48 12345
發新話題
打印

[國際新聞] 奧巴馬警告:大量工作將被人工智能取代 AI

【技術跟人走】杜克大學研發的十億像素超級相機如何落入中國手中?
https://hk.finance.yahoo.com/news/%E3%80%90%E6%8A%80%E8%A1%93%E8%B7%9F%E4%BA%BA%E8%B5%B0%E3%80%91%E6%9D%9C%E5%85%8B%E5%A4%A7%E5%AD%B8%E7%A0%94%E7%99%BC%E7%9A%84%E5%8D%81%E5%84%84%E5%83%8F%E7%B4%A0%E8%B6%85%E7%B4%9A%E7%9B%B8%E6%A9%9F-030020586.html

五年前,一批杜克大學(Duke University)的科學家利用美國國防部的贊助開發了一款十億級像素相機,以為美國海軍提供遠程監控。

2016年杜克大學負責上述研究的首席研究員David Brady移居中國開啟他的業務之後,這項從未被美國政府採用的技術現在正被中國警方使用,可以幫助識別近一個足球場那麼遠的距離之外的人員。

在中國更容易獲得啟動資金、製造供應鏈以及該國對高科技相機的需求不斷擴大等因素吸引了David Brady,他最初在美國的初創公司未能贏得財務支持者和客戶。在移居中國的兩年內,他的公司獲得了製造首款商用相機的足夠資金。

這個項目的東移讓人們了解到中國正如何成為人工智能(AI)等創新技術的全球參與者。中國正越來越多地利用其經濟實力獲得海外人才和專業技術。《華爾街日報》(Wall Street Journal)對VentureSource數據的分析顯示,中國投資者牽頭的AI風險融資去年已飆升至25億美元,而五年前基本沒有這方面的投資。中國國家和地方政府也已將資金投入到私募風險投資基金。

美國政府官員對中國試圖到2030年超過美國成為全球AI領導者的野心感到擔憂。特朗普政府上個月承諾增加聯邦支出以保持美國在科技行業的領先地位。依據處理緊急國家安全問題的法律,美國財政部正提議阻止中國公司收購美國先進技術,同時一些國會議員已經呼籲加強對一些美國公司的出口管制,他們稱這些公司的技術可能被中國警方及其他安全部門使用。

涉及AI監控的初創公司正在中國蓬勃興起,每年中國政府向公共安全項目的投入達到300億美元,包括一個旨在覆蓋公共廣場、主要道路口以及火車站的龐大攝錄網絡。為了滿足這一需求,David Brady的安科迪智能技術有限公司(Aqueti China Technology Inc.)開發了配備19個鏡頭的Mantis設備,其處理器可以將圖像集合到一個1億像素的畫面框內,使用者可以對任何特殊的細節進行放大。

這些相機與人臉識別技術相結合,幫助警方進行人員識別,成為政府追蹤犯罪分子以及普通公民的監控網絡的一部分。安科迪的中國合伙人王惠東(William Wang)表示,安科迪的相機目前已經安裝在北京的天安門廣場附近、上海附近的崑山市主要街道上以及其他城鎮。

王惠東幫助安科迪從一名上海政府前官員那裡獲得了約500萬美元早期投資,後者現在營運着一間風投基金。這名投資者表示一直在尋找可以帶回中國的技術。王惠東表示,安科迪在兩次融資中籌集了約2,800萬美元資金,遠超此前在美國籌到的資金規模。安科迪2013年曾嘗試在眾籌網站Kickstarter籌集2.5萬美元,但最終只籌得1,007美元。

為了獲得以上投資,擔任杜克大學崑山校區光電教授的David Brady採取了一條不太尋常的路徑。他沒有創立一家合資企業,而是將其最初的美國業務併入安科迪中,並獲得了使用該相機技術的許可。杜克大學擁有這一相機技術專利。

David Brady表示:「我們還能在什麼地方製造這些相機呢?這自然是一個中國項目。」他透露除了融資,製造這些相機的供應鏈也在中國,即使是在美國籌集到資金,也會花在中國。

杜克大學發言人表示,美國國務院已批准該技術可作為商用技術出口。David Brady表示,杜克大學已告訴他不能向贊助最初原型機的美國國防部防務高等研究計劃署(Defense Advanced Research Projects Agency, 簡稱Darpa)申請任何新資助。

Darpa表示其對該研究的資助已於2015年3月結束,安科迪對該技術的商用開發與其無關。安科迪仍然面臨獲得足夠買家的任務,以便在主要被中國巨頭(很多與政府有聯繫)主導的市場上賺錢。不過安科迪仍在對生產進行投資:該公司在中國有50名員工,並計劃擴大其相機的生產規模。該公司的相機每部售價約1.5萬美元。

仍為安科迪第一大股東的David Brady表示,他對自己助力中國監控系統並無顧慮。批評者稱中國的監控系統會引發侵犯人權的問題。他對此表示:「一個政府並不需要藉助科技來實施壓迫。」
EL34

TOP

Google AI預測病人死期95%準確
https://hk.news.appledaily.com/international/daily/article/20180621/20426719

人工智能(AI)威力越來越厲害,Google指他們用AI技術預測入院病人死期,準確率高達95%,比真人醫生還要準。

分析21.6萬份資料
Google屬下「醫學大腦」(Medical Brain)(圖)團隊早前與美國多家大學研究員發表論文,指他們使用模仿人類大腦神經網絡的AI系統,分析21.6萬個入院紀錄資料,去預測入院病人會在留院期間死亡、留院逾一星期或30日內再次入院等,在預測留院期間死亡方面準確率達95%,高於醫生以一般方法預測的86%準確率,預測長期留院和再入院準確率分別為86%和77%。亦高於一般方法的76%和70%。

在其中一個病例,AI系統在分析逾17.5萬項資料後,指一名患轉移性乳癌女病人,留院期間死亡或然率是20%,醫院採取常用的「早期預警計分法」,則計出她有9%機會死,結果她在兩周內離世。研究員指AI預測準確度較優勝,是因為醫生一般只會用病歷中有限資料作預測,AI則會將醫院所蒐集得的所有病人資料一併考慮。
EL34

TOP

AI威脅逼近!銀行潛在裁員力度遠超金融風暴時
http://money18.on.cc/finnews/news_breaking_content.html?section=exp&article_id=bkn-20180612192206445-0612_00842_001

外電報道,花旗集團的機構客戶業務行政總裁福雷斯(Jamie Forese)表示,佔花旗投行員工總數近五分之二的營運崗位「最適合由電腦來處理」。他預計未來5年內,其2萬名技術與運營人員中,最多將有一半被裁員。

福雷斯的講法是呼應了德意志銀行前陀手John Cryan的說法,他稱德銀最多可能會有一半的員工被技術取代。

市場預料,人工智能(AI)發展,更多的銀行崗位被機器取代,整個行業的潛在裁員力度將遠甚於2007年至2017年。根據英國《金融時報》的研究,在那十年間,全球前十大投行中的八家總計裁員近6萬人。

銀行崗位「智能化」的呼聲並為一家之言,巴克萊投行部負責人思羅斯比稱,未來投行將僱傭更少的員工賺更多的錢,機器將接手「價值較低的任務」。高盛同時指出這一進程不會很快終結。

人工智慧應用開始搶走銀行人員飯碗
https://www.digitimes.com.tw/tech/dt/n/shwnws.asp?CnlID=9&cat=310&cat1=&cat2=&id=533654&packageid=12960&ct=2
EL34

TOP

MIT研發念力控制機械人 讀心知是否做錯任務
http://hk.on.cc/int/bkn/cnt/news/20180622/bknint-20180622210327492-0622_17011_001.html

判讀腦病掃描影像 國產AI戰勝25位名醫
https://hk.news.yahoo.com/%E5%88%A4%E8%AE%80%E8%85%A6%E7%97%85%E6%8E%83%E6%8F%8F%E5%BD%B1%E5%83%8F-%E5%9C%8B%E7%94%A2ai%E6%88%B0%E5%8B%9D25%E4%BD%8D%E5%90%8D%E9%86%AB-225031648.html

全球首場判讀腦病掃描影像的「人機大戰」日前落幕,人工智能(AI)以逾8成的正確率完勝25位全球腦神經科名醫。專家認為,AI暫無法取代醫生工作,但可協助醫生提高工作效率。

由中國國家神經系統疾病臨床醫學研究中心等研發的醫療AI「BioMind天醫智」,日前在北京對戰25名來自全球的神經系統名醫。

「人機大戰」分為A、B兩組進行︰A組共225道題,對顱內腫瘤的磁力共振掃描(MRI)影像和電腦斷層掃描(CT)影像進行判讀,參賽的15名醫生用了30分鐘完成答題,正確率為66%。而「天醫智」僅用15分鐘完成答題,正確率高達87%。

B組是針對腦血管病的CT影像、MRI影像進行判讀,並預測血腫情況。「天醫智」與10名醫生,需要在30分鐘內完成30道題,AI再度以83%的正確率勝過人類醫生的63%。

正確率逾8成

北京天壇醫院神經影像學中心主任高培毅指,目前天壇醫院的醫生「看片」診斷時間達18個小時,而「天醫智」僅需不到10分鐘,「希望未來AI能把醫生解放出來,讓醫生有更多時間做研究,也令患者節省看病成本」。

哈佛醫學院神經外科兼放射學教授亞歷珊德拉·戈爾比認為,AI的定位是輔助作用,醫生可參考AI的判讀結果,避免可能發生的誤差。而隨着診斷效率的提升,人類醫生在人文關懷方面的優勢會愈來愈顯著,因此AI是無法取代人類醫生的。

[ 本帖最後由 EL34 於 2018-7-4 21:58 編輯 ]
EL34

TOP

華監控延伸至牆外 盡掃全球8000網站
http://orientaldaily.on.cc/cnt/china_world/20180705/00178_005.html

中國網絡監控實力進一步提升,據境外媒體報道,內地政府已應用新推出的檢測軟件,將網絡監控從本土防火牆翻到牆外,並可滲透到世界各大網站。該軟件不僅在內地建立一萬八千個輿情監控據點,還可監測港澳台等境外八千多個涉敏感資訊的網站。

據報道,中科點擊(北京)科技有限公司研發的「軍犬網絡輿情監控系統」,可監測新聞、論壇、博客、微博、圖片、視頻、QQ群、搜索及文檔等。

軍犬可採53語言資訊

系統還可採集中、英、法、西班牙文、以及維吾爾語、蒙古語、藏語等總共五十三種語言的輿論資訊。

據該公司介紹,軍犬網絡輿情監控系統可助用家從海量互聯網數據中,滿足網絡輿情監測工作所需,現時已為二百六十八個政府客戶、五十一間企業客戶進行網絡輿情監控。

旅居德國的蒙古族維權領袖席海明表示,這反映出當局對民眾的監控已擴大至海外,消息令人擔憂。
EL34

TOP

China is determined to steal A.I. crown from US and nothing, not even a trade war, will stop it
https://www.cnbc.com/2018/05/04/china-aims-to-steal-us-a-i-crown-and-not-even-trade-war-will-stop-it.html

China's 2030 plan envisions building a $1 trillion A.I. industry.
Investors poured $4.5 billion into more than 200 Chinese A.I. companies between 2012 and 2017.
The biggest A.I. venture deal ever was completed last month when Alibaba led a $600 million deal for China-based facial-recognition start-up SenseTime.

As U.S. and Chinese officials engage in highly anticipated trade talks, officials from China have asserted that it will not discuss two of the biggest trade demands from the United States. One is about the U.S. trade deficit; the other is an issue that could become the greatest technology war in history: China's push into artificial intelligence.

The United States has good reason to be concerned about China's hard stance. While the ongoing trade war is grabbing all the headlines, it's the tussle for dominance in the A.I. space that could shape the economic fortunes of the two world powers. Overshadowed by the dazzling A.I. advances made by the United States so far, China has been silently but resolutely building an ecosystem that is feeding and fueling its ambition to become a world leader in A.I. by 2030.

Home to tech behemoths like Google, Microsoft, IBM and Apple, the United States is where the bulk of A.I. innovation has taken place. However, there are growing indications that China, with its own army of tech heavy hitters such as Alibaba, Tencent, and Baidu, is moving rapidly to close the gap. For one, the Asian economic giant has all the ingredients it needs to upstage Silicon Valley: generous government coffers, large population, a thriving research community and a society eager for technological change. Its investment in A.I., chips and electrics cars combined has been estimated at $300 billion.

In line with its 2030 vision, the government of Tianjin, a city a couple of hours from Beijing, plans to build a $5 billion fund to support the A.I. development. Money being no object, China is also building a giant $2.1 billion technology park to facilitate A.I. innovation. In a red-hot market for tech, China's A.I. start-ups can raise funds with relative ease. Investors poured $4.5 billion into more than 200 Chinese A.I. companies between 2012 and 2017, according to a white paper produced by Kai-Fu Lee, a former Google and Microsoft executive who now leads a venture-capital firm, Sinovation Ventures.

China's goal is to foster a $1 trillion A.I. industry by 2030. Last month Chinese A.I. start-up SenseTime raised $600 million in a deal led by Alibaba, reported as the largest-ever in the A.I. space. The deal gave SenseTime an implied valuation in some reports of more than $3 billion, or even as high as $4.5 billion.

"SenseTime is the perfect case study in the difference between Chinese and Western technological development," wrote Nicholas Colas, co-founder of DataTrek Research, in a recent report. "Artificial intelligence, and especially the A.I. that powers visual analytics, is a critical technology for a raft of new products. The Western companies have their own advantages, to be sure. But the Chinese model of government sponsorship and private capital is coming on very strong. SenseTime may be the hit investment of the moment coming out of this approach, but it certainly won't be the last."

US research sector could struggle
By contrast, the U.S. research sector could be struggling for funding and policy support under the Trump administration. The American Association for the Advancement of Science said the White House planned to slash science and technology research funding by 15 percent in 2018. Worse, with the recent immigration clampdown, the United States may soon be struggling to attract and retain highly skilled tech experts from around the world that it needs to keep Silicon Valley at the cutting edge of A.I. research and innovation.

There are indications America's grasp of A.I. primacy may already be slipping. According to the White House's National Artificial Intelligence Research and Development Strategic Plan in A.I. research, China had already surpassed the United States, at least in terms of journal articles that mention "deep learning" or "deep neural network," as far back as 2016.

"Sadly, when it comes to science and innovation, the U.S. is moving in reverse by cutting funding for research, denying climate change and cutting investments in education," said Vivek Wadhwa, a distinguished fellow and adjunct professor at Carnegie Mellon University's College of Engineering and author of The Driver in the Driverless Car: How Our Technology Choices Will Create the Future.

U.S. leaders do not appear to be aware of A.I. developments, said Joshua Gans, business professor at the University of Toronto and co-author of Prediction Machines: The Simple Economics of Artificial Intelligence. "President Obama discussed it [AI] on numerous occasions," he said. "[Research funding cuts] is obviously bad news in terms of its ability to nurture scientific leadership. It is shortsighted and will harm the U.S. in the medium-long term."

China's timeline for global A.I. supremacy by 2030 may appear a tad overambitious, but opinions are divided as to whether it's achievable.

Wadhwa, for one, feels China has many obstacles. "Governments can't make innovation by throwing money at it — this only leads to more corruption and bureaucracy," he said. "Innovation comes from people who have diverse ideas, take risks and challenge authority."

Wadhwa's reservation contrasts with the conviction of William Weightman, a Fulbright Fellow researching intellectual property law in China. "While 12 years seems like a short amount of time to achieve such an ambitious target, it's not outside China's scope," said Weightman. "The world should not underestimate China's ability to mobilize a vast amount of resources to accomplish its goals."

He used the example of the first high-speed rail line connecting Beijing and Tianjin, which was completed just in time for the 2008 Beijing Olympics. "Between the vast amount of state resources and the determination of the central leadership, China has a solid foundation on which to build an innovative A.I. sector," Weightman said.

China's potential advantages are many
China's demographics give it an unmatched advantage. The Asian giant has large consumer data (which fuels A.I.), scant regulation restricting the use of it, a supportive government both in terms of policy and funding, a population not overly concerned about privacy and a vigorous tech start-up culture that now boasts one-third of the world's unicorns — start-ups valued at $1 billion or more.

Wadhwa conceded China clearly had an advantage in data, the key to training today's A.I., but said its importance may be overstated. "There will be a new generation of technologies that don't require as much data," he said. "New A.I. techniques which [will] work much differently than today's."

It may also turn out that other things are more important than data, like a culture of innovation and scientific research. And while there is uncertainty around government support for scientific research, "the U.S. still has the most vibrant innovative economy," Gans said. "It also has leadership in science on this front, and if it can nurture that, it can compete."

China still trails the United States in areas such as A.I. research talent and algorithm development, according to Weightman.
EL34

TOP

【科技戰】華為全方位AI大計曝光 命名「達文西計劃」
https://hk.finance.appledaily.com/finance/realtime/article/20180713/58436733

《The Information》引述消息人士透露,華為最高管理層與員工討論如何將AI融入於所有華為的產品,由基站、雲端資料中心以至手機等等,計劃取名"Project Da Vinci",部分高層稱之為「D計劃」。

達文西計劃的存在從來未被報道過,但可能對於華為產品將來帶來重大影響,其中包括發展華為最新AI晶片,再以程式透過聲音及影像去確認雲端上的資料,不過,這計劃亦可能觸動美國政府的神經。

有分析認為,華為透過「D計劃」,試圖擺脫對於處理器公司Nvidia的倚賴。
KT88

TOP

 48 12345
發新話題