AI in Finance: Can Investors Really Use AI to Beat the Market?
For decades, hedge funds have been using algorithms to try to beat the market. And AI models like ChatGPT are only going to accelerate that.
AI is coming for finance
Goldman predicted that 35% of financial jobs are at risk from AI. And we got an early taste of it last month — when Bloomberg launched BloombergGPT, which can search and understand financial data.
But markets are much more complicated than trip planning. According to WSJ, AI has many challenges to overcome:
- Most market data is “noise,” which has little impact.
- Markets change quickly — making past data less reliable.
- Errors with investing can be more costly than other applications.
And market data is much more limited. Bloomberg’s model uses 50B parameters, while the latest GPT-4 uses over a trillion parameters to mold its answer (>6x the parameters from GPT-3).
Can AI really beat markets?
In a research paper, University of Florida Finance Prof Alejandro Lopez-Lira found evidence that ChatGPT can predict markets:
- He fed ChatGPT a news headline and asked if it was good or bad and to interpret what it meant for the company.
- Then he did this for a range of past headlines and analyzed the returns on the next trading day.
He found that ChatGPT did better in predicting the next day’s movement than at random — generating a “statistically significant predictive power on daily stock market returns.”But over the years, he thinks it’ll be challenging to benefit from this — as more investors use such tools, eliminating the edge.
Even Qraft Technologies has limited the number of clients on its AI-assisted tool — which helps spot major market-moving events. Did someone say AI-powered ETFs? Read on.