人工智能在過去幾年中取得了巨大的進步，但對它當前的功能，也有些宣傳過度的嫌疑。

這是上周五舊金山的一個專家座談會得出的結論。該座談會是由美國計算機學會在頒發(fā)第50屆圖靈獎期間舉辦的，該獎項主要獎勵的是在計算機科學領域做出杰出貢獻者。

加州大學伯克利分校的機器學習專家、計算機科學教授邁克爾·喬丹表示，當前人們對所謂“聊天機器人”的功能宣傳得有些過頭了。很多此類軟件使用的都是一種叫“深度學習”的技術，人們會用海量的會話數(shù)據(jù)對它進行“培訓”，使它知道如何與真人進行互動。

盡管好幾家大型科技企業(yè)和創(chuàng)業(yè)公司都推出了能像真人一樣回答問題的聊天機器人，不過喬丹認為，由于人類的語言極其復雜，機器人憑借深度學習等現(xiàn)有的技術，依然無法完全掌握我們的語言。這些機器人本質(zhì)上其實是在耍“社交手腕”，他們可以針對特定語境做出寬泛的回應，但是它們“說不出關于現(xiàn)實世界的任何真實情況”。

喬丹表示：“我們進入了一個機器學習被炒得火熱的時代。”機器學習雖然有改變整個經(jīng)濟的面貌的潛力，但“我們還沒到那個時候。”

會上，谷歌云機器學習首席科學家、斯坦福大學教授李飛飛表示：“我們生活在人工智能的一個最激動人心、也是被炒得最火熱的時代。”李飛飛參與發(fā)起了ImageNet計算機視覺挑戰(zhàn)賽，這次挑戰(zhàn)賽又讓人工智能重新火了一把。參賽者要運用機器學習技術，在照片中識別諸如貓之類的物體。

李飛飛表示，雖然大家都在談論ImageNet的成功，“但我們很少討論失敗。”她還強調(diào)，為了制造出能像真人一樣“看”東西的計算機，研究人員們付出了難以想象的艱辛。

不過李飛飛仍然信心十足地表示，目前人工智能領域已經(jīng)取得了不少具有里程碑意義的成就，最終它們將為我們帶來更多的突破，其影響會觸及諸如醫(yī)療保健等每一個行業(yè)。她表示：“我們正在進入人工智能的一個新階段。”

OpenAI公司是一家由伊隆·馬斯克注資的人工智能研究公司。該公司的研究總監(jiān)伊利婭·蘇特斯科娃指出，深度學習技術要想取得更多突破，首先取決于計算機硬件技術的持續(xù)發(fā)展，比如英偉達的GPU等等，如此才能確保人工智能系統(tǒng)以比以往更快的速度數(shù)據(jù)海量數(shù)據(jù)。深度學習技術將繼續(xù)與計算機硬件一道飛速發(fā)展，目前看來還沒有絲毫減緩的跡象。

蘇特斯科娃表示：“計算能力是深度學習技術的‘氧氣’。”（財富中文網(wǎng)）

譯者：樸成奎

Artificial intelligence has made great strides in the past few years, but it’s also generated much hype over its current capabilities.

That’s one takeaway from a Friday panel in San Francisco involving leading AI experts hosted by the Association for Computing Machinery for its 50th annual Turing Award for advancements in computer science.

Michael Jordan, a machine learning expert and computer science professor at University of California, Berkeley, said there is “way too much hype” regarding the capabilities of so-called chat bots. Many of these software programs use an AI technique called deep learning in which they are “trained” on massive amounts of conversation data so that they learn to interact with people.

But despite several big tech companies and new startups promising powerful chat bots that speak like humans when prodded, Jordan believes the complexity of human language it too difficult for bots to master with modern techniques like deep learning. These bots essentially perform parlor tricks in which they respond with comments that are loosely related to a particular conversation, but they “can’t say anything true about the real world.”

“We are in era of enormous hype of deep learning,” said Jordan. Deep learning has the potential to change the economy, he added, but “we are not there yet."

Also in the panel, Fei-Fei Li, Google’s (goog, +0.89%) machine learning cloud chief and Stanford University Professor, said “We are living in one of the most exciting and hyped eras of AI.” Li helped build the ImageNet computer-vision contest, which spurred a renaissance in AI in which researchers applied deep learning to identify objects like cats in photos.

But while everyone talks about ImageNet’s success, “we hardly talk about the failures,” she said, underscoring the hard work researchers have building powerful computers that can “see” like humans.

Still, Li is excited that current AI milestones will eventually lead to more breakthroughs that will touch every single industry, like healthcare. “We are entering a new phase in AI,” she said.

What will help usher more breakthroughs in deep learning will be the continuing advancements in powerful computing hardware, like Nvidia's GPUs that make it possible to crunch tremendous amounts of data faster than ever, explained Ilya Sutskever, the research director of Elon Musk-backed AI research group OpenAI. Deep learning will keep booming in tandem with advancements in computing hardware that shows no signs of slowing down.

"Compute has been the oxygen of deep learning," Sutskever said.