It’s hard to live in Seattle without seeing an article on home prices or artificial intelligence every day, but this article about Zillow
is actually a cool demonstration of how much AI is changing the game in industries like real estate.
Zestimates and Redfin estimates have already been getting very good over the last several years. According to both companies
, their median national error rate on estimates is less than 2%, and both companies are putting their money where their mouth is by offering “instant buying” programs to home sellers.
However, models based on location, square footage, and number of bedrooms are missing a lot of the data that human buyers and agents take into account when appraising a house. For example, after seeing a few photos of a house, buyers can quickly assess things like curb appeal, views, and the quality of finishes in a house.
According to Zillow, though, the new Zestimate uses deep learning models to analyze photos, so things like granite countertops and remodeled bathrooms can be factored into a home’s estimated value. Now that these models have more of the data available to them that people have when appraising a home, it will be interesting to see how much better their estimates become.
It will also be interesting to watch how the real estate sites handle “virtual staging
” companies that can quickly clean up, furnish, and even remodel a home using special effects and give customers beautiful photos to send to the machines. 🤖