2017年11月15日 星期三

新科技的美麗與哀愁:AI是阻力還是助力?

新科技的美麗與哀愁:AI是阻力還是助力?

【吳碧娥╱北美智權報 編輯部】
1997 5 11 日,由IBM開發的超級電腦「深藍」(Deep Blue)第一次打敗了當時世界西洋棋冠軍加里·卡斯帕洛夫(Garry Kasparov),成為人工智慧的歷史性事件。台灣IBM公司全球企業諮詢服務事業群總經理賈景光在「眺望2018產業發展趨勢研討會」中指出,只要持續輸入所有棋譜資料,透過人工智慧演算,「電腦或人工智慧打敗人類」這個結果,可說是「回不去了」。不過也不用太悲觀,這只不過是「所有的數據」打敗「一個人」的智慧,未來並不是電腦與人的競爭,人工智慧與人類的關係將發展成「有用AI的人打敗沒有使用AI的人」,主體終究還是「人」。

而身為第一個被人工智慧打敗的西洋棋棋王,卡斯帕洛夫在時隔20年後出版的新書《Deep Thinking: Where Machine Intelligence Ends and Human Creativity Begins》中也呼籲,人們應該更樂觀看待人工智慧,並停止把機器當作對手看待。

在未來五到十年中,IBM 提出的AI AI (Augmented Intelligence) 是以「擴增智慧」為主力發展,賈景光強調,人工智慧是協助人們將工作的一部分自動化,而不是取代人類,透過人工智慧的決策輔助,人們可將其餘的工作做得更好。

在兩三年前,IBM人工智慧品牌華生 (WATSON) 投入試用時,第一個選擇的行業就是醫療產業,當IBM帶著WATSON到醫院請醫生試用時,卻引起醫生大怒,認為是在挑戰醫生的專業和經驗。其實WATSON使用的情境和醫生所想像的並不一樣,WATSON並不是代替醫生下診斷,而是透過事實和證據為基礎的判斷,做為醫生下最後診斷前的第二意見(second opinion)

由於WATSON可以在短短數秒內大量閱讀醫療期刊文獻,這是人類所做不到的。透過資料庫當中巨量的病例資料,WATSON針對病人症狀分析出各種可能的病因,可降低醫生誤診的機率;或是參考其他國家類似病例的用藥結果,提升醫生「用對藥」的機率。經一段時間試用後,超過八成的醫生都願意使用WATSON作為助手,讓病人得到更好的診治。賈景光解釋,AI並不是跟人搶飯碗,電腦的功能就是彙整資料,若人和電腦能各自針對高價值的項目相互協作,才是AI的價值所在。

國內發展AI 仍有「五缺」

國際研究暨顧問機構Gartner發現,從2016年到2017年間,與AI相關的問題諮詢多達4,353件,年成長高達五倍。根據Gartner調查,雖然多數企業都對AI有興趣,但近六成企業仍停留在「資訊蒐集」階段,真正採取行動的目前僅占12%。工研院IEK也在2016年至2017年調查國內六大產業共20家業者發現,業者對於引入人工智慧升級產業有強烈需求,但有許多執行上的共通問題,包括:缺乏對AI的應用、缺乏AI專業人才、缺乏performance認證場域、缺乏個資處理經驗、以及國內市場太小等產業「五缺」。

工研院IEK政府業務服務辦公室計畫副組長楊瑞臨指出,AI最具看好的應用是在「客戶關係與業務行銷管理」,像是客戶關係管理、強化客戶行為預測,不論是B2B或者是B2C都具相當發展潛力;此外,以AI結合IoTAIoT,未來在資源管理以及提升企業的生產力與競爭力上,將是各產業領域的重要發展方向。

楊瑞臨認為,AI將會取代許多人工作的說法應是言過其實,但A支援並協助許多職業提升整體工作成效則可預見。Gartner在今年8月的調查結果顯示,「人才缺乏」是目前人工智慧遇到最大的挑戰,不只影響企業導入AI的時程,甚至是「連人都找不到」。由於AI專業人才不足,導致人才競逐以及磁吸效應所帶來的人才分布不均,供需失衡之下,AI人才價碼水漲船高,也會直接影響AI的應用與改善人類生活及工作的進展期程。企業若能運用線上AI課程及多元的open-source資源,一點一滴逐步導入AI,可降低相當成本與縮短學習曲線。

2020AI機器人市場規模達800億美元

而導入人工智慧,最大目的是降低因應外界變化所需的成本,特別是用在生產製造的機器人上,未來機器人的產業地圖將會因AI而有顯著的改變。IEK預期,涵蓋人工智慧技術的機器人相關產品將以等比級數成長,全球市場規模將在2020年超過至800億美元。工研院IEK機械組機械部分析師黃仲宏指出,隨著人工智慧生態系的不斷擴大,當機器人可透過物聯網(IoT)與網路連結,再加上人工智慧的溝通功能,讓我們看見服務型機器人走向普及的曙光。目前全球主要IT公司產品幾乎都已經與AI脫離不了關係,加上GoogleMicrosoftFBAppleIBM等大廠陸續推出與AI相關的產品,積極搶當人工智慧的龍頭,接著就是要讓AI的價值被機器人彰顯;未來機器人的產業地圖將會因人工智慧而有顯著的改變,而新創公司亦扮演重要角色。

台商不能只做製造和加工

面對新科技,最大的贏家不是加倍複製成功的過去,而是能引導人們轉向成長產業的社會和企業,機器人產業就是其中之一。黃仲宏認為,工業機器人製造者有強者恆強、大者恆大的趨勢,台灣廠商不應只專注於製造和銷售機械手臂產品,儘管CP值高,卻是與德國Kuka、日本Fanuc等現有高市占率的工業機器人製造商硬碰硬,台廠要聚焦於已有的機器人製造發展經驗,與資通訊系統整合,進而運用人工智慧技術讓機器人產品發展後發先至,才能成為未來的優勢產品。


工研院IEK主任蘇孟宗認為,由人工智慧所引領的第四波科技創新正在發生,不論是既有產業的轉型升級,或是新創企業的突破創新,人工智慧都將是發展關鍵,如果能有效運用人工智慧,產業就能提升競爭力。過去台灣在高科技產業的國際分工體系中,扮演製造加工的代工角色,面對AIOT時代來臨,台灣的AI優勢在於製造業的終端資料、各類型資料庫(先進製造、健康醫療等)、及半導體核心運算技術等,應運用優勢扮演垂直整合或生態系領導者的關鍵伙伴,同時透過智慧系統與服務,可望提升製造業附加價值創造,強化供應鏈管理與帶動新需求,也能提高服務業勞動生產力,創造新型態科技服務模式。

Capitalism Has a Problem. Is Free Money the Answer?

Capitalism Has a Problem. Is Free Money the Answer?

By PETER S. GOODMAN NOV. 15, 2017 LONDON —

One need not be a card-carrying revolutionary to deduce that global capitalism has a problem. In much of the world, angry workers denounce a shortage of jobs paying enough to support middleclass life. Economists puzzle over the fix for persistently weak wage growth, just as robots appear poised to replace millions of human workers.

At the annual gathering of the global elite in the Swiss resort of Davos, billionaire finance chieftains debate how to make capitalism kinder to the masses to defuse populism. Enter the universal basic income. The idea is gaining traction in many countries as a proposal to soften the edges of capitalism.

Though the details and philosophies vary from place to place, the general notion is that the government hands out regular checks to everyone, regardless of income or whether people are working. The money ensures food and shelter for all, while removing the stigma of public support.

Some posit basic income as a way to let market forces work their ruthless magic, delivering innovation and economic growth, while laying down a cushion for those who fail. Others present it as a means of liberating people from wretched, poverty-level jobs, allowing workers to organize for better conditions or devote time to artistic exploits. Another school sees it as the required response to an era in which work can no longer be relied upon to finance basic needs.

“We see the increasing precariousness of employment,” said Karl Widerquist, a philosopher at Georgetown University in Qatar, and a prominent advocate for a universal social safety net. “Basic income gives the worker the power to say, ‘Well, if Walmart’s not going to pay me enough, then I’m just not going to work there.’ ” The universal basic income is clearly an idea with momentum.

Early this year, Finland kicked off a two-year national experiment in basic income. In the United States, a trial was recently completed in Oakland, Calif., and another is about to launch in nearby Stockton, a community hard-hit by the Great Recession and the attendant epidemic in home foreclosures. The Canadian province of Ontario is enrolling participants for a basic income trial. Several cities in the Netherlands are exploring what happens when they hand out cash grants unconditionally to people already receiving some form of public support. A similar test is underway in Barcelona, Spain.

A nonprofit organization, GiveDirectly, is proceeding with plans to provide universal cash grants in rural Kenya. As a concept, basic income has been kicked around in various guises for centuries, gaining adherents across a strikingly broad swath of the ideological spectrum, from the English social philosopher Thomas More to the American revolutionary Thomas Paine. The populist firebrand Louisiana governor Huey Long, the civil rights icon Martin Luther King Jr., and the laissez faire economist Milton Friedman would presumably agree on little, yet all advocated some version of basic income.

In a clear sign of its modern-day currency, the International Monetary Fund — not an institution prone to utopian dreaming — recently explored basic income as a potential salve for economic inequality. Not everyone loves the idea. Conservatives fret that handing out money free of obligation will turn people into dole-dependent slackers. In the American context, any talk of a truly universal form of basic income also collides with arithmetic. Give every American $10,000 a year — a sum still below the poverty line for an individual — and the tab runs to $3 trillion a year. That is about eight times what the United States now spends on social service programs. Conversation over.

Labor-oriented economists in the United States are especially wary of basic income, given that the American social safety programs have been significantly trimmed in recent decades, with welfare, unemployment benefits and food stamps all subject to a variety of restrictions. If basic income were to replace these components as one giant program — the proposal that would appeal to libertarians — it might beckon as a fat target for additional budget trimming.

“Tens of millions of poor people would likely end up worse off,” declared Robert Greenstein, president of the Center on Budget and Policy Priorities, a Washington-based research institution, in a recent blog post. “Were we starting from scratch — and were our political culture more like Western Europe’s — U.B.I. might be a real possibility. But that’s not the world we live in.” And some advocates for working people dismiss basic income as a wrongheaded approach to the real problem of not enough quality paychecks. “People want to work,” said the Nobel laureate economist Joseph E. Stiglitz when I asked him about basic income early this year. “They don’t want handouts.” Yet some of the basic income experiments now underway are engineered precisely to encourage people to work while limiting their contact with public assistance.

Finland’s trial is giving jobless people the same amount of money they were already receiving in unemployment benefits, while relieving them of bureaucratic obligations. The bet is that people will use time now squandered submitting paperwork to train for better careers, start businesses, or take part-time jobs. Under the system the trial replaces, people living on benefits risk losing support if they secure other income. In short, basic income is being advanced not as a license for Finns to laze in the sauna, but as a means of enhancing the forces of creative destruction so central to capitalism. As the logic goes, once sustenance is eliminated as a worry, weak companies can be shuttered without concern for those thrown out of work, freeing up capital and talent for more productive ventures.

The trials in the Netherlands, conducted at the municipal level, are similarly geared to paring bureaucracy from the unemployment system. Ditto, the Barcelona experiment. Silicon Valley has embraced basic income as a crucial element in enabling the continued rollout of automation. While engineers pioneer ways to replace human laborers with robots, financiers focus on basic income as a replacement for paychecks.

The experiment in Stockton, Calif. — set to become the first city government to test basic income — is underwritten in part by an advocacy group known as the Economic Security Project, whose backers include the Facebook co-founder Chris Hughes. The trial is set to begin next year, with an undisclosed number of residents to receive $500 a month. The trial in Oakland was the work of Y Combinator, a start-up incubator. Its researchers handed out varying grants to a few dozen people as a simple feasibility test for basic income. The next phase is far more ambitious. The Y Combinator researchers plan to distribute grants to 3,000 people with below-average incomes in two as-yet undisclosed American states. They will hand out $1,000 a month to 1,000 people, no strings attached, and $50 a month to the rest, allowing for comparisons in how recipients use the money, and what impact it has on their lives.

One key element of the basic income push is the assumption that poor people are better placed than bureaucrats to determine the most beneficial use of aid money. Rather than saddle recipients with complex rules and a dizzying array of programs, better to just give people money and let them sort out how to use it. This is a central idea of GiveDirectly’s program in Kenya, where it began a pilot study last year in which it handed out small, unconditional cash grants — about $22 a month — to residents of a single village. The program is now expanding its sights, with plans to hand out grants to some 16,000 people in 120 villages. From a research standpoint, these remain early days for basic income, a time for experimentation and assessment before serious amounts of money may be devoted to a new model for public assistance.


Yet from a political standpoint, basic income appears to have found its moment, one delivered by the anxieties of the working poor combined with those of the wealthy, who see in widening inequality the potential for mobs wielding pitchforks. “The interest is exploding everywhere,” said Guy Standing, a research associate at SOAS University of London. “The debates now are extraordinarily fertile.” 

2017年11月14日 星期二

The Ivory Tower Can’t Keep Ignoring Tech

The Ivory Tower Can’t Keep Ignoring Tech

By CATHY O’NEIL NOV. 14, 2017

These days, big data, artificial intelligence and the tech platforms that put them to work have huge influence and power. Algorithms choose the information we see when we go online, the jobs we get, the colleges to which we’re admitted and the credit cards and insurance we are issued. It goes without saying that when computers are making decisions, a lot can go wrong. Our lawmakers desperately need this explained to them in an unbiased way so they can appropriately regulate, and tech companies need to be held accountable for their influence over all elements of our lives.

But academics have been asleep at the wheel, leaving the responsibility for this education to well-paid lobbyists and employees who’ve abandoned the academy. That means our main source of information on the downside of bad technology — often after something’s gone disastrously awry, such as when we learned that fake news dominated our social media feeds before last year’s presidential election, threatening our democracy — is the media. But this coverage often misses everyday issues and tends to be far too credulous when it does exists.

Much of what should concern us is more nuanced and small scale — and much less understood — than what we see in the headlines. Moreover, we shouldn’t have to depend on journalism to do the tedious, serious work of understanding the problems with algorithms any more than we depend on it to pursue the latest questions in sociology or environmental science. We need academia to step up to fill in the gaps in our collective understanding about the new role of technology in shaping our lives. We need robust research on hiring algorithms that seem to filter out people with mental health disorders, sentencing algorithms that fail twice as often for black defendants as for white defendants, statistically flawed public teacher assessments or oppressive scheduling algorithms.

And we need research to ensure that the same mistakes aren’t made again and again. It’s absolutely within the abilities of academic research to study such examples and to push against the most obvious statistical, ethical or constitutional failures and dedicate serious intellectual energy to finding solutions. And whereas professional technologists working at private companies are not in a position to critique their own work, academics theoretically enjoy much more freedom of inquiry.

To be fair, there are real obstacles. Academics largely don’t have access to the mostly private, sensitive personal data that tech companies collect; indeed even when they study data-driven subjects, they work with data and methods that typically predict much more abstract things like disease or economics than human behavior, so they’re naïve about the effects such choices can have.

The academics who do get close to the big companies in terms of technique get quickly plucked out of academia to work for them, with much higher salaries to boot. That means professors working in computer science and robotics departments — or law schools — often find themselves in situations in which positing any skeptical message about technology could present a professional conflict of interest.

The many data science institutes around the country, which have created lucrative master’s programs to train data scientists, are more focused on trying to get a piece of the big data pie — in the form of collaborations and jobs for their graduates — than they are on asking how the pie should be made. We won’t find any help there. Indeed, while West Coast schools like Stanford and the University of California, Berkeley, are renowned for creating factories that churn out the future engineers and data scientists of Silicon Valley, there are very few coveted permanent, tenure-track jobs in the country devoted to algorithmic accountability.

A final hurdle: There is essentially no distinct field of academic study that takes seriously the responsibility of understanding and critiquing the role of technology — and specifically, the algorithms that are responsible for so many decisions — in our lives. That’s not surprising. Which academic department is going to give up a valuable tenure line to devote to this, given how much academic departments fight over resources already?

There’s one solution for the short term. We urgently need an academic institute focused on algorithmic accountability. First, it should provide a comprehensive ethical training for future engineers and data scientists at the undergraduate and graduate levels, with case studies taken from real-world algorithms that are choosing the winners from the losers.

Lecturers from humanities, social sciences and philosophy departments should weigh in. Second, this academic institute should offer a series of workshops, conferences and clinics focused on the intersection of different industries with the world of A.I. and algorithms. These should include experts in the content areas, lawyers, policymakers, ethicists, journalists and data scientists, and they should be tasked with poking holes in our current regulatory framework — and imagine a more relevant one. Third, the institute should convene a committee charged with reimagining the standards and ethics of human experimentation in the age of big data, in ways that can be adopted by the tech industry.

There’s a lot at stake when it comes to the growing role of algorithms in our lives. The good news is that a lot could be explained and clarified by professional and uncompromised thinkers who are protected within the walls of academia with freedom of academic inquiry and expression. If only they would scrutinize the big tech firms rather than stand by waiting to be hired. Cathy O’Neil is a data scientist and author of the book “Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy.


2017年11月7日 星期二

Waymo’s Autonomous Cars Cut Out Human Drivers in Road Tests

Waymo’s Autonomous Cars Cut Out Human Drivers in Road Tests

By DAISUKE WAKABAYASHI
NOV. 7, 2017 SAN FRANCISCO —

The self-driving car is edging closer to becoming driverless. Waymo, the autonomous car company from Google’s parent company Alphabet, has started testing a fleet of self-driving vehicles without any backup drivers on public roads, its chief executive said Tuesday. The tests, which will include passengers within the next few months, mark an important milestone that brings autonomous vehicle technology closer to operating without any human intervention.

Dozens of companies are testing self-driving technology on public roads across the United States and some autonomous features are available in today’s cars. But Waymo is believed to be the first company to test vehicles on public roads without a driver ready to take over in an emergency. “Our ultimate goal is to bring our fully self-driving technology to more cities in the U.S. and around the world,”

John Krafcik, Waymo’s chief, said in prepared remarks at a technology conference in Portugal on Tuesday. “Fully self-driving cars are here.” The tests are a show of engineering prowess by Waymo at a time when traditional automakers and other tech companies like Uber race to develop similar vehicles. Waymo is limiting the trials to a region around Phoenix, where it has been conducting a ride-testing program this year, and plans to expand the testing area over time.

The company said it planned to use the driverless vehicles to launch a commercial ride-hailing service for the general public, but did not offer any detail on when, where or how. Waymo said its driverless cars hit the public roads last month. The company did not say whether it was testing the driverless cars in environments considered challenging for autonomous vehicles, like bridges or tunnels, or more difficult conditions, like driving at night or in rain and snow — usually not a big concern in the dry Phoenix climate.

While the prospect of cars without emergency drivers may raise concerns among some passengers, Waymo said it had confidence in the safety of its self-driving technology. It has included backup systems like a secondary computer to take over if the main computer fails. And though the cars are driverless, they are not entirely without humans, at least for now.

Waymo employees sit in the back seat of the cars, monitoring them, a company spokesman, Johnny Luu, said. Once passengers join the tests, they will be able to contact Waymo support staff with a button inside the car. If the cars are involved in a crash, they are programmed to respond appropriately, including pulling off the road on their own.

Driverless cars are regulated by a patchwork of state laws. Arizona, like many states, has no restrictions against operating an autonomous vehicle without a person in the driver’s seat. On the other hand, California, where Waymo is headquartered, requires any self-driving car to have a safety driver sitting in the front. In December, Waymo published a report for California’s Department of Motor Vehicles about how frequently its car “disengaged” — deactivating its autonomous mode because of a system failure or safety risk and forcing a human driver to take over. In the report, Waymo said this happened once every 5,000 miles the cars drove in 2016, compared with once every 1,250 miles in 2015.

Consumer Watchdog, a frequent critic of Alphabet, said that data demonstrated that the cars are not ready to drive without any human intervention and that Waymo was following the Silicon Valley model of “beta testing” a new technology on the public. “It’s the wrong approach when you’re dealing with self-driving cars,” said John M. Simpson, a director at Consumer Watchdog.

“When things go wrong with a robot car, you kill people.” Researchers believe self-driving cars can be safer than cars operated by human drivers because they are programmed to adhere strictly to traffic laws, they don’t get distracted, and they usually refrain from taking unnecessary risks. Timothy Tait, a spokesman for the Arizona Department of Transportation, said the state was on pace to exceed 1,000 automobile-related fatalities this year and that its top priority is the public’s safety — particularly by advancing efforts to reduce crashes and deaths on its roads. “We are closely monitoring emerging technologies like self-driving cars that may ultimately support safer travel and open up opportunities for populations who today are unable to drive for themselves,” he said in a statement.

Waymo, which started as a research and development project for Google in 2009, maintains what many in the industry consider a technological advantage over its competitors. Waymo said its autonomous vehicles had driven more than 3.4 million miles on actual roads — with safety drivers — as well as running 10 million miles every day in a virtual simulator. In his remarks, Mr. Krafcik said Waymo sees a ride-hailing taxi service as the first commercial application of the company’s driverless car technology, though there could be other uses in logistics and public transportation.

Taking the human out of the equation will fundamentally change transportation and change how people buy cars, said Mr. Krafcik, who was an executive at Hyundai Motors before joining Google. “Because you’re accessing vehicles rather than owning, in the future, you could choose from an entire fleet of vehicle options that are tailored to each trip you want to make,” he said. “They can be designed for specific purposes or tasks.”



2017年11月5日 星期日

Building A.I. That Can Build A.I.

Building A.I. That Can Build A.I.
Google and others, fighting for a small pool of researchers, are looking for automated ways to deal with a shortage of artificial intelligence experts. 

By CADE METZ
NOV. 5, 2017 SAN FRANCISCO —

They are a dream of researchers but perhaps a nightmare for highly skilled computer programmers: artificially intelligent machines that can build other artificially intelligent machines.

With recent speeches in both Silicon Valley and China, Jeff Dean, one of Google’s leading engineers, spotlighted a Google project called AutoML. ML is short for machine learning, referring to computer algorithms that can learn to perform particular tasks on their own by analyzing data. AutoML, in turn, is a machine-learning algorithm that learns to build other machine-learning algorithms. With it, Google may soon find a way to create A.I. technology that can partly take the humans out of building the A.I. systems that many believe are the future of the technology industry.

The project is part of a much larger effort to bring the latest and greatest A.I. techniques to a wider collection of companies and software developers. The tech industry is promising everything from smartphone apps that can recognize faces to cars that can drive on their own. But by some estimates, only 10,000 people worldwide have the education, experience and talent needed to build the complex and sometimes mysterious mathematical algorithms that will drive this new breed of artificial intelligence.

The world’s largest tech businesses, including Google, Facebook and Microsoft, sometimes pay millions of dollars a year to A.I. experts, effectively cornering the market for this hard-to-find talent. The shortage isn’t going away anytime soon, just because mastering these skills takes years of work.

The industry is not willing to wait. Companies are developing all sorts of tools that will make it easier for any operation to build its own A.I. software, including things like image and speech recognition services and online chatbots. “We are following the same path that computer science has followed with every new type of technology,” said Joseph Sirosh, a vice president at Microsoft, which recently unveiled a tool to help coders build deep neural networks, a type of computer algorithm that is driving much of the recent progress in the A.I. field. “We are eliminating a lot of the heavy lifting.”

This is not altruism. Researchers like Mr. Dean believe that if more people and companies are working on artificial intelligence, it will propel their own research. At the same time, companies like Google, Amazon and Microsoft see serious money in the trend that Mr. Sirosh described. All of them are selling cloud-computing services that can help other businesses and developers build A.I. “There is real demand for this,” said Matt Scott, a co-founder and the chief technical officer of Malong, a start-up in China that offers similar services. “And the tools are not yet satisfying all the demand.”

This is most likely what Google has in mind for AutoML, as the company continues to hail the project’s progress. Google’s chief executive, Sundar Pichai, boasted about AutoML last month while unveiling a new Android smartphone. Eventually, the Google project will help companies build systems with artificial intelligence even if they don’t have extensive expertise, Mr. Dean said.

Today, he estimated, no more than a few thousand companies have the right talent for building A.I., but many more have the necessary data. “We want to go from thousands of organizations solving machine learning problems to millions,” he said.

Google is investing heavily in cloud-computing services — services that help other businesses build and run software — which it expects to be one of its primary economic engines in the years to come. And after snapping up such a large portion of the world’s top A.I researchers, it has a means of jump-starting this engine.

Neural networks are rapidly accelerating the development of A.I. Rather than building an image-recognition service or a language translation app by hand, one line of code at a time, engineers can much more quickly build an algorithm that learns tasks on its own. By analyzing the sounds in a vast collection of old technical support calls, for instance, a machine-learning algorithm can learn to recognize spoken words.

But building a neural network is not like building a website or some run-of-the-mill smartphone app. It requires significant math skills, extreme trial and error, and a fair amount of intuition. Jean-François Gagné, the chief executive of an independent machine-learning lab called Element AI, refers to the process as “a new kind of computer programming.” In building a neural network, researchers run dozens or even hundreds of experiments across a vast network of machines, testing how well an algorithm can learn a task like recognizing an image or translating from one language to another. Then they adjust particular parts of the algorithm over and over again, until they settle on something that works.

Some call it a “dark art,” just because researchers find it difficult to explain why they make particular adjustments. But with AutoML, Google is trying to automate this process. It is building algorithms that analyze the development of other algorithms, learning which methods are successful and which are not. Eventually, they learn to build more effective machine learning. Google said AutoML could now build algorithms that, in some cases, identified objects in photos more accurately than services built solely by human experts.

Barret Zoph, one of the Google researchers behind the project, believes that the same method will eventually work well for other tasks, like speech recognition or machine translation. This is not always an easy thing to wrap your head around. But it is part of a significant trend in A.I. research. Experts call it “learning to learn” or “meta-learning.” Many believe such methods will significantly accelerate the progress of A.I. in both the online and physical worlds.

At the University of California, Berkeley, researchers are building techniques that could allow robots to learn new tasks based on what they have learned in the past. “Computers are going to invent the algorithms for us, essentially,” said a Berkeley professor, Pieter Abbeel. “Algorithms invented by computers can solve many, many problems very quickly — at least that is the hope.” This is also a way of expanding the number of people and businesses that can build artificial intelligence. These methods will not replace A.I. researchers entirely.

Experts, like those at Google, must still do much of the important design work. But the belief is that the work of a few experts can help many others build their own software. Renato Negrinho, a researcher at Carnegie Mellon University who is exploring technology similar to AutoML, said this was not a reality today but should be in the years to come. “It is just a matter of when,” he said.



2017年11月1日 星期三

The Upside of Being Ruled by the Five Tech Giants

NY Times TECHNOLOGY
The Upside of Being Ruled by the Five Tech Giants
Farhad Manjoo
STATE OF THE ART
NOV. 1, 2017

The tech giants are too big. But what if that’s not so bad? For a year and a half — and more urgently for much of the last month — I have warned of the growing economic, social and political power held by the five largest American tech companies: Apple, Amazon, Google, Facebook and Microsoft. Because these companies control the world’s most important tech platforms, from smartphones to app stores to the map of our social relationships, their power is growing closer to that of governments than of mere corporations. That was on stark display this week, when executives from two of the five, Facebook and Google, along with a struggling second-tier company, Twitter, testified before Congress about how their technology may have been used to influence the 2016 election.

Yet ever since I started writing about what I call the Frightful Five, some have said my very premise is off base. I have argued that the companies’ size and influence pose a danger. But another argument suggests the opposite — that it’s better to be ruled by a handful of responsive companies capable of bowing to political and legal pressure. In other words, wouldn’t you rather deal with five horse-size Zucks than 100 duck-size technoforces?

The insatiable appetite of digital technology to alter everything in its path is among the most powerful forces shaping the world today. Given all the ways that tech can go wrong — as we are seeing in the Russia influence scandal — isn’t it better that we can blame, and demand fixes from, a handful of American executives when things do go haywire?

That’s not ridiculous. Over the last few weeks, several scholars said there are good reasons to be sanguine about our new tech overlords. Below, I compiled their best arguments about the bright side of the Five.

The Five Can Be Governed
Tech is inherently messy. The greatest human inventions tend to change society in ways that are more profound than anyone ever guesses, including the people who created them. This has clearly been true for the technologies we use today, and will be even more true for the stuff we will get tomorrow. The internet, mobile phones, social networks and artificial intelligence will make a mess of the status quo — and it will be our job, as a society, to decide how to mitigate their downsides.

One benefit of having five giant companies in charge of today’s tech infrastructure is that they provide a convenient focus for addressing those problems. Consider Russian propaganda. People have worried about the internet’s capacity to foster echo chambers and conspiracy theories almost since it began; in fact, in several cases over the last two decades — from 9/11 to the Swift Boat Veterans for Truth to birtherism — the internet did play a key role in the propagation of misinformation. But because those rumors and half-truths spread in a digital media landscape that was not owned and operated by giant companies — one in which information was passed along through a Wild West of email, discussion boards and blogs — it was never conceivable to limit that era’s equivalent of fake news.

Today, it suddenly is. Because Facebook, Google and Twitter play such a central role in modern communication, they can be hauled before Congress and either regulated or shamed into addressing the problems unleashed by the technology they control. This does not mean they will succeed in fixing every problem their tech creates — and in some cases their fixes may well raise other problems, like questions about their power over freedom of expression. But at least they can try to address the wide variety of externalities posed by tech, which may have been impossible for an internet more fragmented by smaller firms.

“This is new stuff everybody is dealing with — it’s not easy,” said Rob Atkinson, president of the Information Technology and Innovation Foundation, a think tank, and co-author of “Big Is Beautiful,” a coming book that extols the social and economic virtues of big companies. (The foundation is funded, in part, by donations from tech companies.) “

So when you discover a problem, scale makes that easier. You’ve got one or two big firms, and they have a lot of public pressure to be a responsible actor.” The Five Hate One Another’s Guts Over the last few weeks, many people at large tech companies have repeatedly responded to my questions about the dangers posed by big tech with a funny argument: Yes, they would say, the other tech giants really are worrisome — so why was I including their company in that group? It was an odd line.

As an outsider to these companies, I tend to worry about the collective power of the Five, especially the way they have managed to control the fortunes of innovative start-ups. But none of the Five see themselves as part of a group — each of them worries about the threat posed by start-ups and by the other four giants, which means that none feels it has the luxury to slow down in creating the best new stuff. This dynamic — where each company competes mightily against the others — suggests some reason for optimism, said Michael Lind, who wrote “Big Is Beautiful” with Mr. Atkinson.

“As long as their innovation rents are recycled into research and development that leads to new products, then what’s to complain about?” You can see this in their product road maps. None of the Five has slowed down investing intended to further expand its area of control — for instance, Google keeps investing in search, Facebook is still spending heavily to create new social-networking features, and Amazon remains relentless in creating new ways to let people shop.

At the same time, they are all locked in intense battles for new markets and technologies. And not only do they keep creating new tech, but they are coming at it in diverse ways — with different business models, different philosophies and different sets of ethics. “So, why pre-emptively say that maybe we’ll be harmed in the future — that in 2030 they’ll jack up their prices or something?” Mr. Lind asked. “Well, deal with that as it comes.”

The Five Are American Grown
The Five achieved their dominance because they operate in areas that provide massive returns to scale. Thanks to economic dynamics like network effects — where a product, like Facebook, gets more useful as more people use it — it was perhaps inevitable that we would see the rise of a handful of large companies take control of much of the modern tech business. But it wasn’t inevitable that these companies would be based in and controlled from the United States. And it’s not obvious that will remain the case — the top tech companies of tomorrow might easily be Chinese, or Indian or Russian or European. But for now, that means we are dealing with companies that feel constrained by American laws and values. Yes, this is jingoistic; the idea of a handful of American tech giants controlling much of society has helped push regulators internationally to try to limit their power.

But we would almost certainly do the same if a bunch of foreign companies attempted to take over our economy. At least it’s our own giants that we have to fear. I don’t mean this list to get the Five off the hook. How we deal with their efforts to capture more power over the economy and our society is perhaps the next great question facing America. But this is a complex problem precisely because there are both advantages and disadvantages to their size. As I said, tech is messy.



2017年10月21日 星期六

黑田政策動向 AI看臉就知道

黑田政策動向 AI看臉就知道
2017-10-22
經濟日報 編譯劉忠勇/綜合外電


未來要從主要央行找尋政策動向的蛛絲馬跡,或許不必再靠央行聲明,而只要看央行總裁的臉色。日本人工智慧(AI)研究人員宣稱,他 們利用電腦視覺技術來解讀日銀總裁黑田東彥臉部表情的細微變化,找到了可以預測日銀決策的辦法。 據英國Sky新聞報導,這套技術由微軟開發,是AI辨識服務產品的一環,利用演算法來分析影像,從憤怒、輕蔑、厭惡、害怕、高興、無 表情、悲傷、驚訝八種表現,來做信心程度的評比。微軟和野村證券研究團隊發現,黑田大部分時間是無表情,但也透露出其他微妙的情 感變化。 研究團隊指出,黑田會在重大政策變動之前的記者會上,會顯出細微「憤怒」和「厭惡」的表情 具體來說,日銀20161月推出負利率和9月實施「殖利率曲線管制」的一個多月前,從黑田臉上就可以測知他對現有政策力有未逮之 感。研究也發現,重大政策宣布之後,黑田顯現的「悲傷」的情緒就少了。