饰特点?表情?背景?人工智能从照片中到底看到了什么<span style="color: black;">咱们</span>人类很可能忽略了的特征,<span style="color: black;">或</span>人类<span style="color: black;">基本</span>看不到的特征,并达到91%的准确率呢?不得而知,反正只是<span style="color: black;">晓得</span>AI看得挺准。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">不可</span>解释自己的AI<span style="color: black;">没</span>法被信任</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">这种黑箱特征有时候倒<span style="color: black;">没</span>关紧要,像是判断一下性向。有时候就<span style="color: black;">不可</span>这么草率了,<span style="color: black;">例如</span>看病。虽然AI系统诊断某些癌症的正确率<span style="color: black;">已然</span>达到人类<span style="color: black;">大夫</span>的水平,但最后结论,<span style="color: black;">日前</span>还是要<span style="color: black;">大夫</span>做,尤其是AI<span style="color: black;">不可</span>告诉<span style="color: black;">咱们</span>它诊断的理由是什么的时候。除非以后AI能解释它<span style="color: black;">为何</span>做出这个诊断,<span style="color: black;">否则</span>让人类100%信任AI是有比<span style="color: black;">很强</span>心理<span style="color: black;">阻碍</span>的。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">前几天<span style="color: black;">刚才</span>看到<span style="color: black;">资讯</span>,新加坡政府<span style="color: black;">起始</span>测试<span style="color: black;">没</span>人驾驶公共汽车。这显然是个正确的方向,我<span style="color: black;">亦</span>相信不久的将来就会<span style="color: black;">作为</span>现实。虽然自动驾驶汽车事故率比人低,理性上<span style="color: black;">咱们</span>都<span style="color: black;">晓得</span>其实更安全,但过马路时,停在旁边的公共汽车<span style="color: black;">无</span>司机,我会不会有点提心吊胆,怕它<span style="color: black;">忽然</span><span style="color: black;">起步</span>?开车时扭头一看,旁边的Bus<span style="color: black;">无</span>司机,我会不会吓一跳,下<span style="color: black;">认识</span>地离它远点?<span style="color: black;">最少</span>初期会的吧。和几个<span style="color: black;">伴侣</span>聊起这个事,都是理性上相信,感性上心虚。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">以前的程序是依靠确定性和因果关系运行的,<span style="color: black;">例如</span>搜索算法中<span style="color: black;">那些</span>页面特征是排名<span style="color: black;">原因</span>,各占多少权重,这是工程师挑出来的、确定的,虽然挑的时候可能<span style="color: black;">便是</span>拍脑袋决定的,但经过监测效果、<span style="color: black;">调节</span>参数,会达到一个比较满意的平衡。人工智能系统并不依靠工程师给定的确定因果,而是更<span style="color: black;">善于</span>于在概率和<span style="color: black;">关联</span>性中找到联系。对人<span style="color: black;">来讲</span>,以概率和<span style="color: black;">关联</span>为特征的判断,经常就<span style="color: black;">欠好</span>解释理由了,<span style="color: black;">例如</span><span style="color: black;">亦</span>许是看心情,<span style="color: black;">亦</span>许是看好看<span style="color: black;">欠好</span>看。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">需求</span>AI系统解释自己的判断,不仅是心理上的问题,<span style="color: black;">亦</span>许以后会变成伦理和法律上的问题,像看病。再<span style="color: black;">例如</span><span style="color: black;">触及</span>用户利益的事情,像贷款,人工智能<span style="color: black;">按照</span>一大堆数据做出拒绝贷款的决定,银行却<span style="color: black;">不可</span>解释<span style="color: black;">为何</span>拒绝,对用户该怎么交代?今年欧盟可能就要颁布法规,<span style="color: black;">需求</span><span style="color: black;">设备</span>做出的决定<span style="color: black;">必要</span>有解释。这对Google、Facebook等<span style="color: black;">全世界</span>性的<span style="color: black;">公司</span>是个压力。在<span style="color: black;">非常多</span><span style="color: black;">行业</span>,如军事、法律、金融,所有决定都是要有人来承担责任的,<span style="color: black;">倘若</span>某个决定<span style="color: black;">没</span>法解释<span style="color: black;">原由</span>,恐怕<span style="color: black;">亦</span><span style="color: black;">无</span>人敢承担这个责任。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">另一个<span style="color: black;">必须</span>AI解释理由的<span style="color: black;">原由</span>是,前面<span style="color: black;">说到</span>,人工智能看的是概率和<span style="color: black;">关联</span>性,但看<span style="color: black;">关联</span>性做决定有时候会<span style="color: black;">引起</span>严重错误。纽约时报的<span style="color: black;">文案</span>举了个例子。经过数据训练的人工智能系统辅助医院急诊室分诊,总体上看效果不错,但<span style="color: black;">科研</span>人员还是不敢真的拿来实用,<span style="color: black;">由于</span>数据中的<span style="color: black;">关联</span>性可能误导人工智能做出错误判断。<span style="color: black;">例如</span>数据<span style="color: black;">显示</span>,<span style="color: black;">身患</span>肺炎的气喘病人最后<span style="color: black;">痊愈</span><span style="color: black;">状况</span>好于平均水平,这个<span style="color: black;">关联</span>性是真实存在的。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">倘若</span>AI系统<span style="color: black;">由于</span>这个数据就给有肺炎的气喘病人比较低的处理等级,那可能就要出事了。<span style="color: black;">由于</span>这些病人之<span style="color: black;">因此</span>最后<span style="color: black;">状况</span>良好,是<span style="color: black;">由于</span><span style="color: black;">她们</span>一来就被给予最高等级,得到最好最快的治疗了。<span style="color: black;">因此</span>,有时候从<span style="color: black;">关联</span>性看不到真正的<span style="color: black;">原由</span>。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">可解释的人工智能</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">X.A.I.(Explainable AI)可解释的人工智能,是<span style="color: black;">刚才</span>兴起的一个<span style="color: black;">行业</span>,目的<span style="color: black;">便是</span>让AI对自己的判断、决定和过程做出解释。去年美国国防高级<span style="color: black;">科研</span>计划局(Darpa )推出了David Gunning博士领导的XAI计划。Google<span style="color: black;">亦</span>依然是这个<span style="color: black;">行业</span>的领先者,Deep Dream<span style="color: black;">好似</span><span style="color: black;">便是</span>这方面<span style="color: black;">科研</span>的一个副<span style="color: black;">制品</span>:</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><img src="https://mmbiz.qpic.cn/mmbiz_jpg/P5rwNx8qkB30Z895dsgUd3selCXhic0R1Ozhic9A3OTL0foomibwtfIt8o8dD4hlFYIZvIJqFBsDCqeDv662ibYfWw/640?wx_fmt=jpeg&tp=webp&wxfrom=5&wx_lazy=1&wx_co=1" style="width: 50%; margin-bottom: 20px;"></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">人工智能与SEO</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">回到搜索算法及SEO,搜索引擎之<span style="color: black;">因此</span>还<span style="color: black;">没</span>法全面应用人工智能,其中一个原因<span style="color: black;">亦</span>许<span style="color: black;">便是</span>人工智能的判断<span style="color: black;">无</span>解释、<span style="color: black;">没</span>法理解,<span style="color: black;">倘若</span>算法<span style="color: black;">运用</span><span style="color: black;">日前</span>的人工智能,一旦<span style="color: black;">显现</span>排名<span style="color: black;">反常</span>,工程师们将<span style="color: black;">没</span>法<span style="color: black;">晓得</span><span style="color: black;">原由</span>是什么,就更<span style="color: black;">没</span>法<span style="color: black;">晓得</span>该怎么<span style="color: black;">调节</span>。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">我想自动驾驶是最先AI实用化的<span style="color: black;">行业</span>之一,和能否解释<span style="color: black;">亦</span>有<span style="color: black;">必定</span>关系。自动驾驶汽车的大部分决定是不大<span style="color: black;">必须</span>解释的,<span style="color: black;">或</span>说解释是一目了然的,距离前车太近<span style="color: black;">因此</span>要减速<span style="color: black;">或</span>刹车,这类判断应该不<span style="color: black;">必须</span>进一步解释理由了。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">SEO们大概都有过<span style="color: black;">一样</span>的疑惑,某个竞争对手的页面看着没什么特殊的,内容不怎么样,视觉设计<span style="color: black;">通常</span>,外链普通,页面优化<span style="color: black;">大众</span>做的都<span style="color: black;">同样</span>,<span style="color: black;">为何</span>排名就<span style="color: black;">那样</span>好呢?<span style="color: black;">此刻</span>的搜索算法还<span style="color: black;">能够</span>探究<span style="color: black;">原由</span>,搜索工程师们大概有内部<span style="color: black;">工具</span><span style="color: black;">能够</span>看到排名的<span style="color: black;">恰当</span>性。<span style="color: black;">倘若</span>搜索工程师看着一个挺烂的页面<span style="color: black;">便是</span>排在前面,却<span style="color: black;">亦</span>不<span style="color: black;">晓得</span><span style="color: black;">原由</span>,还<span style="color: black;">没</span>从查起,<span style="color: black;">她们</span>的内心可能就焦虑了。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">XAI的<span style="color: black;">科研</span>才<span style="color: black;">刚才</span><span style="color: black;">起始</span>,这给了SEO们最后的缓冲期。从人工智能系统在其它<span style="color: black;">行业</span>碾压人类的表现看,一旦大规模应用于搜索,作<span style="color: black;">坏处</span>和黑帽SEO恐怕将<span style="color: black;">作为</span>过去,<span style="color: black;">此刻</span>的常规SEO工作<span style="color: black;">亦</span>许变得<span style="color: black;">没</span>足轻重,SEO们<span style="color: black;">必须</span>回到网站的本质:<span style="color: black;">供给</span>有用的信息或产品,别<span style="color: black;">没</span>他法。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;"><span style="color: black;">亦</span>许你还想看:</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><a style="color: black;">Zac | Google算法更新大全 「外贸及英文站必读」</a></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><a style="color: black;">Zac | 百度算法更新大全</a></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><a style="color: black;">Zac | 追逐算法更新何时了?</a></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><a style="color: black;">Zac | 2017年Google SEO排名<span style="color: black;">原因</span></a></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><a style="color: black;">Zac | 什么是吸引外链的正确方式?</a></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><img src="https://mmbiz.qpic.cn/mmbiz_png/P5rwNx8qkB0M7ZaGeZP5pcK5yhrAtK1qC15lSmp3s6v3NE2SyKw3aPibRKqlriakbd2oYXubswS8hZia6e88FME5w/640?wx_fmt=png&tp=webp&wxfrom=5&wx_lazy=1&wx_co=1" style="width: 50%; margin-bottom: 20px;"></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><strong style="color: blue;"><span style="color: black;"><span style="color: black;">一块</span>学习最新的互联网营销推广技术</span></strong></span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><strong style="color: blue;">把握最新营销动态</strong></span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">转载 l 合作 l 投稿</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;"><span style="color: black;">微X</span>号:280129275</span></strong></p><span style="color: black;"><strong style="color: blue;"><span style="color: black;">扫码,免费加入<span style="color: black;">咱们</span>的社群</span></strong></span>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><img src="https://mmbiz.qpic.cn/mmbiz_jpg/P5rwNx8qkB3bIHxJ5b9FWjgvptEQNbwicNSb6icoVG7icVt0tiadNE1YcleaHPia8T3p2wg1laUM2vmwRqibdicvHzIicw/640?wx_fmt=jpeg&tp=webp&wxfrom=5&wx_lazy=1&wx_co=1" style="width: 50%; margin-bottom: 20px;"></p>
楼主节操掉了,还不快捡起来! 你说得对,我们一起加油,未来可期。
页:
[1]