这是艰苦的道路，所以你必须是斗士。做事情要有决心，遇到问题要有恒心，解决问题要有自信心。
邓亚萍
Thursday, March 26, 2009
Wednesday, March 18, 2009
Future of Wall Street: Rise of the Machines
http://finance.yahoo.com/news/Future-of-Wall-Street-Rise-of-portfolio-14165069.html
Max Dama was 17 years old when he decided to become a "quant." He was already interested in finance, thanks to his father, a real-estate investor, but it wasn’t until he worked on an independent project developing a video game for a college competition, (Dama’s job was to program the behavior of virtual animals acting in groups inside the game - his team won first place), that he saw the intersection between finance and artificial intelligence.
"The two just mix naturally," Dama told me recently. Now 19, Dama is a sophomore at Cal Berkeley, where he is developing a computer-based model that uses "machine learning," a branch of artificial intelligence (AI), to identify patterns in asset prices. He is currently testing his model at the Interactive Brokers Collegiate Trading Olympiad, an investing competition for students that runs for eight weeks, concluding on March 6th. After college, Dama hopes to land a job on Wall Street before starting his own hedge fund.
Dama represents a new generation of math and computer science whizzes looking to push the boundaries of quantitative finance, particularly through the use of artificial intelligence. He's preparing to enter the field at a time of intense soul-searching among many of its practitioners. The word "quant" is used broadly to refer to financial engineers who use mathematical formulas and computer models to develop financial products, assess risk, and exploit tiny price differences across markets using thousands of automatically executed rapid-fire trades. Because of their association with complex mortgage and debt-related products at the heart of the sub-prime meltdown, quants have acquired a reputation as out-of-control financial gunslingers operating at the margins of financial regulation, deploying "black-box" computer models neither their bosses nor the investing public understood.
But while there is general agreement that too much faith has been placed upon the efficacy of quantitative models, many in the field argue that the backlash has gone too far. "It’s as if a religion has failed and the witch hunt has begun," a senior risk manager told me. (Like many quants I spoke to, he insisted on anonymity.) "The mood has swung from 'quants are Gods' to 'quants are duds.' The truth lies somewhere in between."
One area that is sure to grow is the use of artificial intelligence in investment decision-making, sometimes referred to as the "second generation" of quantitative finance. In particular, machine learning shows great promise, according to Robert Levinson, a computer science professor at the University of California, Santa Cruz. "Machine learning allows the discovery of exploitable patterns that the human mind wouldn't even consider or have time to imagine," Levinson said by email.
Levinson runs the Machine Intelligence group at UCSC Baskin School of Engineering, and also serves as chief scientific officer at Alphacet, a software company that has developed a trading platform derived in part from Levinson's original research on AI in chess. "The amount of data is ever-increasing, trading intervals are growing shorter and shorter, and the competition is becoming more and more fierce," Levinson said. "It is only through the ability to make lightning fast calculations on massive amounts of data that financial institutions will be able to keep up."
But AI isn't being used just to power previously unimaginable advances in trading speed, timing and asset selection. It's also being used to mimic the human brain, without the messy emotions that color our decisions. Legend Advisory, a Florida-based division of mutual fund giant Wadell and Reed, manages over $1 billion using a computer model called the Asset Allocation Neural Network, or AANN. (The company pronounces the model "Anne" and refers to it as "she").
Although neural networks have been used on Wall Street for years, their deployment, along with other branches of AI including machine learning, is set to increase dramatically. By 2010 (next year), 50 percent of all stock trades will be generated automatically by computer algorithms, according to Aite Group, a Boston-based financial consulting firm.
AANN's goal is a familiar one: the expulsion of human emotion from investment analysis. "AANN is a human being without a heart," said Shashi Mehrotra, Legend Advisory's chief investment officer. "She's got only one job: to predict the relative behavior of certain asset classes so we can invest our clients' money better." AANN, which ranks seven asset classes, including cash, domestic and international equities, and various kinds of bonds, never experiences fear or greed, and never has a "bad day at the office," Mehrotra said.
The increasing use of complex mathematical formulas and computer programs in finance represents a double-edged sword. Over-reliance on these tools can set the table for the kind of chaos that has gripped financial markets for the last 18 months. At the same time, there is no doubt that mathematics and computers make the processing of vast amounts of data more manageable. What seems clear, however, is that humans must take ultimate responsibility for the technology, and never allow machines to decide whether to "open the pod bay doors," in the famous words of Dave Bowman, the astronaut from Stanley Kubrick's film 2001: A Space Odyssey.
In other words, the growing sophistication of quantitative finance technology makes the role of humans more, not less, critical. "Human judgment is tremendously important," said Wall Street quant veteran David Kelly, chief executive of Moment Analytics, which provides risk management systems and analytical consulting to hedge funds, by email. "All models have limitations and it is up to humans to understand them and take the controls off auto-pilot when necessary."
Max Dama was 17 years old when he decided to become a "quant." He was already interested in finance, thanks to his father, a real-estate investor, but it wasn’t until he worked on an independent project developing a video game for a college competition, (Dama’s job was to program the behavior of virtual animals acting in groups inside the game - his team won first place), that he saw the intersection between finance and artificial intelligence.
"The two just mix naturally," Dama told me recently. Now 19, Dama is a sophomore at Cal Berkeley, where he is developing a computer-based model that uses "machine learning," a branch of artificial intelligence (AI), to identify patterns in asset prices. He is currently testing his model at the Interactive Brokers Collegiate Trading Olympiad, an investing competition for students that runs for eight weeks, concluding on March 6th. After college, Dama hopes to land a job on Wall Street before starting his own hedge fund.
Dama represents a new generation of math and computer science whizzes looking to push the boundaries of quantitative finance, particularly through the use of artificial intelligence. He's preparing to enter the field at a time of intense soul-searching among many of its practitioners. The word "quant" is used broadly to refer to financial engineers who use mathematical formulas and computer models to develop financial products, assess risk, and exploit tiny price differences across markets using thousands of automatically executed rapid-fire trades. Because of their association with complex mortgage and debt-related products at the heart of the sub-prime meltdown, quants have acquired a reputation as out-of-control financial gunslingers operating at the margins of financial regulation, deploying "black-box" computer models neither their bosses nor the investing public understood.
But while there is general agreement that too much faith has been placed upon the efficacy of quantitative models, many in the field argue that the backlash has gone too far. "It’s as if a religion has failed and the witch hunt has begun," a senior risk manager told me. (Like many quants I spoke to, he insisted on anonymity.) "The mood has swung from 'quants are Gods' to 'quants are duds.' The truth lies somewhere in between."
One area that is sure to grow is the use of artificial intelligence in investment decision-making, sometimes referred to as the "second generation" of quantitative finance. In particular, machine learning shows great promise, according to Robert Levinson, a computer science professor at the University of California, Santa Cruz. "Machine learning allows the discovery of exploitable patterns that the human mind wouldn't even consider or have time to imagine," Levinson said by email.
Levinson runs the Machine Intelligence group at UCSC Baskin School of Engineering, and also serves as chief scientific officer at Alphacet, a software company that has developed a trading platform derived in part from Levinson's original research on AI in chess. "The amount of data is ever-increasing, trading intervals are growing shorter and shorter, and the competition is becoming more and more fierce," Levinson said. "It is only through the ability to make lightning fast calculations on massive amounts of data that financial institutions will be able to keep up."
But AI isn't being used just to power previously unimaginable advances in trading speed, timing and asset selection. It's also being used to mimic the human brain, without the messy emotions that color our decisions. Legend Advisory, a Florida-based division of mutual fund giant Wadell and Reed, manages over $1 billion using a computer model called the Asset Allocation Neural Network, or AANN. (The company pronounces the model "Anne" and refers to it as "she").
Although neural networks have been used on Wall Street for years, their deployment, along with other branches of AI including machine learning, is set to increase dramatically. By 2010 (next year), 50 percent of all stock trades will be generated automatically by computer algorithms, according to Aite Group, a Boston-based financial consulting firm.
AANN's goal is a familiar one: the expulsion of human emotion from investment analysis. "AANN is a human being without a heart," said Shashi Mehrotra, Legend Advisory's chief investment officer. "She's got only one job: to predict the relative behavior of certain asset classes so we can invest our clients' money better." AANN, which ranks seven asset classes, including cash, domestic and international equities, and various kinds of bonds, never experiences fear or greed, and never has a "bad day at the office," Mehrotra said.
The increasing use of complex mathematical formulas and computer programs in finance represents a double-edged sword. Over-reliance on these tools can set the table for the kind of chaos that has gripped financial markets for the last 18 months. At the same time, there is no doubt that mathematics and computers make the processing of vast amounts of data more manageable. What seems clear, however, is that humans must take ultimate responsibility for the technology, and never allow machines to decide whether to "open the pod bay doors," in the famous words of Dave Bowman, the astronaut from Stanley Kubrick's film 2001: A Space Odyssey.
In other words, the growing sophistication of quantitative finance technology makes the role of humans more, not less, critical. "Human judgment is tremendously important," said Wall Street quant veteran David Kelly, chief executive of Moment Analytics, which provides risk management systems and analytical consulting to hedge funds, by email. "All models have limitations and it is up to humans to understand them and take the controls off auto-pilot when necessary."
Subscribe to:
Posts (Atom)
