It was an experiment straight out of “2001: A Space Odyssey.”
In an effort to gauge the accuracy of AI technology, researchers presented real estate journalist John Rebchook with several recommended listings. The recommendations, which were based on a home Rebchook liked, came in two sets – one from Denver-area brokers, another from “Find More Genius,” a computer algorithm that Denver-area broker Creed Smith had created. In the end, the humans did not stand a chance. Over three separate days of testing, Rebchook’s favorite recommendation came from the algorithm each day, as did two of his three overall favorites.
“I couldn’t really differentiate between what the bots picked and the human picked,” Rebchook said. “I guess if computers can beat us at chess and Go, they can best us at picking houses, too.”
Artificial intelligence is hardly a novel concept. Siri informs iPhone users about weather forecasts and traffic reports. IBM’s Watson computer system obliterated Ken Jennings in Jeopardy! And computer chess engines are so sophisticated that every year, the International Computer Games Association organizes the World Computer Chess Championship, where all the players are microprocessors, supercomputers and other forms of hardware.
So in this world of computers and “deep learning” technologies, the question is not “if” AI will transform real estate, but “when.”
AI in today’s real estate world
Although real estate has yet to experience a Deep Blue-level revolution, there are still many fascinating forms of artificial intelligence currently at work, including:
- Facebook Messenger – No mere messaging app, Facebook Messenger now allows its users to search for apartments. Via a partnership with Trulia, users send a message to a Trulia bot, indicating where in the U.S. they would like to rent; the bot, in kind, responds with various listings, and can provide summaries and images of the listings.
- CRMs – Lead management systems are ubiquitous, and many are incorporating AI features. Boomtown, for instance, matches buyers to properties and sends out automated drip email communications, while Signpost, which recently joined the Zillow Tech Connect program, uses automated marketing to drive five-star reviews and social media posts.
- Address Report – This fascinating program allows consumers to perform “background checks” on properties, and learn of their history, market health and even amenities (right down the the speed of the elevators). Even more, it delivers dynamic content to subscribers based on a property’s location, providing notifications when certain data points change.
- CityBldr – A new product from Seattle-based startup Everyhome, CityBldr pairs real estate developers with underused properties and potential land assemblages, saving money on real estate commissions in the process. Using 180 data points, the product ranks properties on their best use and estimates a return on investment. It also tells users when competitors are trying to get a permit for a property, and allows users to send owners a purchase offer.
What the AI future holds
One of the most precise AI systems in the world is also among the most widely used – the recommendations system from Netflix. According to data from the company, roughly 75 percent of the movies and shows that Netflix users watch are directly based on its recommendations system, which uses algorithms to study user behavior and match them with similar programming.
That kind of “deep learning,” which entails finding patterns from large quantities of data, is likely the coming revolution in the homebuying experience. Instead of the search engines of today, which pair consumers with homes based on a set number of factors – for example, three bedrooms, two bathrooms and a basement – a deep learning system would incorporate a home shopper’s personality traits and preferences, pairing them with a home in a neighborhood that perfectly complements their lifestyle.
In a recent Onboard Informatics article, Ira Monko described the potential of such technology: “Today, portals seems to display more information in their effort to provide value. That’s way off. Portals don’t need to display more information – they need to display the right information.”