IBM沃森服務(wù)將登陸亞馬遜、微軟、谷歌等云平臺(tái)
多年來,IBM使用沃森(Watson)為企業(yè)客戶提供數(shù)據(jù)分析服務(wù)——但需要在IBM的云端使用?,F(xiàn)在,為了拓展沃森的覆蓋范圍,IBM的這項(xiàng)服務(wù)將登陸競(jìng)爭(zhēng)對(duì)手的云平臺(tái)。
IBM于周二表示,將允許企業(yè)將部分與沃森有關(guān)的軟件用于存儲(chǔ)在競(jìng)爭(zhēng)對(duì)手企業(yè)云數(shù)據(jù)中心的底層數(shù)據(jù),包括微軟的Azure和亞馬遜網(wǎng)絡(luò)服務(wù)??蛻暨€可以將沃森用于內(nèi)部數(shù)據(jù)中心的數(shù)據(jù)。
“這種做法將使該服務(wù)更加公開,這尚屬首次。”IBM的數(shù)據(jù)和人工智能總經(jīng)理羅伯·托馬斯說。
這一舉措標(biāo)志著IBM的一次轉(zhuǎn)變,到目前為止,這種轉(zhuǎn)變似乎也只是開放云計(jì)算服務(wù)和沃森技術(shù)。例如,2016年,IBM時(shí)任首席財(cái)務(wù)官、現(xiàn)任全球市場(chǎng)高級(jí)副總裁馬丁·施勒特在財(cái)報(bào)電話會(huì)議上對(duì)分析師說:“沃森在我們的云上運(yùn)行,我們的技術(shù)要在自己的云上運(yùn)行?!?/p>
但I(xiàn)BM正在調(diào)整戰(zhàn)略,提升沃森的吸引力。IBM之所以改變心意,因?yàn)楣镜墓苍品?wù)市場(chǎng)份額縮水,最多只能在亞馬遜網(wǎng)絡(luò)服務(wù)和微軟Azure后排名第三位。
Hurwitz&Associates的董事總經(jīng)理丹·基爾希稱IBM的新服務(wù)“非常重要”,因?yàn)槠髽I(yè)更愿意使用不依賴于單一供應(yīng)商的技術(shù)。
“如果告訴客戶,如果所有的東西都買我們的,只能用我們的軟件和我們的資源,就會(huì)和客戶脫節(jié)。”基爾希說。
研究公司451 Research的聯(lián)合創(chuàng)始人及研究副總裁尼克·佩申思稱此舉意味著“IBM承認(rèn)現(xiàn)在是混合云的世界”,他指的是公司希望使用多家云計(jì)算供應(yīng)商,希望一些軟件能在自己內(nèi)部數(shù)據(jù)中心運(yùn)行。
亞馬遜和微軟等云計(jì)算公司正在推動(dòng)人工智能的發(fā)展,因?yàn)槭褂蒙疃葘W(xué)習(xí)等技術(shù)通常需要通過云服務(wù)處理海量數(shù)據(jù),佩申思說。IBM決定向?qū)κ止镜脑品?wù)開放沃森,這可能表明“IBM承認(rèn)亞馬遜網(wǎng)頁服務(wù)、微軟和谷歌在機(jī)器學(xué)習(xí)中處于領(lǐng)先地位”,他說。
基爾希表示,這種做法可以幫助IBM贏得那些擔(dān)心被捆綁在同一家公司人工智能項(xiàng)目云服務(wù)中的客戶?,F(xiàn)在他們的選擇相對(duì)更自由。
IBM的新服務(wù)Cloud Private for Data(ICP for Data)基于備受歡迎的開源IT軟件Kubernetes,該軟件由谷歌開發(fā),用于管理數(shù)據(jù)中心的基礎(chǔ)架構(gòu)。這款自由軟件被開發(fā)人員廣泛使用,因?yàn)樗麄兛梢杂盟鼧?gòu)建復(fù)雜的應(yīng)用程序,這些應(yīng)用程序由多個(gè)數(shù)據(jù)庫和服務(wù)提供支持,理論上這些數(shù)據(jù)庫和服務(wù)不必束縛于單個(gè)云計(jì)算服務(wù)供應(yīng)商。
借助這項(xiàng)新技術(shù),IBM客戶可以使用沃森搭建機(jī)器學(xué)習(xí)模型、創(chuàng)建數(shù)字聊天機(jī)器人等。
托馬斯表示,IBM尚未給這項(xiàng)新服務(wù)定價(jià),可能會(huì)根據(jù)客戶的需求進(jìn)行分級(jí)定價(jià)。此外,雖然該服務(wù)目前僅兼容亞馬遜網(wǎng)頁服務(wù)等北美地區(qū)的云供應(yīng)商,但未來可能會(huì)加入中國科技巨頭阿里巴巴等亞洲供應(yīng)商,托馬斯稱。
然而,托馬斯表示,沃森的最佳運(yùn)行方式還是通過IBM自己的云計(jì)算服務(wù)。使用該服務(wù)的客戶可以從多種硬件和計(jì)算機(jī)芯片配置中進(jìn)行選擇,最大限度地發(fā)揮沃森的功能。
“我們對(duì)IBM的云服務(wù)充滿信心,如果客戶在其他地方試用了我們的產(chǎn)品,他們最后會(huì)被IBM的云及其特別之處吸引?!蓖旭R斯說。
雖然有些公司擔(dān)心在使用某些軟件或IT產(chǎn)品時(shí)會(huì)被捆綁在特定公司的云端基礎(chǔ)架構(gòu)中,但佩申思表示他們目前仍處于人工智能使用的初期階段,仍然愿意將本公司的數(shù)據(jù)上傳到亞馬遜網(wǎng)頁服務(wù)等云平臺(tái),并依靠它們發(fā)展機(jī)器學(xué)習(xí)軟件。目前,在機(jī)器學(xué)習(xí)項(xiàng)目中綁定特定供應(yīng)商還不是一個(gè)值得擔(dān)心的問題,但I(xiàn)BM認(rèn)為這種情況必將出現(xiàn)變化。
“我們沒有聽到有很多公司砸門說:‘這太過分了,我們要求提升可移植性?!迸迳晁颊f。
他還說道:“你可以說IBM正試圖收復(fù)機(jī)器學(xué)習(xí)領(lǐng)域的失地。多年來,沃森因?yàn)槠渌放频某霈F(xiàn)失去了一些光彩。他們正試圖收復(fù)失地。”(財(cái)富中文網(wǎng)) 譯者:Agatha |
For years, IBM has offered corporate customers its Watson data crunching technology—but only if they used it on IBM’s cloud computing service. Now, in an effort to expand Watson’s reach, IBM is also making it available on competing cloud services.
IBM said Tuesday that it would allow businesses to use some of IBM’s Watson-related software with underlying data that is stored in rival cloud data centers like Microsoft’s Azure and Amazon Web Services. Customers will also be able to use Watson with data stored in their own data centers.
“It is enabling a level of openness that hasn’t been available to date,” said Rob Thomas, IBM general manager of data and AI.
The move marks a departure for IBM, which, until now, hasn’t seemed anything but open with its cloud computing service and Watson technologies. In 2016, for example, then-IBM CFO and current IBM senior vice president of global markets Martin Schroeter told analysts during an earnings call, “Watson runs on our cloud, and our technology will run on IBM’s cloud.”
But IBM is now shaking up its strategy in an effort to broaden Watson’s appeal. The change of heart comes as IBM’s public cloud languishes in third place, at best, in terms of market share behind AWS and Microsoft Azure.
Dan Kirsch, a managing director at Hurwitz & Associates, called IBM’s new service “really significant” because businesses are increasingly seeking technology that’s not dependent on a single vendor.
“You look out of touch with your customers if you say you can only have our software and our capabilities if you adopt everything only from us,” Kirsch said.
Nick Patience, a founder and research vice president for 451 Research, said the move is “an acknowledgment by IBM that it’s a hybrid cloud world,” referring to firms wanting to use more than one cloud computing vendor as well as keeping some software running in their internal data centers.
Cloud computing companies like Amazon and Microsoft are pushing artificial intelligence because using technologies like deep learning typically require processing huge amounts of data through their cloud services, Patience explained. IBM deciding to make Watson available on competing cloud services could be a sign that “IBM is acknowledging that AWS, Microsoft, and Google are leading the pack in machine learning,” he said.
Kirsch explained that being more open could help IBM win customers that are concerned about being locked into one company’s cloud service for AI projects. In this way, they’ll be relatively free to pick and chose.
IBM’s new service, Cloud Private for Data (ICP for Data), is built on the popular open source IT software Kubernetes, which was developed by Google as a way to manage data center infrastructure. The free software has become popular with developers because they can use the technology to build complex apps that are powered by multiple databases and services that theoretically don’t have to be tethered to a single cloud provider.
With the new technology, IBM customers can use Watson to do things like build machine-learning models and create digital chatbots.
Thomas said IBM hasn’t set a price for the new service and it will likely be tiered based on the customers’ needs. Additionally, while the service is currently compatible with only North American cloud providers like AWS, Asian cloud vendors like China tech giant Alibaba could be added in the future, Thomas said.
Still, Thomas argues that the best way to run Watson is through IBM’s own cloud computing service. Customers using it can choose from multiple hardware and computer chip configurations that squeeze the most power out of Watson.
“We are confident in the IBM cloud that if clients try our products anywhere, they will eventually be drawn to IBM cloud and the uniqueness it provides,” said Thomas.
Although companies are concerned about being locked into a specific company’s cloud infrastructure when it comes to certain software and IT products, Patience said they are currently in the early stages of using AI and are still willing to upload their corporate data to cloud services like AWS and rely on them for machine learning software. Being locked into a particular vendor when it comes to machine learning projects isn’t yet much of a concern, but IBM is betting that will change.
“We don’t hear a load of companies banging on the door saying, ‘This is outrageous, we demand more portability,'” Patience said.
He continued: “You could say IBM is trying to take back the initiative in machine learning here. The Watson brand has lost a little bit of luster over the years as others has come along. They are trying to take it back.” |