The Pitfalls and Opportunities of AI

By Arianna Marino | December 17, 2024

 

By Chris Haran, Managing Director and Chief Technology Officer at Midwest Real Estate Data (MRED)

 

To predict the future, we need to look to the past. 

 

April 26, 1993. Bill Chee, serving as President of NAR, gives a speech warning of “Lions Over the Hill” coming for the real estate industry. Grappling with the idea of MLSs losing control of their data to syndication, Chee extolls the leadership of organized real estate to focus on external threats rather than in-fighting. If the MLS and its members could not figure out how to control data distribution to consumers, other entities would. 

 

Over thirty years later, the industry is consumed with lawsuits, changing regulations, and the need for consolidation… all while another technology lion is charging over the hill.

 

If syndication was about losing control of data, then artificial intelligence is losing control of knowledge. 

 

Artificial intelligence (AI) is defined as the theory and development of computer systems able to perform tasks that normally require human intelligence. AI covers a huge umbrella of techniques and technologies, has been around for decades, and is still very misunderstood.** Until recently, AI belonged only to the realm of tech giants. Then, ChatGPT appeared.

 

ChatGPT is the most popular of a segment of AI called Large Language Models (LLMs) that belongs to the family of generative AI. These models change the paradigm of information discovery from search (think Google, Bing, Yahoo, etc.) to find. Instead of looking through pages of results, the LLM looks through all the data and returns the answer you need.

 

Or, do they? 

 

LLMs work by doing something most MLSs despise; scraping data. They trawl any public data sources they can find, suck up as much information as they can to drop into their system, and then get their algorithms to work training on that data learning how to interpret and make predictions. Then, when someone asks the LLM a question, out pops the answer! 

 

However, never has the axiom ‘garbage in, garbage out’ mattered more than when it comes to this space. AI (so far) has not passed the Turing Test, which tests a machine’s ability to exhibit intelligent behavior indistinguishable from a human being. Meaning, the AI can’t (yet) think for itself. 

When AI is providing ‘answers’, it is doing its best to pull together everything it has searched to interpret what is the most likely response. This means that this information is susceptible to malicious data via astroturfing or just plain bad information from stupid people (not that anyone is ever wrong online). If that isn’t enough, all generative AI is also capable of hallucinations: just straight up providing wrong answers based on bias, bad training, or even the model making incorrect assumptions. Why yes, that does seem almost disturbingly human.

 

This mass scraping and consuming of data sets off the Bat-Signal for copyright lawyers everywhere. Data usage and ownership has been boiling since the advent of Web 2.0 and social media, and generative AI will cause that to bubble over. When even media giants like the New York Times are concerned about the effect of a large language model on their business, it means that every business needs to pay attention to their data ownership. Copyright has always been a hotspot for the MLS, and these tools make it easier than ever for that data control to slip away. 

 

Generative AI accelerates the trend of deepfakes, making it more difficult to believe what your eyes and ears are telling you. Just ask Johnny Cash’s estate with his latest ‘cover’ of Barbie Girl by Aqua. While much of it is easy to laugh at, the reality is that much of the population is not equipped to identify what is fake and real. Social engineering gathers tons of information from simple Facebook quizzes, enticing users to give answers to common security questions used online. What will be given away when your ‘boss’ calls you asking for key financial information?   

 

So, with all these pitfalls, real estate ought to put our foot down and demand to keep AI out, right? Absolutely not! 

 

AI is our generational opportunity to learn, take advantage, and evolve. Generative AI works best as an augmentation or enhancement to your current abilities and resources. Think of the MLS or brokerage and their often small or non-existent marketing departments; imagine being able to tap into a resource that can help you craft email and social media campaigns, create imagery, or help you take that rough-around-the-edges communication to members and polish it up in whatever tone you desire. Use computer vision to help your compliance proactively spot violations in images before they leave your database, and to fill in details for listing input from your members’ submitted photos and media. For technology, teams working with programs like GitHub Copilot or Amazon Q partner with your in-house developers to check code, translate into other programming languages, and increase productivity up and down the tech stack. Have an executive that is a bit shy in front of the camera or too packed a schedule for the microphone? Create a digital twin in visual or audio format, write up a script, and have constant communication to your members from your executives.

 

All of that still requires people ensuring information is correct, but it provides quicker delivery and more options for the small businesses that make up real estate. Going beyond what is available today, think of creating your own proprietary large language models and other generative AI for MLS members to help them find the answers about the local market, MLS policies, interpretation of forms, and so much more. When harnessed correctly, real estate agents become super-charged with information that makes them more valuable to consumers and the marketplace. 

 

When (not if) you enter into artificial intelligence for your organization, a few questions to ponder:

• Who owns what?
• Is what I’m seeing or hearing real?
• Why am I getting these results?
• Who is behind the curtain?

 

Syndication, artificial intelligence, whatever is next… if we in the industry focus on what makes life better for those buying, selling, and leasing, we will charge over the hill. 

 

*None of the preceding article was written using large language model tools, though the temptation was high with a looming deadline.
**Thanks to my MaiLS group at CMLX3. All that work is getting legs.