Goldman Sachs Hedges Its Bets with AI
Goldman may be moving slower on AI than it’s counterparts.
Sign up to uncover the latest in emerging technology.
Goldman Sachs may be adding another tool to its quickly-growing AI banking toolkit.
The financial institution is working on a system for a “regularization-based” asset-hedging tool. Basically, this system uses AI to predict the performance of a hedging portfolio relative to a certain asset it’s tracking. (A hedging portfolio includes assets to limit the risk of major drops in the price of another asset or stock.)
This tool uses an AI model trained on data from user-defined “observation period.” Then the parameters (a.k.a. the things that tell the AI model what to do) for the model are chosen from a past “validation period,” and the model is back-tested to see how it would have performed during that time frame.
Goldman said this mitigates the problem of “overfitting” seen in conventional AI-based hedging tools, or when a model is too tailored to its training data and performs poorly when given new data. This issue limits “the predictive power of such tools in generating effective risk hedging portfolios for the future performance of the target asset.”
The end result is a more flexible, more accurate hedging tool, according to the company. The end-user is given metrics on how the model performed at making its predictions, allowing them to “tailor the model to their own particular needs and preferences,” rather than being inflexible and only able to solve highly specific problems as conventional tools do. If the user makes changes, the model is retrained based on those parameters.
Users also don’t have to frequently rebalance their hedging portfolio, or adjust it to match their level of risk tolerance, which the company says can “result in significant transaction costs.”
A patent like this follows the same trend as other recent Goldman filings: The firm recently sought to patent an AI-based dashboard to help traders break down large amounts of data to visualize asset prices and transaction parameters. This patent adds to the company’s efforts to automate and optimize the jobs of its traders (potentially allowing the firm to save money on labor costs).
The financial institution itself is quite bullish on AI. Goldman said in April that generative AI could raise global GDP by 7% over the next 10 years, and forecasted earlier this month that investment in the technology could reach $200 billion by 2025. In May, the firm’s CIO Marco Argenti told the Wall Street Journal that the company was testing out generative AI, trying to “prioritize certain use cases,” and researching and investing in “a way we consider absolutely safe.”
But Goldman is seemingly slower to jump on the bandwagon than other firms: JPMorgan Chase has more than 300 use cases for AI in practice, according to CEO Jamie Dimon’s letter to shareholders. The firm’s COO Daniel Pinto also announced earlier this week that it would invest $1 billion or more a year in AI tech. And outside of traditional finance companies, plenty of fintechs and credit card companies are utilizing AI for everything from customer retention to financial planning to fraud detection.
“Right now, we have a lot of proof of concepts ongoing. Nothing is at the production stage,” Goldman’s Argenti told WSJ of its generative AI efforts.
Though Goldman is clearly showing interest in AI, the firm doesn’t seem to have the same urgency to embed the tech into its operations as other financial companies.