Summary List Placement
Financial firms are putting more resources towards using complex forms of machine learning. And that means they’ll also need to develop ways to better understand the tech and explain it both internally and to regulators.
A recent survey by data giant Refinitiv on the use of artificial intelligence and machine learning in financial services found that 75% of respondents were using some form of deep learning, a type of ML that includes a series of complex, ever-evolving calculations. Those surveyed included data scientists, quants, and executives at a variety of financial firms.
Geoffrey Horrell, head of Refinitiv Labs in London, told Business Insider that the surge of interest in deep learning means firms will need to invest resources in understanding these complex algorithms.
Deep-learning techniques can be used to handle more complex problems thanks to their ability to ingest and understand significant amounts of data. However, that additional firepower comes at a cost.
“As you see that growth in deep learning, and particularly around the text analysis part of deep learning, there’s that need to have better explainability,” Horrell said.
Wall Street’s AI leaders are already considering explainability
The increased focus on being able to explain and interpret machine-learning techniques stems from the top, Horrell said. Leaders in the field have already set up governance teams for their AI models and are filling roles focused on ensuring the tech is used appropriately and fairly, he added.
For all the benefits the use of artificial intelligence offers— streamlining manual processes and faster analysis of data — the tech still has its hangups. AI models have been found to be racially biased, often times because of the datasets they ingest.
As a result, the industry has been hesitant about where to deploy the tech, especially when it pertains to decisions directly impacting customers.
In May, Business Insider reported that Bank of America hired Diane Daley, a former Citigroup executive, to lead its enterprise governance function. Daley’s responsibilities include AI policies, standards, and oversight.
Cathy Bessant, the bank’s chief operations and technology officer, has long been outspoken about the responsible use of AI. Bank of America was a founding donor of Harvard’s Kennedy School of Government’s Council on the Responsible Use of Artificial Intelligence in 2018.
In August 2019, Business Insider reported some of the largest Wall Street banks, including Morgan Stanley and Citi, were in the process of forming a working group.
“There is a lot of interest and activity in this whole area,” Horrell said. “The people who are leading are talking about this, and I think things may follow rapidly.”
Regulators also want firms to show their work when it comes to AI
The increased use of machine learning has also caught regulators’ eyes, Horrell said. In particular, rule makers are keen to understand how the technology would be applied to making decisions that directly impact customers.
Things like credit decisions, know-your-customer processes, and anti-money laundering tools all fall into that category.
In short, anytime a choice is being made about whether a service will be given to someone, regulators want to understand how the firm reached that conclusion, Horrell said.
And while regulators have yet to put forth specific rules around how they want firms to monitor AI usage, considerations are already being made. The European Union currently preparing its first set of rules around AI.
“The idea that regulation might come, again, means that people will want to get interested in this,” he said.
To be sure, it’s not just regulators that are motivating firms. The companies themselves are eager to understand how these technologies reach different decisions.
While the use of AI for investment research or trading ideas might not catch the eye of regulators, it is important for firms to understand how those techniques will work under different market conditions.
Horrell said investors want to make sure AI-based models are robust enough to withstand various market conditions, as opposed to doing just enough data-mining to find some correlations that might not stand the test of time.
The rise of quantamental strategies — the use of quantitative and fundamental techniques — has further increased those efforts.
As traditional funds look to incorporate more quantitative modelling into their process, it’s important for them to understand how the tech works, Joshua Pantony, CEO of Boosted.ai., a fintech that helps fundamental managers use quantitative skills.
“Explainable machine learning that can tell you what it’s doing has probably been the single most important success factor we’ve seen in clients trying to take the quantamental approach,” he said.