DealBook: New Breed of SAC Capital Hire Is at Center of Insider Trading Case

When Mathew Martoma walked onto the trading floor at SAC Capital Advisors six years ago, he represented a new breed of employee at the giant hedge fund.

Steven A. Cohen, SAC’s billionaire founder, had burnished his reputation as a market wizard by surrounding himself with hard-charging traders — many of them former college jocks and frat boys who thrived in the fund’s competitive, testosterone-fueled environment.

But the brainy and unassuming Mr. Martoma, armed with a Stanford business degree and an expertise in biomedicine, was part of a wave of SAC hires in a crack new research unit. They were just as driven but had more distinguished pedigrees, hailing from top investment banks and elite schools. They were drawn to the firm, in part, by the lavish annual bonuses Mr. Cohen bestowed upon his top performers, sometimes reaching into the tens of millions of dollars.

When Mr. Martoma walks into Federal District Court in Manhattan on Monday morning, he will represent something else: the latest in a growing list of former SAC employees who find themselves accused of breaking the law.

The case against Mr. Martoma, made in a criminal complaint filed by the government last week, represents a watershed moment in its multiyear investigation of insider trading at SAC. For the first time, prosecutors have linked Mr. Cohen to trading activity that the government contends was illegal.

Mr. Martoma has rebuffed efforts by federal authorities to persuade him to plead guilty and cooperate, said a person briefed on the investigation who was not authorized to discuss the case. But if he were to “flip,” Mr. Martoma could help the government with its investigation of Mr. Cohen.

The government has placed Mr. Martoma, 38, at the center of what it calls the most lucrative insider-trading scheme it has ever uncovered. Mr. Martoma is charged with corrupting a doctor who had access to secret drug data, then using that information to gain profits and avert losses totaling $276 million. Mr. Martoma closely collaborated with Mr. Cohen on the questionable trades, prosecutors contend. Mr. Cohen, 56, of Greenwich, Conn., has not been charged, and there is no allegation that he knew the information was confidential. Through a spokesman, Mr. Cohen said that he had at all times acted appropriately.

Charles A. Stillman, a lawyer for Mr. Martoma, who will appear before a federal magistrate judge Monday, said he expected his client to be exonerated.

But with Mr. Martoma’s arrest last week, the clouds over SAC, which is based in Stamford, Conn., and Mr. Cohen have darkened. The government has now implicated five former SAC employees in its sweeping investigation into insider trading; three have admitted their crimes. Three other SAC alumni have also been charged with illegal trading after they left the firm; two have pleaded guilty.

Former employees of Mr. Cohen, all of whom spoke on the condition of anonymity, said that the case against Mr. Martoma highlighted SAC’s high-stress, pressure-packed culture. They described a ruthless place where those who helped Mr. Cohen make money would earn fortunes, while laggards could be fired abruptly, even for a single wrong-way trade.

Though SAC, with about 1,000 employees, manages about $14 billion in assets and has pushed into more esoteric investment strategies, at its core the firm buys and sells stocks. Mr. Cohen and his staff are known for relentlessly digging for information about publicly traded companies to form a “mosaic,” building a complete picture of the company’s prospects that gives the firm an edge over other investors.

SAC hired Mr. Martoma to help Mr. Cohen gain that edge. The son of Indian immigrants, Mr. Martoma was born Ajai Mathew Mariamdani Thomas, but changed his name in 2003, according to legal records. Raised in Merritt Island, Fla., outside Cape Canaveral, Mr. Martoma graduated summa cum laude from Duke University in 1995, where he studied biomedicine, ethics and public policy. After college, Mr. Martoma worked in Washington at the National Human Genome Research Institute.

He spent a year and a half at Harvard Law School, then dropped out to earn a business degree at Stanford University. He blended his passion for medicine and a desire to work on Wall Street by pursuing a career as a stock analyst covering health care companies. After a stint at a smaller hedge fund, Sirios Capital Management in Boston, Mr. Martoma joined SAC.

He became part of a new unit, CR Intrinsic, which was set up as a research engine of SAC. CR Intrinsic (the CR stands for Cumulative Return) was led by Matthew Grossman, an ambitious young analyst who became Mr. Cohen’s right-hand man. Mr. Grossman had worked at Tiger Management, the hedge fund known for its rigorous research and longer-term investment horizon.

With a deep network of contacts in the pharmaceutical and biotech fields, Mr. Martoma made a mark at CR Intrinsic. The volatile health care stocks in which Mr. Martoma specialized had long been favorites of Mr. Cohen’s, offering the potential for big returns through betting on the outcome of events like clinical trials for promising drugs.

To bolster his knowledge, Mr. Martoma tapped into expert-network firms, which employ consultants who match money managers with industry specialists, including public company employees.

For an information-driven hedge fund like SAC, the temptation to exploit the expert-network relationship was immense, two former employees said.

Two of the former SAC employees who have admitted to insider trading said they used expert-network firms to obtain secret information about public companies. And of the roughly 70 insider trading cases that federal prosecutors in Manhattan have brought in the last three years, more than a dozen have involved expert networks.

Mr. Martoma’s case began in 2006, when the expert-network firm Gerson Lehrman Group connected him to Sidney Gilman, a neurology professor at the University of Michigan and a specialist in Alzheimer’s disease. Dr. Gilman, who moonlighted as a consultant for Gerson Lehrman, helped oversee clinical trials for bapineuzumab, or bapi, a new Alzheimer’s drug being jointly developed by Elan and Wyeth.

He also brazenly leaked to Mr. Martoma secret data about the trials throughout 2008, according to the government, violating his duty to the drug companies and breaching his agreement with Gerson Lehrman not to divulge confidential information to money managers. Dr. Gilman earned $108,000 from his work for SAC, the government said.

At first, Dr. Gilman’s positive updates on the Alzheimer’s drug trials emboldened SAC to make big bets on Elan and Wyeth, prosecutors said. Mr. Martoma worked closely with Mr. Cohen on the investments, highlighting the drug companies in his weekly “best ideas” list submitted to Mr. Cohen. SAC accumulated $700 million worth of Elan and Wyeth shares, making it one of the fund’s largest bets.

Prosecutors say that in July 2008 Dr. Gilman received more complete results about bapi showing problems with the drug’s efficacy. He then shared those results with Mr. Martoma, the government contends.

With the public announcement of the data just a week away, Mr. Martoma e-mailed Mr. Cohen on a Sunday, according to the complaint. Within the hour, the two were on the phone and spoke for 20 minutes, prosecutors say.

Over several days, SAC not only sold its entire positions in Elan and Wyeth, but shorted, or bet against, the drug companies’ shares, the government said. On July 29, the companies announced results of the drug trial and their shares sank. SAC avoided about $194 million in losses by selling the stocks and then made $83 million by shorting them, according to court filings. Still, SAC paid Mr. Martoma a $9.4 million bonus in 2008 that was largely attributable to his contributions on the Elan and Wyeth trades, prosecutors said. But the firm fired him in early 2010 after his stock picks flagged. The case against Mr. Martoma stemmed in part from a referral made to federal securities regulators. In 2008, the Financial Industry Regulatory Authority observed unusual short-sales in the drug stocks and noted the abnormal activity, regulators said.

Prosecutors recently reached a nonprosecution agreement with Dr. Gilman, meaning they will not charge him. The Justice Department rarely strikes nonprosecution agreements with individuals. The government’s deal with Dr. Gilman, legal experts say, could put pressure on Mr. Martoma to strike a plea deal and cooperate.

Alain Delaquérière contributed reporting.

A version of this article appeared in print on 11/26/2012, on page B1 of the NewYork edition with the headline: New Breed of SAC Capital Hire Is at Center of Insider Trading Case.

Read More..

White House Presses for Drone Rule Book

WASHINGTON — Facing the possibility that President Obama might not win a second term, his administration accelerated work in the weeks before the election to develop explicit rules for the targeted killing of terrorists by unmanned drones, so that a new president would inherit clear standards and procedures, according to two administration officials.

The matter may have lost some urgency after Nov. 6. But with more than 300 drone strikes and some 2,500 people killed by the Central Intelligence Agency and the military since Mr. Obama first took office, the administration is still pushing to make the rules formal and resolve internal uncertainty and disagreement about exactly when lethal action is justified.

Mr. Obama and his advisers are still debating whether remote-control killing should be a measure of last resort against imminent threats to the United States, or a more flexible tool, available to help allied governments attack their enemies or to prevent militants from controlling territory.

Though publicly the administration presents a united front on the use of drones, behind the scenes there is longstanding tension. The Defense Department and the C.I.A. continue to press for greater latitude to carry out strikes; Justice Department and State Department officials, and the president’s counterterrorism adviser, John O. Brennan, have argued for restraint, officials involved in the discussions say.

More broadly, the administration’s legal reasoning has not persuaded many other countries that the strikes are acceptable under international law. For years before the Sept. 11, 2001, attacks, the United States routinely condemned targeted killings of suspected terrorists by Israel, and most countries still object to such measures.

But since the first targeted killing by the United States in 2002, two administrations have taken the position that the United States is at war with Al Qaeda and its allies and can legally defend itself by striking its enemies wherever they are found.

Partly because United Nations officials know that the United States is setting a legal and ethical precedent for other countries developing armed drones, the U.N. plans to open a unit in Geneva early next year to investigate American drone strikes.

The attempt to write a formal rule book for targeted killing began last summer after news reports on the drone program, started under President George W. Bush and expanded by Mr. Obama, revealed some details of the president’s role in the shifting procedures for compiling “kill lists” and approving strikes. Though national security officials insist that the process is meticulous and lawful, the president and top aides believe it should be institutionalized, a course of action that seemed particularly urgent when it appeared that Mitt Romney might win the presidency.

“There was concern that the levers might no longer be in our hands,” said one official, speaking on condition of anonymity. With a continuing debate about the proper limits of drone strikes, Mr. Obama did not want to leave an “amorphous” program to his successor, the official said. The effort, which would have been rushed to completion by January had Mr. Romney won, will now be finished at a more leisurely pace, the official said.

Mr. Obama himself, in little-noticed remarks, has acknowledged that the legal governance of drone strikes is still a work in progress.

“One of the things we’ve got to do is put a legal architecture in place, and we need Congressional help in order to do that, to make sure that not only am I reined in but any president’s reined in terms of some of the decisions that we’re making,” Mr. Obama told Jon Stewart in an appearance on “The Daily Show” on Oct. 18.

In an interview with Mark Bowden for a new book on the killing of Osama bin Laden, “The Finish,” Mr. Obama said that “creating a legal structure, processes, with oversight checks on how we use unmanned weapons, is going to be a challenge for me and my successors for some time to come.”

The president expressed wariness of the powerful temptation drones pose to policy makers. “There’s a remoteness to it that makes it tempting to think that somehow we can, without any mess on our hands, solve vexing security problems,” he said.

Despite public remarks by Mr. Obama and his aides on the legal basis for targeted killing, the program remains officially classified. In court, fighting lawsuits filed by the American Civil Liberties Union and The New York Times seeking secret legal opinions on targeted killings, the government has refused even to acknowledge the existence of the drone program in Pakistan.

But by many accounts, there has been a significant shift in the nature of the targets. In the early years, most strikes were aimed at ranking leaders of Al Qaeda thought to be plotting to attack the United States. That is the purpose Mr. Obama has emphasized, saying in a CNN interview in September that drones were used to prevent “an operational plot against the United States” and counter “terrorist networks that target the United States.”

But for at least two years in Pakistan, partly because of the C.I.A.’s success in decimating Al Qaeda’s top ranks, most strikes have been directed at militants whose main battle is with the Pakistani authorities or who fight with the Taliban against American troops in Afghanistan.

In Yemen, some strikes apparently launched by the United States killed militants who were preparing to attack Yemeni military forces. Some of those killed were wearing suicide vests, according to Yemeni news reports.

“Unless they were about to get on a flight to New York to conduct an attack, they were not an imminent threat to the United States,” said Micah Zenko, a fellow at the Council on Foreign Relations who is a critic of the strikes. “We don’t say that we’re the counterinsurgency air force of Pakistan, Yemen and Somalia, but we are.”

Then there is the matter of strikes against people whose identities are unknown. In an online video chat in January, Mr. Obama spoke of the strikes in Pakistan as “a targeted, focused effort at people who are on a list of active terrorists.” But for several years, first in Pakistan and later in Yemen, in addition to “personality strikes” against named terrorists, the C.I.A. and the military have carried out “signature strikes” against groups of suspected, unknown militants.

Originally that term was used to suggest the specific “signature” of a known high-level terrorist, such as his vehicle parked at a meeting place. But the word evolved to mean the “signature” of militants in general — for instance, young men toting arms in an area controlled by extremist groups. Such strikes have prompted the greatest conflict inside the Obama administration, with some officials questioning whether killing unidentified fighters is legally justified or worth the local backlash.

Many people inside and outside the government have argued for far greater candor about all of the strikes, saying excessive secrecy has prevented public debate in Congress or a full explanation of their rationale. Experts say the strikes are deeply unpopular both in Pakistan and Yemen, in part because of allegations of large numbers of civilian casualties, which American officials say are exaggerated.

Gregory D. Johnsen, author of “The Last Refuge: Yemen, Al Qaeda and America’s War in Arabia,” argues that the strike strategy is backfiring in Yemen. “In Yemen, Al Qaeda is actually expanding,” Mr. Johnsen said in a recent talk at the Brookings Institution, in part because of the backlash against the strikes.

Shuja Nawaz, a Pakistan-born analyst now at the Atlantic Council in Washington, said the United States should start making public a detailed account of the results of each strike, including any collateral deaths, in part to counter propaganda from jihadist groups. “This is a grand opportunity for the Obama administration to take the drones out of the shadows and to be open about their objectives,” he said.

But the administration appears to be a long way from embracing such openness. The draft rule book for drone strikes that has been passed among agencies over the last several months is so highly classified, officials said, that it is hand-carried from office to office rather than sent by e-mail.

Read More..

Indian Prostitutes’ New Autonomy Imperils AIDS Fight

Kuni Takahashi for The New York Times

Sex workers in Mumbai’s long-established red-light district, where brothels are dwindling.

MUMBAI, India — Millions once bought sex in the narrow alleys of Kamathipura, a vast red-light district here. But prostitutes with inexpensive mobile phones are luring customers elsewhere, and that is endangering the astonishing progress India has made against AIDS.

Indeed, the recent closings of hundreds of ancient brothels, while something of an economic victory for prostitutes, may one day cost them, and many others, their lives.

“The place where sex happens turns out to be an important H.I.V. prevention point,” said Saggurti Niranjan, program associate of the Population Council. “And when we don’t know where that is, we can’t help stop the transmission.”

Cellphones, those tiny gateways to modernity, have recently allowed prostitutes to shed the shackles of brothel madams and strike out on their own. But that independence has made prostitutes far harder for government and safe-sex counselors to trace. And without the advice and free condoms those counselors provide, prostitutes and their customers are returning to dangerous ways.

Studies show that prostitutes who rely on cellphones are more susceptible to H.I.V. because they are far less likely than their brothel-based peers to require their clients to wear condoms.

In interviews, prostitutes said they had surrendered some control in the bedroom in exchange for far more control over their incomes.

“Now, I get the full cash in my hand before we start,” said Neelan, a prostitute with four children whose side business in sex work is unknown to her husband and neighbors. (Neelan is a professional name, not her real one.)

“Earlier, if the customer got scared and didn’t go all the way, the madam might not charge the full amount,” she explained. “But if they back out now, I say that I have removed all my clothes and am going to keep the money.”

India has been the world’s most surprising AIDS success story. Though infections did not appear in India until 1986, many predicted the nation would soon become the epidemic’s focal point. In 2002, the C.I.A.’s National Intelligence Council predicted that India would have as many as 25 million AIDS cases by 2010. Instead, India now has about 1.5 million.

An important reason the disease never took extensive hold in India is that most women here have fewer sexual partners than in many other developing countries. Just as important was an intensive effort underwritten by the World Bank and the Bill and Melinda Gates Foundation to target high-risk groups like prostitutes, gay men and intravenous drug users.

But the Gates Foundation is now largely ending its oversight and support for AIDS prevention in India, just as efforts directed at prostitutes are becoming much more difficult. Experts say it is too early to identify how much H.I.V. infections might rise.

“Nowadays, the mobility of sex workers is huge, and contacting them is very difficult,” said Ashok Alexander, the former director in India of the Gates Foundation. “It’s a totally different challenge, and the strategies will also have to change.”

An example of the strategies that had been working can be found in Delhi’s red-light district on Garstin Bastion Road near the old Delhi railway station, where brothels have thrived since the 16th century. A walk through dark alleys, past blind beggars and up narrow, steep and deeply worn stone staircases brings customers into brightly lighted rooms teeming with scores of women brushing each other’s hair, trying on new dresses, eating snacks, performing the latest Bollywood dances, tending small children and disappearing into tiny bedrooms with nervous men who come out moments later buttoning their trousers.

A 2009 government survey found 2,000 prostitutes at Garstin Bastion (also known as G. B.) Road who served about 8,000 men a day. The government estimated that if it could deliver as many as 320,000 free condoms each month and train dozens of prostitutes to counsel safe-sex practices to their peers, AIDS infections could be significantly reduced. Instead of broadcasting safe-sex messages across the country — an expensive and inefficient strategy commonly employed in much of the world — it encircled Garstin Bastion with a firebreak of posters with messages like “Don’t take a risk, use a condom” and “When a condom is in, risk is out.”

Surprising many international AIDS experts, these and related tactics worked. Studies showed that condom use among clients of prostitutes soared.

“To the credit of the Indian strategists, their focus on these high-risk groups paid off,” said Dr. Peter Piot, the former executive director of U.N.AIDS and now director of the London School of Hygiene and Tropical Medicine. A number of other countries, following India’s example, have achieved impressive results over the past decade as well, according to the latest United Nations report, which was released last week.

Sruthi Gottipati contributed reporting in Mumbai and New Delhi.

Read More..

G.E. Looks to Industry for the Next Digital Disruption

SAN RAMON, Calif. — When Sharoda Paul finished a postdoctoral fellowship last year at the Palo Alto Research Center, she did what most of her peers do — considered a job at a big Silicon Valley company, in her case, Google. But instead, Ms. Paul, a 31-year-old expert in social computing, went to work for General Electric.

Ms. Paul is one of more than 250 engineers recruited in the last year and a half to G.E.’s new software center here, in the East Bay of San Francisco. The company plans to increase that work force of computer scientists and software developers to 400, and to invest $1 billion in the center by 2015. The buildup is part of G.E’s big bet on what it calls the “industrial Internet,” bringing digital intelligence to the physical world of industry as never before.

The concept of Internet-connected machines that collect data and communicate, often called the “Internet of Things,” has been around for years. Information technology companies, too, are pursuing this emerging field. I.B.M. has its “Smarter Planet” projects, while Cisco champions the “Internet of Everything.”

But G.E.’s effort, analysts say, shows that Internet-era technology is ready to sweep through the industrial economy much as the consumer Internet has transformed media, communications and advertising over the last decade.

In recent months, Ms. Paul has donned a hard hat and safety boots to study power plants. She has ridden on a rail locomotive and toured hospital wards. “Here, you get to work with things that touch people in so many ways,” she said. “That was a big draw.”

G.E. is the nation’s largest industrial company, a producer of aircraft engines, power plant turbines, rail locomotives and medical imaging equipment. It makes the heavy-duty machinery that transports people, heats homes and powers factories, and lets doctors diagnose life-threatening diseases.

G.E. resides in a different world from the consumer Internet. But the major technologies that animate Google and Facebook are also vital ingredients in the industrial Internet — tools from artificial intelligence, like machine-learning software, and vast streams of new data. In industry, the data flood comes mainly from smaller, more powerful and cheaper sensors on the equipment.

Smarter machines, for example, can alert their human handlers when they will need maintenance, before a breakdown. It is the equivalent of preventive and personalized care for equipment, with less downtime and more output.

“These technologies are really there now, in a way that is practical and economic,” said Mark M. Little, G.E.’s senior vice president for global research.

G.E.’s embrace of the industrial Internet is a long-term strategy. But if its optimism proves justified, the impact could be felt across the economy.

The outlook for technology-led economic growth is a subject of considerable debate. In a recent research paper, Robert J. Gordon, a prominent economist at Northwestern University, argues that the gains from computing and the Internet have petered out in the last eight years.

Since 2000, Mr. Gordon asserts, invention has focused mainly on consumer and communications technologies, including smartphones and tablet computers. Such devices, he writes, are “smaller, smarter and more capable, but do not fundamentally change labor productivity or the standard of living” in the way that electric lighting or the automobile did.

But others say such pessimism misses the next wave of technology. “The reason I think Bob Gordon is wrong is precisely because of the kind of thing G.E. is doing,” said Andrew McAfee, principal research scientist at M.I.T.’s Center for Digital Business.

Today, G.E. is putting sensors on everything, be it a gas turbine or a hospital bed. The mission of the engineers in San Ramon is to design the software for gathering data, and the clever algorithms for sifting through it for cost savings and productivity gains. Across the industries it covers, G.E. estimates such efficiency opportunities at as much as $150 billion.

Some industrial Internet projects are already under way. First Wind, an owner and operator of 16 wind farms in America, is a G.E. customer for wind turbines. It has been experimenting with upgrades that add more sensors, controls and optimization software.

The new sensors measure temperature, wind speeds, location and pitch of the blades. They collect three to five times as much data as the sensors on turbines of a few years ago, said Paul Gaynor, chief executive of First Wind. The data is collected and analyzed by G.E. software, and the operation of each turbine can be tweaked for efficiency. For example, in very high winds, turbines across an entire farm are routinely shut down to prevent damage from rotating too fast. But more refined measurement of wind speeds might mean only a portion of the turbines need to be shut down. In wintry conditions, turbines can detect when they are icing up, and speed up or change pitch to knock off the ice.

Read More..

Scientists See Advances in Deep Learning, a Part of Artificial Intelligence

Hao Zhang/The New York Times

A voice recognition program translated a speech given by Richard F. Rashid, Microsoft’s top scientist, into Mandarin Chinese.

Using an artificial intelligence technique inspired by theories about how the brain recognizes patterns, technology companies are reporting startling gains in fields as diverse as computer vision, speech recognition and the identification of promising new molecules for designing drugs.

The advances have led to widespread enthusiasm among researchers who design software to perform human activities like seeing, listening and thinking. They offer the promise of machines that converse with humans and perform tasks like driving cars and working in factories, raising the specter of automated robots that could replace human workers.

The technology, called deep learning, has already been put to use in services like Apple’s Siri virtual personal assistant, which is based on Nuance Communications’ speech recognition service, and in Google’s Street View, which uses machine vision to identify specific addresses.

But what is new in recent months is the growing speed and accuracy of deep-learning programs, often called artificial neural networks or just “neural nets” for their resemblance to the neural connections in the brain.

“There has been a number of stunning new results with deep-learning methods,” said Yann LeCun, a computer scientist at New York University who did pioneering research in handwriting recognition at Bell Laboratories. “The kind of jump we are seeing in the accuracy of these systems is very rare indeed.”

Artificial intelligence researchers are acutely aware of the dangers of being overly optimistic. Their field has long been plagued by outbursts of misplaced enthusiasm followed by equally striking declines.

In the 1960s, some computer scientists believed that a workable artificial intelligence system was just 10 years away. In the 1980s, a wave of commercial start-ups collapsed, leading to what some people called the “A.I. winter.”

But recent achievements have impressed a wide spectrum of computer experts. In October, for example, a team of graduate students studying with the University of Toronto computer scientist Geoffrey E. Hinton won the top prize in a contest sponsored by Merck to design software to help find molecules that might lead to new drugs.

From a data set describing the chemical structure of 15 different molecules, they used deep-learning software to determine which molecule was most likely to be an effective drug agent.

The achievement was particularly impressive because the team decided to enter the contest at the last minute and designed its software with no specific knowledge about how the molecules bind to their targets. The students were also working with a relatively small set of data; neural nets typically perform well only with very large ones.

“This is a really breathtaking result because it is the first time that deep learning won, and more significantly it won on a data set that it wouldn’t have been expected to win at,” said Anthony Goldbloom, chief executive and founder of Kaggle, a company that organizes data science competitions, including the Merck contest.

Advances in pattern recognition hold implications not just for drug development but for an array of applications, including marketing and law enforcement. With greater accuracy, for example, marketers can comb large databases of consumer behavior to get more precise information on buying habits. And improvements in facial recognition are likely to make surveillance technology cheaper and more commonplace.

Artificial neural networks, an idea going back to the 1950s, seek to mimic the way the brain absorbs information and learns from it. In recent decades, Dr. Hinton, 64 (a great-great-grandson of the 19th-century mathematician George Boole, whose work in logic is the foundation for modern digital computers), has pioneered powerful new techniques for helping the artificial networks recognize patterns.

Modern artificial neural networks are composed of an array of software components, divided into inputs, hidden layers and outputs. The arrays can be “trained” by repeated exposures to recognize patterns like images or sounds.

These techniques, aided by the growing speed and power of modern computers, have led to rapid improvements in speech recognition, drug discovery and computer vision.

Deep-learning systems have recently outperformed humans in certain limited recognition tests.

Last year, for example, a program created by scientists at the Swiss A. I. Lab at the University of Lugano won a pattern recognition contest by outperforming both competing software systems and a human expert in identifying images in a database of German traffic signs.

The winning program accurately identified 99.46 percent of the images in a set of 50,000; the top score in a group of 32 human participants was 99.22 percent, and the average for the humans was 98.84 percent.

Read More..

Scientists See Advances in Deep Learning, a Part of Artificial Intelligence

Hao Zhang/The New York Times

A voice recognition program translated a speech given by Richard F. Rashid, Microsoft’s top scientist, into Mandarin Chinese.

Using an artificial intelligence technique inspired by theories about how the brain recognizes patterns, technology companies are reporting startling gains in fields as diverse as computer vision, speech recognition and the identification of promising new molecules for designing drugs.

The advances have led to widespread enthusiasm among researchers who design software to perform human activities like seeing, listening and thinking. They offer the promise of machines that converse with humans and perform tasks like driving cars and working in factories, raising the specter of automated robots that could replace human workers.

The technology, called deep learning, has already been put to use in services like Apple’s Siri virtual personal assistant, which is based on Nuance Communications’ speech recognition service, and in Google’s Street View, which uses machine vision to identify specific addresses.

But what is new in recent months is the growing speed and accuracy of deep-learning programs, often called artificial neural networks or just “neural nets” for their resemblance to the neural connections in the brain.

“There has been a number of stunning new results with deep-learning methods,” said Yann LeCun, a computer scientist at New York University who did pioneering research in handwriting recognition at Bell Laboratories. “The kind of jump we are seeing in the accuracy of these systems is very rare indeed.”

Artificial intelligence researchers are acutely aware of the dangers of being overly optimistic. Their field has long been plagued by outbursts of misplaced enthusiasm followed by equally striking declines.

In the 1960s, some computer scientists believed that a workable artificial intelligence system was just 10 years away. In the 1980s, a wave of commercial start-ups collapsed, leading to what some people called the “A.I. winter.”

But recent achievements have impressed a wide spectrum of computer experts. In October, for example, a team of graduate students studying with the University of Toronto computer scientist Geoffrey E. Hinton won the top prize in a contest sponsored by Merck to design software to help find molecules that might lead to new drugs.

From a data set describing the chemical structure of 15 different molecules, they used deep-learning software to determine which molecule was most likely to be an effective drug agent.

The achievement was particularly impressive because the team decided to enter the contest at the last minute and designed its software with no specific knowledge about how the molecules bind to their targets. The students were also working with a relatively small set of data; neural nets typically perform well only with very large ones.

“This is a really breathtaking result because it is the first time that deep learning won, and more significantly it won on a data set that it wouldn’t have been expected to win at,” said Anthony Goldbloom, chief executive and founder of Kaggle, a company that organizes data science competitions, including the Merck contest.

Advances in pattern recognition hold implications not just for drug development but for an array of applications, including marketing and law enforcement. With greater accuracy, for example, marketers can comb large databases of consumer behavior to get more precise information on buying habits. And improvements in facial recognition are likely to make surveillance technology cheaper and more commonplace.

Artificial neural networks, an idea going back to the 1950s, seek to mimic the way the brain absorbs information and learns from it. In recent decades, Dr. Hinton, 64 (a great-great-grandson of the 19th-century mathematician George Boole, whose work in logic is the foundation for modern digital computers), has pioneered powerful new techniques for helping the artificial networks recognize patterns.

Modern artificial neural networks are composed of an array of software components, divided into inputs, hidden layers and outputs. The arrays can be “trained” by repeated exposures to recognize patterns like images or sounds.

These techniques, aided by the growing speed and power of modern computers, have led to rapid improvements in speech recognition, drug discovery and computer vision.

Deep-learning systems have recently outperformed humans in certain limited recognition tests.

Last year, for example, a program created by scientists at the Swiss A. I. Lab at the University of Lugano won a pattern recognition contest by outperforming both competing software systems and a human expert in identifying images in a database of German traffic signs.

The winning program accurately identified 99.46 percent of the images in a set of 50,000; the top score in a group of 32 human participants was 99.22 percent, and the average for the humans was 98.84 percent.

Read More..

Military Analysis: For Israel, Gaza Conflict Is Test for an Iran Confrontation

Menahem Kahana/Agence France-Presse — Getty Images

An Israeli missile is launched from a battery. Officials said their antimissile system shot down 88 percent of all assigned targets.

WASHINGTON — The conflict that ended, for now, in a cease-fire between Hamas and Israel seemed like the latest episode in a periodic showdown. But there was a second, strategic agenda unfolding, according to American and Israeli officials: The exchange was something of a practice run for any future armed confrontation with Iran, featuring improved rockets that can reach Jerusalem and new antimissile systems to counter them.

It is Iran, of course, that most preoccupies Prime Minister Benjamin Netanyahu and President Obama. While disagreeing on tactics, both have made it clear that time is short, probably measured in months, to resolve the standoff over Iran’s nuclear program.

And one key to their war-gaming has been cutting off Iran’s ability to slip next-generation missiles into the Gaza Strip or Lebanon, where they could be launched by Iran’s surrogates, Hamas, Hezbollah and Islamic Jihad, during any crisis over sanctions or an Israeli strike on Iran’s nuclear facilities.

Michael B. Oren, the Israeli ambassador to the United States and a military historian, likened the insertion of Iranian missiles into Gaza to the Cuban missile crisis.

“In the Cuban missile crisis, the U.S. was not confronting Cuba, but rather the Soviet Union,” Mr. Oren said Wednesday, as the cease-fire was declared. “In Operation Pillar of Defense,” the name the Israel Defense Force gave the Gaza operation, “Israel was not confronting Gaza, but Iran.”

It is an imprecise analogy. What the Soviet Union was slipping into Cuba 50 years ago was a nuclear arsenal. In Gaza, the rockets and parts that came from Iran were conventional, and, as the Israelis learned, still have significant accuracy problems. But from one point of view, Israel was using the Gaza battle to learn the capabilities of Hamas and Islamic Jihad — the group that has the closest ties to Iran — as well as to disrupt those links.

Indeed, the first strike in the eight-day conflict between Hamas and Israel arguably took place nearly a month before the fighting began — in Khartoum, the capital of Sudan, as another mysterious explosion in the shadow war with Iran.

A factory said to be producing light arms blew up in spectacular fashion on Oct. 22, and within two days the Sudanese charged that it had been hit by four Israeli warplanes that easily penetrated the country’s airspace. Israelis will not talk about it. But Israeli and American officials maintain that Sudan has long been a prime transit point for smuggling Iranian Fajr rockets, the kind that Hamas launched against Tel Aviv and Jerusalem over recent days.

The missile defense campaign that ensued over Israeli territory is being described as the most intense yet in real combat anywhere — and as having the potential to change warfare in the same way that novel applications of air power in the Spanish Civil War shaped combat in the skies ever since.

Of course, a conflict with Iran, if a last-ditch effort to restart negotiations fails, would look different than what has just occurred. Just weeks before the outbreak in Gaza, the United States and European and Persian Gulf Arab allies were practicing at sea, working on clearing mines that might be dropped in shipping lanes in the Strait of Hormuz.

But in the Israeli and American contingency planning, Israel would face three tiers of threat in a conflict with Iran: the short-range missiles that have been lobbed in this campaign, medium-range rockets fielded by Hezbollah in Lebanon and long-range missiles from Iran.

The last of those three could include the Shahab-3, the missile Israeli and American intelligence believe could someday be fitted with a nuclear weapon if Iran ever succeeded in developing one and — the harder task — shrinking it to fit a warhead.

A United States Army air defense officer said that the American and Israeli militaries were “absolutely learning a lot” from this campaign that may contribute to a more effective “integration of all those tiered systems into a layered approach.”

The goal, and the challenge, is to link short-, medium- and long-range missile defense radar systems and interceptors against the different types of threats that may emerge in the next conflict.

Even so, a historic battle of missile versus missile defense has played out in the skies over Israel, with Israeli officials saying their Iron Dome system shot down 350 incoming rockets — 88 percent of all targets assigned to the missile defense interceptors. Israeli officials declined to specify the number of interceptors on hand to reload their missile-defense batteries.

Before the conflict began, Hamas was estimated to have amassed an arsenal of 10,000 to 12,000 rockets. Israeli officials say their pre-emptive strikes on Hamas rocket depots severely reduced the arsenal of missiles, both those provided by Iran and some built in Gaza on a Syrian design.

Read More..

The Shrewd Shopper Carries a Smartphone

Tim Gruber for The New York Times

From left, Tara Niebeling, Sarah Schmidt, Bridget Jewell and Erin Vande Steeg are members of the social media team at the Mall of America in Bloomington, Minn.

Retailers are trying to lure shoppers away from the Internet, where they have increasingly been shopping to avoid Black Friday madness, and back to the stores. The bait is technological tools that will make shopping on the busiest day of the year a little more sane — and give shoppers an edge over their competition.

Those with smartphones in hand will get better planning tools, prices and parking spots. Walmart has a map that shows shoppers exactly where the top Black Friday specials can be found. A Mall of America Twitter feed gives advice on traffic and gifts, and the Macy’s app sends special deals for every five minutes a shopper stays in a store.

“The crazy mad rush to camp out and the crazy mad rush to hit the doorbusters have really made people think, ‘I’m just going to stay home on Black Friday,’ ” said Carey Rossi, editor in chief of ConsumerSearch.com, a review site. “This is going to invite some people back and say, ‘You know what? It doesn’t have to be that crazy.’ ”

Part of the retailers’ strategy is to slap back at online stores like Amazon.com, which last year used apps to pick off shoppers as they browsed in physical stores. But the stores are also recognizing that shopping on the Friday after Thanksgiving need not require an overnight wait in line, a helmet and elbow pads. A smartphone gives shoppers enough of an edge.

“This takes away that frantic Black Friday anxiety,” said Lawrence Fong, co-founder of BuyVia, an app that sends people price alerts and promotions. “While there’s a sport to it, life’s a little too short.”

Denise Fouts, 45, who works repairing fire and water damage in Chandler, Ariz., was already using apps, including Shopkick, Target’s app and one called Black Friday, before Thanksgiving to prepare for Black Friday. “There still are going to be the crowds, but at least I already know ahead of time what I’m going specifically for,” Ms. Fouts said.

Last week, Macy’s released an update to its app with about 300 Black Friday specials and their location. In the Herald Square store, for instance, the $49.99 cashmere sweater specials will be in the Broadway side of the fifth-floor women’s department.

“With the speed that people are shopping with on Black Friday, they need to be really efficient about how they’re spending their time,” said Jennifer Kasper, group vice president for digital media at Macy’s.

When shoppers keep the app open, Macy’s will start sending special deals to the phone every five minutes. The deals are not advertised elsewhere.

Walmart has had an app for several years, but recently introduced an in-store mode, which shows things like the current circular or food tastings when a shopper is near a certain location. Twelve percent of Walmart’s mobile revenue now comes from when a person is inside a store.

For Black Friday, the app will have a map of each store, with the precise location of the top sale items — so planners can determine the best way to run. “The blitz items are not where you think they would be, because for traffic reasons, maybe the hot game console is in the lawn and garden center,” said Gibu Thomas, senior vice president for mobile and digital for Walmart Global eCommerce.

Target is also testing a way-finding feature on its app at stores that include some in Seattle, Chicago and Los Angeles. If a shopper types in an item, the app will give its location.

Other app makers are betting that shoppers want apps that pull in information from many stores.

RedLaser, an eBay app, lets shoppers use their phones to compare prices and recently started using location data to give shoppers personalized promotions when they walk into stores, including items not on store shelves at Best Buy, for instance. RetailMeNot, which offers e-commerce coupons, now has offline coupons that will pop up on users’ cellphones when they step near 500 malls on Black Friday.

“Consumers are not going to download 40 different apps for 40 different stores,” said Cyriac Roeding, co-founder of Shopkick, a location-based app that gives shoppers points, redeemable for perks, when they walk into stores or scan certain items.

For Black Friday, Shopkick is publishing what it calls a little black book with the top doorbusters. Shoppers will earn extra points and rewards for shopping on Black Friday.

Read More..

The New Old Age Blog: Home Health Care Help by the Hour

Here is a bright idea that ought to spread: You call your home care agency and say you will need your mother’s aide for the normal two hours on Monday and Wednesday, but for just half an hour (to drive her to a doctor’s office) on Tuesday, then 90 minutes on Thursday. And the agency says, “Sure.”

It sounds logical to hire someone to help — with bathing, dressing, errands, meal preparation, medication reminders – for only as many hours as an older adult needs assistance. But it is actually unusual for companies to offer such flexibility.

The majority of agencies require a four-hour minimum. Having to spend $80  — the national average cost for home care is $21 an hour — if you only need $40 worth of help is a big barrier for families trying to keep their elderly relatives living at home longer. A few agencies allow you to hire for fewer than four hours, but at higher rates.

But Mission Healthcare in San Diego, Calif., a three-year-old agency that began with Medicare-certified skilled home nursing and hospice care, expanded to general home care this summer and decided that clients should be able to specify how much help they want – in 15-minute increments — and will pay for.

“We’ll come for as long as they need us to,” said Mark Kimsey, one of Mission’s four directors. “In one hour, a well-trained caregiver can get the client bathed and dressed, prepare three meals and have them organized for the day.” (I have to think that is a speedy caregiver with a not-too-frail client, but still … )

Can Mission, which charges $19 to $20 an hour, actually make money this way? Though overall the agency serves 1,100 clients, its fledgling home care business is still small: 30 aides caring for just 60 clients. The aides can get benefits if they work enough hours, a bonus for them and for consumers (better employees, lower turnover), but an additional cost for the company.

The directors say they are profitable already, and that the approach will succeed because more people will like the flexibility and potential savings, and sign up. It is also true, let’s acknowledge, that a person who can get by with an hour or two a day for now may well need more help eventually, a boost for Mission’s bottom line.

Competitors are no doubt watching to see if this works. I’m curious, too, to see if the policy catches on. “Consumers can change the marketplace if they want to,” Mr. Kimsey said. It would be nice to think that is true.

Would you use this kind of hourly service, if it were available?

Paula Span is the author of “When the Time Comes: Families With Aging Parents Share Their Struggles and Solutions.”

Read More..

The New Old Age Blog: Home Health Care Help by the Hour

Here is a bright idea that ought to spread: You call your home care agency and say you will need your mother’s aide for the normal two hours on Monday and Wednesday, but for just half an hour (to drive her to a doctor’s office) on Tuesday, then 90 minutes on Thursday. And the agency says, “Sure.”

It sounds logical to hire someone to help — with bathing, dressing, errands, meal preparation, medication reminders – for only as many hours as an older adult needs assistance. But it is actually unusual for companies to offer such flexibility.

The majority of agencies require a four-hour minimum. Having to spend $80  — the national average cost for home care is $21 an hour — if you only need $40 worth of help is a big barrier for families trying to keep their elderly relatives living at home longer. A few agencies allow you to hire for fewer than four hours, but at higher rates.

But Mission Healthcare in San Diego, Calif., a three-year-old agency that began with Medicare-certified skilled home nursing and hospice care, expanded to general home care this summer and decided that clients should be able to specify how much help they want – in 15-minute increments — and will pay for.

“We’ll come for as long as they need us to,” said Mark Kimsey, one of Mission’s four directors. “In one hour, a well-trained caregiver can get the client bathed and dressed, prepare three meals and have them organized for the day.” (I have to think that is a speedy caregiver with a not-too-frail client, but still … )

Can Mission, which charges $19 to $20 an hour, actually make money this way? Though overall the agency serves 1,100 clients, its fledgling home care business is still small: 30 aides caring for just 60 clients. The aides can get benefits if they work enough hours, a bonus for them and for consumers (better employees, lower turnover), but an additional cost for the company.

The directors say they are profitable already, and that the approach will succeed because more people will like the flexibility and potential savings, and sign up. It is also true, let’s acknowledge, that a person who can get by with an hour or two a day for now may well need more help eventually, a boost for Mission’s bottom line.

Competitors are no doubt watching to see if this works. I’m curious, too, to see if the policy catches on. “Consumers can change the marketplace if they want to,” Mr. Kimsey said. It would be nice to think that is true.

Would you use this kind of hourly service, if it were available?

Paula Span is the author of “When the Time Comes: Families With Aging Parents Share Their Struggles and Solutions.”

Read More..