原創譯文 |對話谷歌首席數據官沙克特——機器學習將如何發展

轉載聲明

本文為燈塔大數據原創內容，歡迎個人轉載至朋友圈，其他機構轉載請在文章開頭標註：

「轉自：燈塔大數據；」

人工智慧當下的發展勢頭可謂是「炙手可熱」，在科技領域也掀起了一場激烈的角逐。亞馬遜、微軟、谷歌和IBM紛紛開始投資人工智慧項目，研發出的人工智慧應用可以說是五花八門，從無人駕駛車到新型癌症治療法，逐漸在各個領域滲透發展。

雖然大部分科技巨頭紛紛加入這場人工智慧競賽，但仍有不少人認為，目前階段處於領先地位的還是谷歌。

谷歌當前的戰略是兼并人工智慧領域的創業公司，將人才和資源收入麾下，以實現谷歌人工智慧方面的發展。

不久前，谷歌的首席執行官桑德爾·皮查伊（Sundar Pichai）對外表示谷歌正在向「以人工智慧為首位的公司」轉型。

去年三月份，谷歌子公司Deepmind的產品AlphaGo在與圍棋冠軍李世石的對弈中，以4:1完勝李世石，獲得了世界媒體的廣泛關注，也讓人們開始注意到谷歌人工智慧方面的發展。

沙克特·庫爾瑪（Saket Kumar）是谷歌現任的首席數據官。他在創新分析領域有著超過15年的工作經驗，在將數據轉化成洞察力來進行科學決策方面造詣頗深，被奉為思想領袖。

在加入谷歌之前，沙克特就已經在廣告業、石油天然氣行業、醫療保健和製造業等多個領域的分析項目中取得了成功。加入谷歌之後，他帶領的數據學家團隊專註於為谷歌的頂級客戶提供更加高效的營銷策略。

沙克特·庫爾瑪博士將出席2017年6月5日至6日在美國舊金山馬里奧特聯合廣場舉辦的「機器學習創新峰會」，記者有幸在他演講之前對他進行了採訪。

記者：您認為在近期，最重要的機器學習應用是什麼？

沙克特：這個問題很難回答啊。現在很多行業和消費活動都進行了電子化，由此產生的數據量也在不斷增加。這對機器學習來說是一件好事，因為現在有了更多的數據集和案例能夠用來分析和學習。

比如說圖像識別、音頻轉寫、不同語言間翻譯等等。我認為近期最重要的機器學習應用可能當屬消費者行為分析應用了。

現在谷歌、臉書、亞馬遜和其他科技巨頭都已經擁有了足夠多的數據，並且已經打造了各自的知識庫，使他們能夠夠產出自己的機器學習解決方案。

記者：日前，KDnuggets網站發起一項投票，有51%的投票者認為在未來十年內，人類數據學家的大部分預測分析和數據科學方面的工作能夠交給機器。

您認為數據學家這個角色正面臨威脅嗎？您認為科技的發展會對人才市場造成什麼樣的影響呢？政府是否準備好迎接這個新時代了呢？

沙克特：不可否認的趨勢是會有越來越多的自動化工具和改進工具幫助人們完成預測分析。但是，我並不認為這會威脅到人類數據學家的工作機會。

機器將代替他們去做那些枯燥的數據處理和清除工作，基礎的數據分析和建模可能也將由機器自動完成。

儘管如此，對於那些懂得數據學、演算法等深度領域知識的人來說，他們根本不用愁工作的問題，因為他們還有將這些基於數學分析的信息傳達給商業決策者的重要任務。

記者：在機器學習領域，有哪些新技術或者新想法是讓您覺得特別感興趣，或者特別看好的呢？

沙克特：現在視頻和多媒體消費領域也採用了機器分析，這是讓我特別激動的。圖像和視頻識別技術還在持續改進中。在這個方面機器學習能夠提供很大幫助，因為機器學習能夠分析視頻，並給出真實的消費者互動反饋。

記者：在未來幾年，您預計機器學習行業會遇到哪些挑戰？您認為該如何克服這些困呢？

沙克特：目前機器學習行業出現了很多好的趨勢，如計算和儲存的成本都在降低。但是很多公司依然存在缺陷。比如說，有不少公司招不到優秀的數據學家。

很多大公司，特別是位於矽谷的那些公司，都面臨著招不到合適人才的困難，因為現在這方面的人才實在是不多，供他們選擇的人才範圍很小。

英文原文

Interview With Saket Kumar, Chief DataScientist At Google

'The most important applications willlikely be analyzing consumer behavior'

The race to lead the way in AI is hottingup. Amazon, Microsoft, Google, and IBM are among those to have invested heavilyin research of the technology, with applications ranging from driverless carsto improved cancer treatment.

Arguably leading the way is Google, whosemain focus has been on acquiring innovative startups in the field and bringingthem under their umbrella.

Sundar Pichai, Google』s chief executive officer,recently said that the company was 『really transitioning to becoming anAI-first company.』

Perhaps its most eye catching demonstration was the victorylast year of Google Deepmind』s AlphaGo over Go star Lee Sedol, but even moreexciting things are happening behind the scenes.

Dr. Saket Kumar is Chief Data Scientist atGoogle. He has more than 15 years experience as an innovative analyticspractitioner and thought leader, with a focus on translating data into insightsfor decision makers.

He has led successful analytics assignments in multipleindustries, including advertising, oil & gas, healthcare, andmanufacturing. At Google, he leads a team of data scientists focused onimproving marketing effectiveness for top tier clients.

We sat down with him ahead of hispresentation at the Machine Learning Innovation Summit, taking place this June5-6 at the Marriott Union Square in San Francisco.

Where do you think machine learning』s mostimportant applications will be in the near future?

This is a hard question. We see tons ofbusiness and consumer activities being digitized. The amount of data that getsdigitized continues to grow.

Machine learning is great for situations wherethere are large data sets and cases to learn from. Examples of this includeImage identification, voice transcription, translation etc.

The most importantapplications will likely be analyzing consumer behavior as companies likeGoogle, Facebook, Amazon, and others have tons of such data and have developedlarge knowledge base that they can leverage to build ML solutions.

In a recent KDnuggets poll, 51% ofrespondents said that they expect most expert-level Predictive Analytics/DataScience tasks currently done by human Data Scientists to be automated to happenwithin the next decade.

Do you think the data scientist』s role is really underthreat? What kind of impact do you think it will have on the job market ingeneral, and are governments prepared?

There is going to be automation andimprovement in the tools that help with predictive analytics. However, I do notsee any threat to the work done by human data scientists.

A lot of drudgery ofprocessing and cleaning data will hopefully go away. The base analysis andmodeling is likely going to be commoditized.

Despite this, there will still bea role for people who know data, algorithms, deep domain knowledge, and caneffectively communicate math based insights to business leaders.

Are there any new technologies or ideas inthe machine learning space that you find particularly exciting or believe willbe especially important in the next few years?

I am excited about the intersection ofvideo/multi-media consumption and analytics. Image and video recognition isstill work in progress.

There is a lot of exciting stuff that can be done withrespect to what ML sees in videos and actual consumption/interaction responseof the consumers.

What challenges do you foresee holding backmachine learning from achieving its potential? How do you think these could beovercome?

There are a lot of positive trends(computing and storage costs going down). There are still data silos with andacross organizations.

One obvious one is the lack of qualified data scientists.Most companies - with the exception of large silicon valley companies -struggle to get right talent as the pool to draw from is not large.

翻譯：燈塔大數據

閱讀原文了解更多詳