Panellists at a CFA Society Toronto “Innovation Series” event assert that the power of recent generative artificial intelligence (AI) tools like ChatGPT promise to revolutionize fundamental investing.

The May 17 in-person panel discussion, “From Algorithms to Action: How AI and Technology Are Driving Smarter Investment Decisions,” at CFA Society Toronto offices provided insights from expert panellists Jonathan Briggs, Director of AI-based investment firm Oracle Alpha Inc., Michael Marrale, CEO of data-based analytics solutions firm M Science, and Clare Flynn Levy, Founder and CEO of Essentia Analytics.

All three panellists agreed that the vast volumes of data and other accumulated content created by investment managers are ripe for AI-based analysis that can amplify signals in the data and generate insights that are difficult for humans to generate on their own.

This has the potential to make generative AI “massively disruptive for fundamental analysis,” Briggs said.

ChatGPT is next-level

Generative AI independently generates new original content from existing data that the AI model has been pre-trained on. In the AI subfield of natural language processing, breakthroughs in the ability to contextualize and transform inputted data, and to massively scale up the scope of training data, have revolutionized so-called large language models (LLMs) like ChatGPT (a generative pre-trained transformer (GPT) chatbot).

Briggs likened such GPT AI models to freestyle battle rap, where musicians have an idea where they want to go, but build the song one piece at a time, with every word changing the direction of the next word. A GPT model also takes a step-by-step approach to forecasting, changing the distribution of probabilities as it progresses.

This makes generative AI a good tool for fundamental research, which relies on intuition and the ability to create a comprehensive value proposition around a long-term investment, said Briggs, who previously spent two years as Managing Director of the Alpha Generation Lab and six years as the Head of Quantitative Equity Research at the Canada Pension Plan Investment Board (CPPIB).

AI can explore unasked questions and improve the process of understanding complex investment dynamics. It can remove noise from analysis and forecasts and can generate theses and counter-theses more efficiently and thoroughly than a human investment team, he said.

“It’s a thought partner that can accelerate that process of probing and understanding,” he said.

Game on

Investment managers are responding. A recent CFA Institute Global Survey showed that a combination of finance and AI (or big data) skills was the most sought-after experience in future hiring decisions, said Fred Pinto, CFA, CEO of the Society and event moderator. The survey also showed that data analysis in the core business was the main activity where AI and big data were being applied at the organizational level.

Initiatives are underway to feed the decades of accumulated investment industry data and billions of documents into LLMs like ChatGPT. US asset manager Bridgewater Associates is reportedly investing heavily in AI.¹ Bloomberg recently announced BloombergGPT, intended to allow its customers to run GPT tasks on data available via Bloomberg terminals.

“There’s a big AI bid in corporate America right now,” said M Science’s Marrale, adding that he has seen promising results running corporate earnings recaps through ChatGPT.

Privacy concerns

But just as the power of AI makes data sets relevant for specific investment outcomes more valuable, data privacy concerns arise. Regulatory bodies, such as the Consumer Financial Protection Bureau and Federal Trade Commission in the US, are focusing on the complexities of data privacy and protection raised by AI, Marrale said.

The response may be the development of large language model expert networks that would allow companies to leverage proprietary data without compromising its confidentiality, Briggs said. He envisions a future with a small number of generalized LLMs serving the investment management industry.

Despite the excitement since the November 2022 launch of ChatGPT, AI is not actually new in investment management, particularly when it comes to mundane tasks. Machine-learning algorithms have replaced manual, time-consuming tagging of vast data sets. Computer coding is changing from a task done by people to one done by AI with human oversight and direction. The development of a growing array of open source LLMs, like Databricks’ Dolly LLM, along with vendor solutions like GitHub’s Copilot, has further propelled AI’s adoption in the industry.

AI needs big data to work

However, there are limits to applying AI in investment management, said Flynn Levy, the third panellist. Her company, Essentia Analytics, uses machine learning to find statistically significant upside/downside capture patterns in large, long-only equity funds, turning the findings into English through language generation.

“We’re helping fund managers see themselves in a data-driven mirror and make better decisions going forward.”

Strictly speaking, that process isn’t AI, Flynn Levy said, since it doesn’t involve the large data sets needed for AI to be reliable. She said you can’t use AI to analyze fund manager behaviour for performance insights.

M Science has used AI to reduce time spent by staff on data tagging by almost 75 percent, but it’s a misconception that machines will ultimately replace us, said Marrale.

The session concluded with Marrale noting, “I don’t think any of us is going away,” adding that AI “is not always perfect, and we do have to have human eyes on it. We just don’t have to do everything ourselves anymore.”