作為微軟的首席技術(shù)官,凱文·斯科特面臨的挑戰(zhàn)在于:如何讓微軟在各種科技趨勢中立于不敗之地。
斯科特于兩年前成為了微軟的首席技術(shù)官,在此之前曾執(zhí)掌領(lǐng)英的軟件工程業(yè)務(wù),長達六年。2016年,微軟以260億美元收購了領(lǐng)英。他在微軟的首項任務(wù)便是弄清楚公司紛繁復(fù)雜的業(yè)務(wù)部門所使用的各種科技,并且衡量其實用性,然后確保每個部門都能用上最受歡迎的技術(shù)。
斯科特說,這一舉措反映了微軟首席執(zhí)行官薩蒂亞·納德拉所奉行的理念。納德拉花費了大量的時間來思考“該做什么”,以及“我們會因不做什么而感到后悔?!?/p>
微軟錯過了幾場技術(shù)革命,尤其是智能手機的崛起,而其對手蘋果和谷歌則把握住了這一機會。納德拉希望斯科特能夠確保類似事情不會再次上演。
在以下經(jīng)過編輯的《財富》采訪中,斯科特討論了人工智能、微軟繼續(xù)開發(fā)混合現(xiàn)實技術(shù)[微軟對虛擬和增強現(xiàn)實技術(shù)的統(tǒng)稱]的舉措,以及深度學(xué)習(xí)所帶來的挑戰(zhàn)。
《財富》:微軟的人工智能與其他公司有何不同?
凱文·斯科特:從骨子里來講,微軟是一家平臺公司。聽過比爾·蓋茨長期以來對平臺公司定義的人可能會知道,平臺公司致力于打造能夠創(chuàng)造各種機遇的技術(shù),而利用這些機遇的公司有著各自不同的經(jīng)濟價值觀。我們正在做大整個市場的規(guī)模。例如,個人電腦帶來了一個巨大的經(jīng)濟機遇。在我們看來,人工智能基本上也是如此。
當我想到平臺時,我會想到Windows,其他公司可以利用這一系統(tǒng)打造各種應(yīng)用。你覺得人工智能也是如此嗎?
如今我們所力推的事情是:眾多人工智能技術(shù)如今依然十分復(fù)雜,很多公司難以跟上其步伐。如今,數(shù)以萬計的開發(fā)人員可能都是機器學(xué)習(xí)/數(shù)據(jù)科學(xué)的堅實擁躉。幾乎我們所接觸過的每一位客戶都在思考如何使用人工智能來幫助其更好地經(jīng)營業(yè)務(wù)。我們不能指望所有公司都能請得起那些專業(yè)領(lǐng)域的博士和機器學(xué)習(xí)工程師。按照當前人工智能行業(yè)的發(fā)展速度,現(xiàn)有的人才數(shù)量難以滿足市場需求。
我們所面臨的一個挑戰(zhàn)是打造能夠降低準入門檻的科技,這樣,更多的開發(fā)人員便可以在其產(chǎn)品和服務(wù)中使用機器學(xué)習(xí)。在這一領(lǐng)域,微軟自身可謂是自成一派,因為我們有約5.5萬名開發(fā)人員,但并非所有人都是機器學(xué)習(xí)/數(shù)據(jù)科學(xué)專家。
我覺得,學(xué)會接受“很多實驗都將失敗”這個事實是探索深度學(xué)習(xí)人工智能技術(shù)的公司所面臨的一個挑戰(zhàn)。
作為平臺提供商,微軟有責(zé)任為人們提供更好的工具,引導(dǎo)人們走上能夠通往成功的康莊大道。
我覺得,人們確實得有心理準備,因為會有失敗的情形。人們必須抱著試驗的心態(tài)來從事這件事情。它不是定理證明,并非只是按照固定的套路就能搞定,而且不會出什么意外。它更像是實驗室的工作。
那些深諳技術(shù)的公司也都已經(jīng)適應(yīng)了這種“實驗然后失敗”的流程。我們只是通過自己的努力了解到,第一個實驗沒有獲得成功,接下來還得再接再厲。當你獲得成功的時候,在實驗中付出的所有成本都是值得的。
你在人工智能方面都有哪些經(jīng)歷?
我正在寫一本有關(guān)人工智能的書,內(nèi)容涉及人們?yōu)槭裁磻?yīng)該樂觀看待存在人工智能的未來。而且反過來講,我認為人工智能甚至對美國農(nóng)村地區(qū)的人也是有百利而無一害。
我來自弗吉尼亞中部坎貝爾郡農(nóng)村的一個貧窮家庭,我所在的這個小鎮(zhèn)名為格拉迪斯。為了寫書,一年前我回訪了小鎮(zhèn)。鎮(zhèn)上所有的工業(yè)早在數(shù)年前就已經(jīng)消失。煙草、紡織品、家具制造業(yè)都不見了。但當?shù)爻霈F(xiàn)了一些有意思的行業(yè),其中一些便是基于人工智能和高級自動化。
格拉迪斯發(fā)生了什么變化?
在我的同學(xué)中,有的家里連續(xù)5代人都是煙草種植者。在煙草市場沒落后,他們的生意便一落千丈,不得不另謀出路。這些人有著不俗的創(chuàng)業(yè)頭腦,而且認為科技將在其業(yè)務(wù)中發(fā)揮重要的作用。
他們以前用于種植煙草的地塊如今已經(jīng)種上了草皮,其單位經(jīng)濟效益并不比煙草差。部分原因在于他們使用了多種高級自動化拖拉機以及較為復(fù)雜的科技,并運用它們在異常遼闊的土地上種植草皮。這個工作的勞動力需求比種植煙草更大,但借助這一技術(shù),他們雇用的人數(shù)沒有發(fā)生變化。因此科技并沒有削減工作崗位。
在地平線上,你可以看見無人機從作物上飛過,從空中進行檢查。這并不是說不需要人工,而是可以讓無人機更加頻繁地飛越地塊,并獲得更多有關(guān)地塊情況的數(shù)據(jù),以便讓人們更好地調(diào)整肥料和水。
得益于這一技術(shù),公司無需為了獲得最佳單位經(jīng)濟效益而建造一個大型工廠,并雇員數(shù)千名員工。人們可以在當?shù)卦O(shè)立一家企業(yè),然后雇用30名員工,然后將這個只有30人的弗吉尼亞坎貝爾郡企業(yè)發(fā)展成為一家國際性公司。有人認為,如果存在100家企業(yè),每個企業(yè)有1萬名員工,工作機會并不會回到城鎮(zhèn),然而,如果當?shù)負碛?0萬家企業(yè),每家公司有100個高技能崗位,情況就會不一樣。
這類工作的薪資更高嗎?
是的,確實如此。
有人擔(dān)心,盡管自動化會讓公司更加高效,但受益的是管理層,不是雇員。
我覺得這兩種事情都可能發(fā)生,因此我們應(yīng)該謹慎對待。在寫這本書時,包括與沃爾瑪這種規(guī)模的客戶以及中小企業(yè)的客戶進行交談時,我發(fā)現(xiàn)有很多事情都值得我們期待。
虛擬現(xiàn)實和增強現(xiàn)實似乎在三年前異?;鸨?,但由于回報速度較慢,如今很多風(fēng)投資本家已不再像以前那樣關(guān)注這一行業(yè)。如果某項技術(shù)并沒有出現(xiàn)預(yù)期的熱度,你會采取什么樣的措施,并如何調(diào)整?
我的工作職責(zé)之一就是確保公司能夠長期維持對某些投資的專注和重視??梢源_定的是,微軟并未減少在混合現(xiàn)實方面的投資[微軟制作了HoloLens增強現(xiàn)實頭戴設(shè)備]。不但沒有減少,反而還有所增加,雖然不是大幅增長,但確實在增長。
如果你自認為是一家平臺公司,你就必須對平臺的未來進行構(gòu)思。我們認為有三種技術(shù)將發(fā)展為重要的平臺,只是它們目前正處于不同的開發(fā)階段。
一個是量子計算,它在今后將變得非常重要;再就是混合現(xiàn)實,我們認為它將先于量子計算,成為一個異常重要的平臺;然而在這之前,“智能邊緣”這一理念將大行其道,人們可以將其看作是物聯(lián)網(wǎng)[與互聯(lián)網(wǎng)相連的設(shè)備]、傳感器和人工智能的結(jié)合體。
我們認為,上述三項事物將成為未來極為重要的平臺。同時,如果要讓某個全球性的平臺得以運轉(zhuǎn),人們必須進行投資,并相信它是可行的,而且它的實現(xiàn)只是時間上的問題,而不是能否的問題。(財富中文網(wǎng)) 譯者:馮豐 審校:夏林 |
As Microsoft’s chief technology officer, Kevin Scott has the challenging job of keeping his company atop all of the tech trends.
Scott became Microsoft’s CTO two years ago after six years directing software engineering at LinkedIn, which Microsoft bought in 2016 for $26 billion. One of his first jobs at Microsoft was to identify all the technologies used by the company’s sprawling business units—to gauge their usefulness—and then make sure that the popular ones were available to every division.
The exercise reflects the philosophy of Microsoft CEO Satya Nadella, Scott says. Nadella not only spends a lot of time thinking about what to do, but also “what are we not doing that we’re going to regret.”
Microsoft missed out on a few tech revolutions, in particular the rise of smartphones, which rivals Apple and Google ended up capitalizing on. Nadella wants Scott to make sure that nothing similar happens again.
In this edited interview with Fortune, Scott talks about artificial intelligence, Microsoft’s continued push into mixed reality [Microsoft lingo for both virtual and augmented reality tech], and the challenges of deep learning.
Fortune: How do you distinguish Microsoft’s AI from other companies?
Kevin Scott: We’re a platform company by DNA. If you listen to how Bill Gates has always defined what a platform company is, it’s one that builds technology that creates all of this opportunity in which you don’t have all of the economic value concentrated in one company. We’re increasing the overall size of the pie. The PC, for instance, created an enormous economic opportunity. We see AI as essentially the same thing.
When I think of platforms, I think of things like Windows, which other companies can build apps on top of. Is this how you see AI?
The thing that we’re pushing hard on is that a lot of AI right now is still unnecessarily difficult for many people to get up to speed on. There are maybe in the high tens of thousands of developers out there who are hardcore machine learning/data science folks. Almost every customer we interact with is thinking about using AI to help its business run better. And you can’t expect each and every one of them to hire a bunch of Ph.Ds and machine learning engineers. At the rate AI is unfolding, not enough of those people exist.
One of our challenges is to build technologies that lower the barriers to entry so a much larger pool of developers can use machine learning in their products and services. Microsoft itself is a microcosm for this because we have about 55,000 developers in the company and not all of them are machine learning/data science experts.
I imagine it’s a challenge for companies exploring the AI technique of deep learning to get used to the idea that a lot of their experiments will fail.
It’s incumbent upon us as platform providers to give people better tools—to guide you in better ways toward paths that get you to success.
I think you do have to expect some of this stuff not to work. You have to get into it with this experimental mindset. It’s not like you’re proving a theorem and you walk through the steps and it’s done and predictable. It’s more like lab science.
The most technologically-savvy companies are used to this trial-and-error process. We just know through our own efforts that the first thing is not going to work, and you have to push and push. When you get the win, it totally covers all of the costs of the experimentation.
What’s your background in AI?
I’m writing a book on AI right now. It’s about why we should be optimistic about a future that includes AI. The contrarian thing is that I think it’s net beneficial even to people in rural parts of the country.
I was a poor kid from rural central Virginia—Campbell County, a little town called Gladys. I went back there a year ago for the book. All the industry there evaporated years ago. Tobacco, textiles, furniture manufacturing all went poof. But some interesting things are emerging there now, some of which is powered by AI and advanced automation.
What’s going on in Gladys?
I went to school with people whose families’ have been tobacco farmers for five generations. Their business basically went sideways when the tobacco markets collapsed, and they had to figure out what to do. They were fairly entrepreneurial and they knew technology would play a role in what they were doing.
All the land that they used to plant tobacco on is sod now, and the unit economics is about as good as tobacco. Part of the reason is that they use a bunch of advanced automation—tractors, and fairly sophisticated technology to let them grow sod on these very large tracks of land. It’s more labor intensive than tobacco was, but with the technology they have about the same number of employees. So technology hasn’t reduced jobs.
On the horizon are things like drones that can fly over crops to do aerial inspections. It’s not that you don’t need a human being, but you can fly over it more frequently and get more data about what’s going on in your field so you can better adjust fertilizers and water.
Because of the technology, you don’t need a giant factory with thousands of people in order to just get your unit economics right. You can start a business and have 30 people working in this place and have that 30-person business in Campbell County, Va. be a global business. Some people believe you won’t have jobs coming back where there are 100 companies with 10,000 jobs a piece, but you’ll have 100,000 companies with 100 higher-skilled jobs each.
Will those jobs pay more?
Yeah. I know for sure.
Some people are concerned that while automation will make companies more efficient, only management will benefit and not the workers.
I think both can happen and I think we should be cautious. What I’ve seen working on this book and talking with customers the size of Walmart all the way down to small and medium sized businesses is that there’s lots of things to be hopeful about.
Virtual reality and augmented reality seemed really big three years ago, and now many venture capital investors aren’t as focused on it because they couldn’t get returns fast enough. How do you plan for and adjust when a technology hasn’t caught on as fast as hoped?
Part of my job is making sure that we maintain our focus and our commitment to some of these investments over long periods. The thing I can say is we have not reduced our investments in mixed reality [Microsoft makes the HoloLens augmented reality headset]. If anything, we increased things—not dramatically up, but it’s growing.
If you’re thinking of yourself as a platform company, you have to be thinking about what the future platforms are going to be. We have three things that we believe are going to be important platforms that are in different stages of development.
One is quantum computing, which at some point is going to be very important. There’s mixed reality, which we think is probably in a shorter time horizon is going to be a very important platform. And on a shorter time horizon than that, this notion of an intelligent edge, which you can think of as a mashup of IOT [Internet-connected devices], sensors, and AI.
We believe all three of those will be extremely important platforms in the future. And to make a global scale platform work, you have to invest and believe it’s real. It’s a question of when and not if. |