The nonprofit Hispanic Organization of Mortgage Experts last week unveiled a new AI initiative to help Spanish-speaking applicants understand and navigate America's mortgage system, which is often confusing even for native speakers.
Although already advanced, the AI will learn on the job, improving its translations by refining the way it deals with lenders' jargon with help from Spanish-speaking mortgage experts. All AI translation systems have issues handling technical and cultural nuances.
Why AI Is Urgently Needed Now
The Associated Press (AP) says the need for such a system was heightened last month as federal housing agencies implemented a March executive order to switch to English-only services. "A nationally designated language is at the core of a unified and cohesive society, and the United States is strengthened by a citizenry that can freely exchange ideas in one shared language," says the White House.
As many as 3.2 million Hispanic households in America have limited English language skills, according to the Census Bureau. Millions more whose first language is other than Spanish face similar problems.
"Over the past four decades, the number of limited English proficient (LEP) households in the US has almost tripled, says the Urban Institute. "Today, close to 5 million heads of household have a limited ability to speak, read, write, or understand English."
To start with, the new AI initiative will prioritize Spanish speakers. However, its founders hope to extend it to those with other native tongues in the future.
Challenges of AI
Many AI translation systems are already excellent for many purposes. Millions who mainly speak minority languages in countries around the world rely on applications such as DeepL and Google Translate.
No doubt, many use these translation applications daily to try to understand contracts, letters and articles written in a language that they only partly understand.
Technical Nuances
However, while these apps can often help with general translations, they frequently encounter problems with technical information.
"AI translation systems often face challenges when it comes to specialized domains like law or medicine, leading to translations that may be less precise or applicable in these specific fields," says XTM, an enterprise translation firm. "In these fields, in which content accuracy is so important, any mistake can prove to be costly, which is why having a human expert translate and/or review the content can prove to be quicker in the long run than having to go through a time-consuming post-editing task."
Mortgage lenders may deal in less complicated and critical information than doctors and lawyers, but there are still huge amounts of jargon and legal concepts involved in a loan application. And borrowers must fully understand the key ones.
Such is machine learning that, one day, AI may be able to fly solo. But, for now, most LEP households will want a competent, bilingual professional to help safely land their mortgages.
Privacy and Security
"AI translation services often require users to input their text for translation," says SimulTrans. "While reputable companies take privacy and security seriously, there is always a risk that sensitive or confidential information could be exposed, even unintentionally."
Last month, we reported on the recently banned practice of credit bureaus selling "trigger leads." When a mortgage lender ran a credit check on an applicant, the bureaus would sell the consumer's personal information to the lender's competitors. And the applicant would be bombarded with mortgage offers.
Imagine the pressure on AI developers to monetize their products. The result could be something much worse than trigger leads.
AI will almost certainly play large and exciting roles in all our lives. And it could very soon be a lifeline for LEP households wanting mortgages, especially if they have their hands held by organizations like the Hispanic Organization of Mortgage Experts. However, for now, we need to approach it with due caution.
