By Ray Birch
TAMPA—Artificial intelligence (AI) will provide a boost for credit unions, provided credit unions do one important thing that will be critical for their futures, says Tom Davis.
As part of an ongoing CUToday.info series on the effect increasingly sophisticated AI will have on credit unions, Davis, the CEO of Trellance, emphasized credit unions will need to understand how the new technology is making decisions that affect members’ lives.
Part I in this series appeared here.
“It's very difficult to tell—with machine learning, deep learning and things like that—how AI is making its decisions,” observed Davis. “When we start to rely on these tools to make decisions that impact our members’ lives, if we can determine how those outcomes are made, I think that is very important.”
Not Just About Code
Until recently, most of the technology and its underlying programming has not been difficult to understand, said Davis.
“It has been easy enough to walk through the code and say, well, you know why this was done for that reason,” said Davis.
But that isn’t so clear in AI-driven systems such as automated lending, as the technology is generative and actually becomes smarter as it sifts through large data sets and recognizes patterns. In those cases it is challenging to determine how the system actually came up with its outcomes, offered Davis.
“Now, we somewhat struggle with that, specifically with financial services,” he said.
Understanding the Why’s
As an example, Davis pointed to credit decisioning. If AI is charged with providing or denying credit to individuals, a lack of understanding why a decision was made creates potential cases of litigation or regulatory scrutiny in which, say, redlining might be alleged.
“Are we on firm ground to say that there was no bias in the system? And did that system allow us, as credit unions focused our mission, to meet our members’ needs?” Davis asked. “Artificial intelligence is trying to mimic how we make decisions and it will get things wrong at times, and we may not be able to understand completely as users of these systems how it got things wrong. Yet, at the same time there is that concern, there is the ability of AI to make decisions that are better than we could have made on our own—helping us to reach better outcomes faster.”
‘Don’t Really Have a Choice’
Davis said he firmly believes credit unions need AI.
“I also believe that we don't really have a choice, to say cannot use it,” explained Davis. “I think we need to accept the fact that AI is trying to mimic how we go about our day, how we make decisions, and we need to understand how we can better work with these models and be able to make sure they meet our members’ needs.”
Davis reminded that AI is already impacting how credit unions make decisions.
“We're using neural networks, which in some ways try to mimic our ability to understand things, like risk, for example. It can detect fraud faster than we can by using larger data sets and detecting patterns,” Davis noted. “However, obviously, as AI helps us better identify fraud, the bad guys are able to use these systems, too. (It will involve) working in reverse to try to anticipate how our systems will pick up fraud. There'll be this constant, back-and-forth battle of good and evil, so to speak.”
AI Needed to Deal With AI
Decisions on use of AI just keep coming at credit unions more quickly, Davis reminded.
Tom Davis
“There's a lot more use-cases and technology available to us today. And I think, as we all know, this is coming at us faster and faster,” said Davis. “As this technology gets more democratized, meaning more folks within credit unions can actually use some it…Twenty or 30 years ago, you heard about AI, things called Watson and computers that were beating chess champions. But then, I don’t know how you really applied AI towards meaningful use-cases in our lives. I think what we're seeing now with this new wave of technology that's available, you can go and get a ChatGPT personal account, for that matter, and you can use it in personal and professional lives.”
Davis acknowledged he is concerned that as credit unions begin to turn to tools such as ChatGPT proprietary CU knowledge can make its way into the AI knowledge base.
“One of my big concerns now is with what am I giving up in order to get those insights,” said Davis.
A Different Hollywood Ending
Davis pointed out credit unions will be in a constant state of dealing with new technology that can change the way that people do their jobs today, possibly even supplanting some of the ways workers do their jobs.
“I know that scares a lot of people because we all want to think we all have careers,” said Davis. “But as people are finding out in Hollywood…”
More in this series: