This AI tool writes real estate descriptions without ever entering a house


It looked like any number of property descriptions you might see online in today’s US housing market. Usually they are written by humans. This one, however, was made up by artificial intelligence.
A Canadian startup called Listing AI uses cutting edge AI to quickly produce computer-generated descriptions that, like the one above, can be surprisingly compelling. All users have to do is give it a few details about the house, and the AI ​​does the rest.

The results need to be improved even further: The real Oakland, Calif. Home that fits the description above (which my family is currently selling) actually has a pressed pewter ceiling in the dining room, rather than the living room. living room, for example. The descriptions that Listing AI created for me are not as specific or well written as those put together by our (human) real estate agent. And I had to provide the website with a lot of information about the different rooms and features in the house and the exterior landscaping – a process that looked a bit like Mad Libs real estate – before the website could come up with several different descriptions. . .

But the overall consistency of the descriptions Listing AI came up with seconds after my submission provides another sign that AI is improving in a task that was traditionally seen as uniquely human – and shows how people may be able to work with it. technology, rather than fear that it could replace us.

It probably won’t do all the work of writing a house description for you, but according to Listing AI co-founder Mustafa Al-Hayali, that’s not the point. He hopes he will complete about 80-90% of the work of crafting a home description, which can be done by a real estate agent or copywriter.

“I don’t think it’s meant to replace a person when it comes to getting a job done, but it’s supposed to make their job a lot easier,” Al-Hayali told CNN Business. “It can generate ideas that you can use.”

There is a basement, but it’s not over

Al-Hayali, whose day job is a senior software developer on the Canada-based e-commerce platform Shopify, said he and his co-founder (Corey Pollock, a senior product manager at Shopify), had the idea of ​​Lister AI after the two became new owners in Toronto. As they went through this process, he said, they noticed inaccurate or even copied property descriptions from a previous period when the property was for sale. Al-Hayali bought his condo in March 2020 – just before Covid forced many people to start relying on virtual tours and other means of learning properties remotely, which made descriptions detailed all the more important.

They built the website over the past month and launched it publicly this week. Listing AI asks a user to provide all kinds of data about the house, and then the software tweaks it so that it can be better used by AI.

After that, the information is processed by GPT-3, an AI model of the non-profit research company OpenAI. GPT-3 has been trained on the text of billions of web pages so that it is adept at responding to written prompts generating everything from news articles to recipes to poetry. (Real estate descriptions are among the hundreds of apps developers have considered for the model, with varying levels of success.)

Users can sign up for a free trial which allows you to generate multiple ads. If you want to generate unlimited descriptions, it currently costs $ 9 per month or $ 84 per year (website says these are introductory prices; regular prices will be $ 12 per month or $ 105 per year. ).

There were obvious mistakes and weird AI decisions in the listings the website made up for me, ranging from misrepresentation (the house does not have a “finished basement”, although it did a basement) to the final sentences (“The living room”) which has just died out. Listing AI also has more social media-friendly versions of its house descriptions, which didn’t seem too different to me, but each came with an asking price for the house (they ranged from around $ 600,000 to $ 2.4 millions of dollars).

However, the overall tenor and style – and the fact that GPT-3 has no knowledge of the actual appearance or floor plan of the house – surprisingly reflected reality.

“This Mediterranean style house is perfect for you!” exclaimed one listing the generated website. It ended with a call to action: “Come take a look today!”

Human contact

After experimenting with Listing AI, I asked the real estate agent representing my family in the sale of my mother-in-law’s house, Scott Ward, what he thought of this kind of tool. Ward, a Red Oak Realty realtor in Oakland, had some issues with Listing AI’s word choices – he would never describe the Oakland home as “rustic-chic,” he said, and he said. hate the word “bragging” in property descriptions.

But he agreed that there are a lot of poorly written house descriptions and that automating the process might help some people. For example, it might be useful if people are unfamiliar with this type of writing process or are selling a number of similar homes, such as rental units.

Still, he said the hypothesis that AI, rather than a person, can be used to showcase a home to potential buyers worries him.

“I think parts of this business can certainly be automated, but there is still too much human contact,” he said. “And I think that’s not necessarily a bad thing.”

Correction: An earlier version of this article misidentified Corey Pollock’s job title at Shopify. He is a senior product manager there.


Previous Buy engagement ring insurance - Forbes Advisor
Next Bank of America will provide groundbreaking funding to BIPOC affordable housing developers.

No Comment

Leave a reply

Your email address will not be published.