MISMO/MBA #MortgageAIForum: Artificial Intelligence Holds Promise, Peril for Mortgage Lending

(Left to right: Camelia Martin from Snapdocs, Tom Clerici from Freedom Mortgage, Zac Maufe from Google Cloud, Martin Kaestner from KPMG LLP, Christos Bettios from NFM Lending)

ARLINGTON, Va.–Artificial Intelligence will fundamentally change the real estate finance business—and nearly everything else—experts said here at the #MortgageAIForum, co-hosted by MBA and MISMO.

The forum, which began with a joint MISMO/MBA networking event at Amazon’s HQ2 building in Arlington, brought together AI experts, mortgage professionals and government officials to discuss opportunities, risks and regulatory interests in this fast-developing area. Various panels addressed different use cases for AI and demonstrated AI products.

Michael Bradley, Senior Vice President of Single-Family Modeling, Econometrics, Data Science and Analytics (MEDA) at Freddie Mac, McLean, Va., said artificial intelligence–computer systems that can perform tasks that normally require human intelligence–“can help us do our work better.” He noted AI helps Freddie Mac work smarter and helps the GSE keep pace with what’s going on in the external market.

Advances in AI technology in recent decades spring from several factors, including the sheer volume of data that has become available and computing advances that allow faster processing of that data, Bradley said. “The third [advance] is machine learning and other techniques that allow us to make great insights from that those huge stores of data.”

The pace of change is “exponential,” Bradley said. He mentioned IBM’s Deep Blue chess computer, which beat former World Chess Champion Garry Kasparov in 1997. “They fed the machine information about how grandmasters played the game. And of course, the machine beat the human because it could go through all the possible moves and permutations much faster than a human can.” Then, just 20 years later in 2017, computer program AlphaZero was able to build a new deep learning neural net to play chess. But AlphaZero did not get fed data about how humans play chess, it received the objective function and the rules of chess, then the machine played against itself, Bradley noted. “And within four hours, AlphaZero created strategies that no human today has beaten,” he said.

Zac Maufe, Head of Regulated Industries at Google Cloud, San Francisco, noted financial services firms have been early adopters and users of technology for a long time. “I remember using machine learning many years ago on tasks that seem kind of quaint now,” he said. “Being able to take data and turn it into structured data to have clean data is going to be even more important now than ever before.”

Maufe said Generative AI, which is artificial intelligence capable of generating text or images, has grown more important. “I think the easiest way to think about this [Generative AI] is it’s a massive productivity improver,” he said, noting applications ranging from coding to analysis.

“The other population is resources that are really hard to hire and retain,” Maufe said. “Think about things like call centers and the amount of training they require.” He cited one firm that sought to cut in half the amount of onboarding/training time its customer call center representatives needed by allowing them to use a ‘phone a friend’ Generative AI feature that helps with more difficult questions. “So along the lines of productivity, we think that there’s significant value that can be generated by the use of this [Generative AI] technology,” he said.

Kevin Levitt, Global Industry Business Development, Financial Services at NVIDIA, Santa Clara, Calif., said there are many applications of AI in the financial services world. “Some in this industry focus on geospatial data. There are recommendation systems about what kind of loan would be best. In this industry there are hundreds of AI-based applications and soon there will be thousands.”

Levitt called last year’s launch of ChatGPT “the AI shot heard ‘round the world,” like the start of the American Revolution. “We are very much on the early part of the journey [to AI] and we will move faster as we learn from each other,” he said. “Our perspective is that there are many different entities competing for financial customers. We believe unless you are working to AI-enable the critical applications across your enterprise, you are going to fall behind those that are. The service that gets delivered will be incrementally better if you leverage AI.”

Levitt said he thinks every corporation will compete around AI in the future. “The key to building successful AI is to build it around the needs of your end users,” he said.