Financial modelling with large language models

AI isn’t new but the advent of the ChatGPT (Generative Pre-trained Transformer) tool in November 2022 from Open AI, with the support of Microsoft, has excited technologists and financial services professionals alike about the end uses it can be put to.

Mining the vast data lakes now available from unstructured social media resources, online news outlets and other data providers covering financial markets, to divine investable patterns and trends with ever greater granularity is one such end use of interest to quants.

Part of the LLMs & Advanced ML Summit, presenters offered quants a roadmap to hone sentiment analysis with AI tools, in order to:

Identify and target investable trends
Better manage risk and opportunity spotting
Generate alpha from media analytics
Score relevance
Construct portfolios

For this experiment, the presenters used Google’s Bard tool to compile tradable thematic baskets. Their process uses natural language processing (NLP) to continuously scan for innovative topics and convert them into tradable thematic investment baskets.

How important is natural language processing as part of a quant strategies?

Very important
Important
Somewhat important
Not important

More generally, AI has numerous applications in finance, such as:

Asset allocation

How do you identify long-term emerging economic trends, such as the rise of ESG metrics, that might move markets and business models. AI can be deployed to optimise long-term allocations accordingly and inform quant trading strategies and models.

Risk management

Procedures can protect and optimise portfolios much more easily than in the past thanks to AI, which can now be quickly built into trading models. It is especially useful if the models are tasked with specific granular end goals looking at currency fluctuations, for example, or market volatility, credit defaults and so on in greater and greater detail. “We modelled all different scenarios with our AI LLM,” a presenter said.

Stock picking

This can be aided by AI analysing financial statements, news and social media sentiment more efficiently. The aim is to consistently identify companies with strong growth potential early on. Picking an investment theme is important, but once the trend has been identified specific companies within it will also need identifying to maximise alpha.

Trading execution

What time, day or week is best to go short or long on particular stocks is a common refrain on financial markets. AI could reduce the human bias and maximise effectiveness.

AI can have its own bias as well of course. For instance, if it is only designed by rich white academics, its output will reflect that. AI might be dangerous too if a human isn’t kept in the loop as the ultimate arbiter of what can and cannot be done by the machine. But the technology is undeniably powerful and transformative – harnessing it in a good way is what will give quants a competitive edge in the year ahead.