(Original title: Robotic Hogwash! Artificial Intelligence Will Not Take Over Wall Street)
Netease Technology News July 20, Wall Street Journal article pointed out that artificial intelligence will not take over Wall Street. Although AI investment is booming, there are still three serious problems in applying AI technology to investment.
Ten years ago, computer-driven traders stopped using algorithms after the algorithms used were doing nothing, leading to billions of dollars in losses and the closure of Goldman's flagship quantitative funds.
Ten years later, artificial intelligence and machine learning became two hot words for automated investment. However, although these new technologies are increasingly prosperous, there are still three serious problems with applying AI to investment: it runs so well that it is often incomprehensible. It only knows the recent history. What's worse is that once it becomes popular, the algorithms are confronting each other in the market, and that will be self-defeating.
Machine learning systems are really good at discovering patterns today. Unfortunately, computers are so good that they often find patterns that do not exist.
"Overfitting" problem
Michael Kollo, chief strategist at Axa IM Rosenberg Equities, talks about neural networks developed by Washington University researchers (an AI based on the brain model) that can distinguish wolves by linking wolves and snow. Pictures and pictures of dogs.
"It can easily identify a stubborn phenomenon and learn a law from it," he said. Using the market data for the past 35 years to train AI, it may form a simple rule: buying bonds. Given that the yield on the 10-year Treasury note has fallen sharply from 13.7% in July 1982 to 2.31% on Monday, it is a good strategy in hindsight – but in the next 35 years, the yield on Treasury bonds will not drop as much.
In the industry, the recognition of non-repeating patterns is called "overfitting" - noting the wolf-independent snow in photographs of wolves, or noticing accidental patterns of past stock prices that have nothing to do with future trends.
David Harding, founder of the hedge fund Winton Group, said finding ways to avoid such false laws is at the core of computer-driven investments. "Avoiding overfitting is a state of mind," he said. "It's the same thing as avoiding wishful thinking."
Anthony Ledford, chief scientist of the quantitative fund Man AHL, said that more advanced machine learning systems sometimes mean that it is less useful. "The more complex your model, the better it will explain the data you use for training, and the less able it is to interpret future data," he said. The model needs to recognize that many trends in the market are meaningless noises. It needs to try to identify important signals, even if it cannot explain some of the historical data.
Many quantitative investors are trying to avoid overfitting. They insist that the rules they use should have economic or behavioral logic. If the computer finds that every three weeks in Kansas rains, the stocks of oil companies listed in Paris rise, then betting that this will happen again in the future is tantamount to betting.
It is difficult to explain the reason for the decision
Unfortunately, it is almost impossible to explain the reasons behind the decisions made by thousands of information input systems. This is a very important issue. The U.S. defense research department is even investing in a project to try. Create AIs that can explain their own decisions.
The lack of transparency in such systems means that even advanced systems are often only pilot products and can only invest less, or they have to arrange manual monitoring of the recommendations generated by the system.
Charles Ellis is a typical example. In November last year, he joined Mediolanum Asset Management Asset Management in Dublin to develop a machine learning system. The earliest put into use system is to provide investment advice for various industries. Ellis said that with 20 years of data on 1,500 variables related to US stocks, the company used a machine learning system called the Random Forest Regression Model to try to avoid overfitting, and the early results were ideal. The system is only used for a small part of the $20 billion investment portfolio, and the final investment decision is still made by the fund manager. Another system that aims to try to predict the economic cycle is also quite promising.
The disadvantage of the random forest approach is that it is difficult to understand the reasons behind the computer making a particular decision. "It's kind of like a black box, because you don't know why that information input produces such a result," Ellis said.
Colo pointed out that when such a system causes losses, if it fails to figure out the cause of the loss, it will inevitably be closed. "Everything will eventually go wrong. Every algorithm has a bad time," he said. "The difference between those surviving algorithms and those that are not surviving is that what they do can be explained clearly."
Some investors do not care about the lack of transparency of the algorithm. Jeffrey Tarrant, of hedge fund investor Protégé燩artners, said that the issue “doesn't bother me at all.†He invested in six funds using AI technology (usually combined with unusual data Source) The investment managers of these funds come from an unusual background. He estimates that 75 funds claim to use AI technology, but he thinks there are only 25 actually using AI.
Only know recent history
Investors who use computer management funds for a long time scoff at the latest AI investment boom. Winton's Harding ridiculed. "We have been saying for 30 years that we can use computer to manage funds, but they are considered fools. Now they are coming out and you can use computers to manage funds."
He applied a machine learning technique to the moving average of futures market prices, which helped him become a billionaire. His team just saw machine learning as another statistical technique for discovering market anomalies.
Twenty years ago, Sushil Wadhwani had used machine learning techniques when he was head of the trading system at hedge fund Tudor. Today, he is running his own automation fund with machine learning - but sometimes he ignores the system's instructions. In 2008, he shut down the system's analysis of European bonds because, after years of verification, he realized that the spread between the best euro zone bond and the worst bond does not depend on economic fundamentals. As the banking system crashed, he knew it was no longer applicable.
He pointed out that "the machine will have a hard time recognizing that unless it knows that it should look at what happened during the economic crisis of the 1930s."
High-frequency trading systems may have enough cases to change the trading mechanism to operate, but they cannot use excessive funds. If many of the new data sets deployed can only be traced back to a decade or two, applying machine learning to long-term investments can be tricky. Computers that do not know history are doomed to repeat mistakes. (Lebang)
Matte Hydrogel Screen Protector
The anti-glare technology used in the Matte Protective Film can reduce glare to eliminate eye fatigue, and make it easier to watch under direct light, which is more friendly to your eyes.
In order to let you enjoy it for a long time, the Frosted Screen Protector uses durable military-grade TPU material, which has strong durability and scratch resistance. It protects the screen from unnecessary scratches.
It has good anti-fingerprint ability, sweat will not remain on the screen surface, and it is easy to clean without affecting touch sensitivity or response speed.
The adhesive layer ensures that you stick the Protective Film in a stress-free manner and maintain a strong adhesion without leaving any sticky residue.
If you want to know more about Matte Screen Protector products, please click the product details to view the parameters, models, pictures, prices and other information about Matte Screen Protector.
Whether you are a group or an individual, we will try our best to provide you with accurate and comprehensive information about the Matte Screen Protector!
Matte Screen Protector, Frosted Screen Protector, Matt Screen Protector, Matte Hydrogel Film, Matt Protective Film, Anti-Glare Screen Protector
Shenzhen Jianjiantong Technology Co., Ltd. , https://www.morhoh-sz.com