Like many managers, Newton Investment Management is exploring uses for applications like ChatGPT. But Newton, a subsidiary of BNY Mellon Investment Management, is skeptical about using large language models alone and is instead pairing the models with fundamental analysis to potentially produce better results.
“When I look at things like ChatGPT and the more general types of models, I’m not entirely sure they are going to be that helpful to us,” said Euan Munro, CEO of Newton. “That’s simply because when you allow your model to look at all the information, it pulls in data sets all over the place. That’s when you are at risk of getting these false positives…It’s not really capable of differentiating between correlation and causality in the way that I think humans are.”
Newton has a potential fix. “I think the future for us is going to be using artificial intelligence, but giving it [fundamental] data sets that we believe are relevant, high-quality, and accurate, rather than just letting it scramble around,” Munro said. He used the inflation rate in Japan to illustrate how fundamental analysis can improve predictions of AI models. Japan’s current inflation level is around 3 percent, higher than the central bank’s target. From a fundamental research perspective, that means the central bank may need to raise interest rates aggressively to curb inflation. But a large language model reliant on historical data would fail to predict a rise in interest rates because it’s statistically unlikely given Japan’s historical pattern of low inflation.
“A combination of humans imagining how the future might be different from the past with the actual modeling is quite powerful, as long as you know how to bring it together,” Munro said.
The rise of ChatGPT, a chatbot developed by Microsoft-backed OpenAI, has pushed quantitative and fundamental managers to explore how to use large language models to improve investment returns. AQR, for example, found in backtests that using large language models would have helped investors outperform the market from 2004 to 2019.
Newton’s exploration of AI comes two years after BNY Mellon decided to merge Newton Investment Management with Mellon’s active equity and fixed income teams. Before the merger, Mellon focused on quantitative research, while Newton emphasized fundamental analysis. Since the consolidation, the two teams have attempted to benefit from each other’s investment strategies. Their approach to the adoption of AI tools is the latest example.
