2017年大数据年中盘点：预测的趋势现况如何

2016年底，我们对大数据在新一年的发展趋势做出了预测。当时与我们的“大数据书呆子网络”进行了充分讨论，最后得到了一份预测报告。现在2017年已经过去一半了，我们想回顾一下我们去年年底时发布的预测，看看当时水晶球告诉我们的结果到底准不准确。

2017年已经过半，是时候回顾一下前半年大数据行业的发展情况。

　　 预测一：机器学习、人工智能和物联网将普及

现实：2017年以来，机器学习，人工智能，以及物联网行业的发展速度一直居高不下，不论是从消费者角度还是从商业角度来看，这三者的应用领域得到了迅速拓展。

IBM的Watson已经家喻户晓，人工智能威胁论时不时就霸占世界各地新闻头条，也有不少人吹捧人工智能是我们抵御日益猖獗的网络攻击的重要防线。

同样地，物联网的发展也令人瞩目，越来越多的家庭和企业采用互联的设备，促进了物联网的快速发展。

物联网的一大成功案例是信息技术服务供应商“EPAM系统”，2017年上半年该公司股票已经升值36%，以及“Skyworks”公司，在上半年其股价增长了47%，在股市的成功表明物联网市场正在蓬勃发展，这些物联网领域的大公司运营情况良好，也足以证明物联网正走进大众视线，成为人们日常生活的一部分。

预测准确性：准确

　 预测二：人工智能准确度加强

现实：人工智能绝对是2017年的一个焦点话题，已经从早期的模糊概念发展成现实中的应用，在此发展过程中，人工智能的准确度在不断增强。

最好的证明是人工智能在医疗领域的应用，如乳腺X光检测、心脏病和中风治疗方面，人工智能的应用提高了治疗的精准性。

在疾病诊断测试中，人类医生得分平均为0.728，而人工智能的分数为0.745到0.764之间（1表示100%准确）。人工智能还能用于预测儿童自闭症的发生概率，在症状出现之前就发现疾病，这在以前是不可想象的。

但是人工智能准确度加强也带来了消极影响，比如在大选中人工智能就扮演了一次不折不扣的黑暗元素。

“CambridgeAnalytica”公司在英国脱欧公投期间，就利用精准人工智能和心理战策略，利用下流手段干预投票，该公司在美国大选期间，还发表了“唐纳德·特朗普，我们的人工智能总统”文章来迷惑大众。

预测准确性：准确

　　 预测三：公司需准备迎接快速运营（operateatspeed）时代

现实：这点我们现在很难去评定它到底准不准确，因为我们当时预测的是说，相较于投入人力财力发展内存技术和量子计算技术，越来越多的公司会选择集中精力扩大市场份额。

因此现在很难说这些公司是否在积极准备中，但是可以肯定的一点是不少数据平台都在积极发展快速运营方面的业务，而很多大公司，像Hortonworks、IBM和SAP现在也都在提供内存服务，未来很可能有越来越多的公司会采用他们的服务。

　 预测四：行业专业化减弱

现实：这项预测现在我们也不好说到底准确不准确，因为2017年才过去一半，我们尚未找到客观的、可量化的方式来衡量这个预测的准确性。

可以肯定的一点是，我们已经看到了一些零零星星的证据来为这一趋势作证，还有我们熟悉的来自不同领域的专家也表达了类似的观点。

这些专家拥有数据应用相关经验，而不仅仅局限于某个领域，这和他们过去“各自为战”没有团队合作的情况完全不同。

越来越多的数据学家开始与外界合作，这样他们就不需要再去分散精力了解其他行业，只需要专注于提高数据应用的效率就够了。

预测准确性：尚待确认

　　 预测五：政府将加大对数据的审查力度

现实：毋庸置疑，数据已经成为各国政府关注的焦点领域之一，尤其是美国。

2016年，我们就已经看到有针对数据的政治讨论，从希拉里邮件门，到俄罗斯黑客窃取美国民主党自由委员会（DNC）数据，再到近期“深根分析”公司（DeepRootAnalytics）两千万客户数据泄露，人们对数据安全问题的讨论一直没有平息过。

尽管对世界各国政府来说，数据安全确实已经成为十分重要的议题，但是在过去一年间，还没有任何国家在这方面加强立法。

唯一具有影响力的法案是欧洲《通用数据保护条例》（GeneralDataProtectionRegulation），该条例出台后，欧盟国家的政府和公司都根据要求加强了数据安全保护工作。

2018年是该条例的截止期限，在最后期限到来之前，各国政府和企业还会在数据安全方面投入更多。

预测准确性：准确

英文原文

Big Data Top Trends 2017 - How Are WeDoing?

Half way through the year, we check back tosee how we're doing

In November 2016 we made our predictionsabout what the market would look like in the coming 12 months. We put ourcollective heads together, spoke to our network of data wonks, and put ourideas down on paper.

As we are now halfway through 2017 we wanted to revisitour predictions and see how clear our glance into the crystal ball was.

　 Prediction: Machine Learning, AI, And IoTTo Become Common

Reality: The pace of change within machinelearning, AI, and the IoT has been high in 2017, with huge leaps forward intheir use across the board, both from a consumer and business perspective.

IBM’s Watson has become practically ahousehold name, the ‘threat’ of AI has seen the term become common on frontpages across the world, and it is being touted as a way of protecting the worldagainst the increasing number of cyber attacks.

Equally we have seen the IoT make hugestrides, with an increasing number of people bringing connected devices intotheir homes and companies making significant gains in IoT.

Some of the successstories so far include EPAM Systems, who have seen their shares increase by 36%in the first six months of 2017 and Skyworks who have seen theirs increase by47% in the same time.

The success of these stocks signifies that the market isbuoyant and some of the key companies in the sector are performing well, whichis a clear indication that the IoT is becoming increasingly common, both inpeople’s homes and perception.

Accuracy: Correct

　 Prediction: Increased AI Accuracy

Reality: AI has taken centre stage in 2017,moving from a relatively indistinct concept to reality, and alongside this,there has been an increased accuracy.

One of the key manifestations of this hasbeen within the medical field, with treatments like Mammograms, heart attacks,and strokes predicted considerably more accurately thanks to AI usage.

In testsconducted against doctor-led diagnosis, AI scored between 0.745 and 0.764 (with1 being 100% accuracy) compared to 0.728 from doctors. It can also predictautism in young children, before symptoms manifest, something that wasimpossible before.

It has also had more negative consequences,with AI at heart of the darker elements of recent elections. Its accuracy usehas been credited with suppressing the votes of certain parts of thepopulation.

This stems from companies like Cambridge Analytica who have beenaccused of underhand methods using precise AI combined with psychologicalwarfare techniques (https://www.theguardian.com/technology/2017/may/07/the-great-british-brexit-robbery-hijacked-democracy)and stemmed headlines like ‘Donald Trump, Our A.I.

President’(https://www.nytimes.com/2017/05/22/opinion/donald-trump-our-ai-president.html).

Accuracy: Correct

　 Prediction: Companies Will Need To PrepareTo Operate At Speed

Reality: This is a prediction that isdifficult to gauge given that we predicted that rather than actively adoptingin-memory and quantum computing techniques, companies would instead bepreparing for their increased presence.

It is therefore hard to say whether ornot companies are actively preparing, but there have certainly been moves fromcompanies who provide many of the data platforms used that suggests they areincreasingly moving towards this kind of service.

Large players likeHortonworks, IBM, and SAP all now offer some kind of in-memory service that islikely to be adopted by companies in the future.

Accuracy: This is probably wrong, ifplatforms are preparing to offer it as part of their package, there will belittle need for companies to do the same.

　 Prediction: Less Industry Specialization

Reality: This is again a prediction that ishard to get an accurate read on half way through the prediction period becausethere aren’t really any objective and quantifiable ways to find solid numberson the subject.

What is certain is that we have seen anecdotal evidence ofthis, with several of our speakers from the past 6-12 months working acrossmany different industries throughout their career to date.

It means that they have experience relatedto the use of data rather than the use of data within a specific industry,which has typically been the case in the past, especially when they wereworking as individuals rather than in teams.

As more data scientists have begunto take a collaborative approach, there has been less of an onus onunderstanding of the business and more on actually using the data effectively.

Accuracy: TBC

　 Prediction: Government Scrutiny On Data

Reality: There is little doubt that datahas taken center stage in terms of government attention, especially in the US.

We have seen it become the biggest talking point in politics over the lastyear, from Hillary Clinton not having good enough data security, Russianhackers stealing data from the DNC, and recently a leak from Deep RootAnalytics which released details from 200 million users.

However, despite it becoming a huge issuefor governments across the world, there is little additional legislation thathas been passed this year that is likely to have an impact on companies.

Considerable work being done in the area though, primarily as a result of theGDPR (General Data Protection Regulation), an EU legislation that requirescompanies and governments who hold data on EU nationals to protect that data inrobust ways.

The approaching 2018 deadline for this has meant that despite alack of original legislation in the area, it is clear that a considerableamount is being done in terms of preparation for it.

Accuracy: Correct

 

责任编辑：陈近梅