Quant Strats: Who Is Better, Machine Learning or a Human Analyst?

“It’s really this idea that, as the tooling gets better and the processes get more and more efficient, where does my strategy come in if we’re all using the same tool?” asked Andrew Gelfand, head of quant and long/short equity alpha capture at Balyasny Asset Management. “Where are the different perspectives coming from? Or, will it just be an AI-style arms race of who is most efficient?”

As the battle for AI heats up, IT leaders are working to understand the technology’s strengths and weaknesses. Here are a few ways that large language models (LLMs) can beat humans, and vice versa.

Machine Learning Dominates in Analytics, Speed and Productivity

The greatest strength of ML is its ability to digest, optimize and sort data 10 million times faster than the human brain. The model can also catch errors humans may miss.

For instance, AI-driven sentiment analysis evaluates market trends in real time using alternative data sources (credit card transactions, social media trends) to forecast stock movement more quickly than traditional human analysis.

These data applications are also allowing financial firms to layer real-time risk management on top of their statistical approaches. This reduces operational inefficiencies and allows teams to build model validations at scale.

AI can also instantly generate Python scripts to process financial data formats such as JSON or XML. This saves analysts hours of time writing code from scratch or adjusting the formats, notes Eugene Miculet, head of data strategy at WorldQuant.

However, increased speed also raises the chances of job consolidation. “AI is becoming so efficient that managers at my level may have less to do,” said Gelfand.

Human Analysts Excel at Context and Nuanced Insights

Since ML models have binary logic, humans (unsurprisingly) are better at navigating the grey areas. These can include ambiguous power dynamics, legal contracts, market trends, motives and the implications of proprietary data. 

A fundamental analyst can recall the context of a project or client request. And in investment research, they may realize that alternative data sources such as credit card transactions skew toward wealthier consumers, for example, leading to a systematic bias in market predictions, Gelfand explained.

Humans are also savvier when it comes to contract negotiations, compliance and due diligence. “They’re going out and doing the behind-the-scenes work to figure out what’s driving these companies. Analysts can also build relationships with companies in a way that models can’t,” he said.

This field work allows humans to call out “look-ahead bias,” which is a shortcoming of AI models, according to Petter Kolm, professor at New York University’s Courant Institute of Mathematical Sciences. 

READ MORE: The four most common ways to use LLMs in quantitative finance.

In Uncharted Markets, Human Judgment Outperforms AI

 “One incredible feature of the human brain is our ability to encounter a completely new scenario and use prior knowledge to rapidly and continuously adapt to the new scenario — even if it contains sensory, emotional or social stimuli we haven’t encountered before,” says Lisa Giocomo, professor of neurobiology at Stanford University, in an interview with Stanford Medicine magazine.

This is why humans can navigate uniquely stressful and unprecedent market conditions. It’s also why fundamental analysts have identified more nontraditional data sources (such as Internet of Things sensor data from a John Deere tractor show) that could provide market intelligence before competitors catch on, said Rich Brown, global head of market data at Jain Global.

 Click the banner below for financial expertise on navigating regulatory compliance.

The Verdict: Analysts Should Use AI as a Force Multiplier

In the contest between humans and ML models, is speed really everything?  And if not, what other factors are essential? Quant Strats experts listed entrepreneurship, ethics, mentorship, sociability and creativity as a few.

That trade-off suggests that organizations should consider AI a force multiplier for finance strategy, not a full replacement.

“Hedge fund and financial analysts are a relatively niche skill,” said Brown. “The amount of time and training that you would have to put into trying to replicate our jobs, I think anybody would rather spend that time on making top-line revenue.”

For Aaron Cheiffetz, senior principal consultant at CDW, it’s about balancing the human with the machine. Folding together AI and “Big Data analytics to detect macro trends” improves quant strategy, “ensuring portfolios remain resilient in changing economic environments,” he says in an interview on the Quant Strats website. But teams can’t layer AI into their infrastructure through platforms such as Snowflake, Amazon Web Services and Google Cloud Platform without a hands-on tech strategy.

“AI can be a valuable asset in systematic strategies, but it must operate within a controlled framework rather than act as an independent decision-maker,” Cheiffetz says.

UP NEXT: What does the finance bank of the future look like?