How to Design Policy in the Age of AI
An expert panel discusses the yet-unknown and steps policy makers should take.
How to Design Policy in the Age of AITechnical traders have long searched for distinctive shapes in stock-price charts in order to predict returns. As artificial intelligence systems become better and better at recognizing visual patterns—for applications such as self-driving cars—can they uncover similar insights based on the shapes found in data? Chicago Booth’s Dacheng Xiu and his coauthors find that they can, producing strategies that are reliably more profitable and less risky than other well-known trading patterns.
For technical stock traders, there’s value in the ability to recognize patterns in price charts. If they’ve seen a certain shape or curve on a chart before, that could turn out to provide a reliable market predictor, and a signal whether to buy or sell. Pattern recognition is also what artificial-intelligence systems use to give self-driving cars the ability to identify pedestrians, stop signs, and other road signals, which raises the question: Could such technology be used to better predict the market?
According to Chicago Booth’s Dacheng Xiu, Yale School of Management’s Bryan Kelly, and Chicago’s CS PhD candidate Jingwen Jiang, the answer is yes. They’ve found that this same technology can be used to detect patterns in stock-price charts, and generate buy-sell signals for investors.
The researchers built a machine-learning-based program using a model called convolutional neural network, or CNN. It examines chart patterns to generate trading strategies. The researchers find that the program can produce strategies that are reliably more profitable and less risky than other current well-known trading patterns, such as momentum and reversal. They also avoid the error-prone hunches of technical investors. This approach can also use detailed data from one market to make predictions about a different market that contains little information for trading and model on its own. For example, a rich data set on US-based markets can be used to predict stock-price performance in a not-so-well-developed foreign market.
The researchers suggest that this convolutional neural network may also be able to analyze assets other than stocks, so long as they share some basic, transferable features. If the decisions made by technical traders rely on visual representations of the data that they analyze, then convolutional neural networks stand to become a powerful tool for portfolio managers of the future.
An expert panel discusses the yet-unknown and steps policy makers should take.
How to Design Policy in the Age of AIChicago Booth’s Gregory D. Bunch discusses how founders and companies could be making better use of AI to develop, test, and operationalize their strategies.
Why AI May Be Your Best StrategistIn a study, the bot cut fat from corporate disclosures to provide effective summaries.
ChatGPT Could Help Investors Make More Informed DecisionsYour Privacy
We want to demonstrate our commitment to your privacy. Please review Chicago Booth's privacy notice, which provides information explaining how and why we collect particular information when you visit our website.