adam121 - Fotolia

AI hype doesn't stop Trulia from using new analytics tools

Online real estate listing site Trulia is using artificial intelligence tools in meaningful ways, overcoming what many people see as distracting hype.

The AI hype train left the station long ago, and many businesses are now seeing if they can -- or even should -- try to catch up to it.

Real estate tracking site Trulia is one enterprise that has found success with artificial intelligence, though the company did it without jumping on the latest and greatest platforms. Instead, it took an incremental approach to AI adoption that is starting to pay off in business value.

"We're at a stage where everyone is trying to define [AI], and there are pockets of success," said Deep Varma, Trulia's vice president of engineering. "So, it's not fully hype. It's just a matter of some industries adopting technologies faster or later."

Examples of AI at Trulia

Trulia has rolled out deep learning and AI-based features on its site over the course of the last few years, with more advanced features coming online more recently. The most straightforward is its recommendation engine, which matches users with properties likely to fit their preferences. There is also a customer engagement model that assesses web data to determine the type and frequency of communication that users prefer.

Deep Varma, vice president of engineering, TruliaDeep Varma

Then, there are the more complex models that are closer to AI. A natural language generation model automatically creates text for neighborhood descriptions on the site. One of the most advanced AI features is a computer vision model. This deep learning algorithm visually evaluates each image uploaded with property listings. It identifies key property features from photos and uses them to shape listings. For example, if a user has shown a preference for a remodeled kitchen and a property that otherwise meets their search criteria has one, the algorithm will identify the photo of the kitchen and surface it at the top of the listing ahead of all the other images.

Varma said this is about more than just AI hype; it's about engaging users and helping them identify the right properties.

"Those photos are telling you the story," he said. "Just by doing this, we have increased inquiries for our listings."

Tools for building AI features

Software vendors are a major source of the AI hype we see today. Varma said that many vendors slap the descriptor AI onto their products, even if it's a stretch to call what they do AI.

Also, a number of general-purpose AI tools are on the market. Varma said he's dubious about these, since a real estate site like his probably doesn't need the same functionality as an auto parts retailer, for example. This often results in tech companies selling products their customers don't need, he said.

So, rather than investing heavily in vendor products, Trulia has built much of its AI infrastructure around open source tools. This may sound intimidating, but Varma said it doesn't mean the company does everything from scratch. For example, Trulia built its computer vision neural network model around prewritten Python code pulled from an online library. Trulia is also using the open source database Redis and stream processing engine Kafka. New applications are written in Scala and Java.

Of course, this doesn't necessarily mean open source tools will be right for every enterprise pursuing AI. Varma said the decision comes down to how critical AI is to the business's operations.

"If you believe something is core to your success, you better invest in that area," he said. "But if that's not your goal, don't go and do it from scratch."

AI is here to stay

This isn't the first time AI has enjoyed major hype. In the past, though, its bubble has inevitably deflated, and people's expectations have always been dashed.

Varma doesn't see that happening this time around. Yes, there is a lot of AI hype right now. But too many businesses are seeing benefits from AI applications for the bottom to completely fall out like it has in the past.

Also, Varma pointed to all the work being done in the open source community to develop technically complex models, like neural networks, and share them with everyone, which he said is helping to standardize the application of deep learning models and make them easier to scale to new industries and use cases.

"AI is still in its infancy, but it's here to stay," Varma said. "AI is going to impact all industries. It's important for all industries to recognize this and see how to use this technology to bring value."

Next Steps

Expert panel: Today's AI doesn't live up to the hype

AI can be used for evil just as easily as for good

Deep learning AI applications evolve beyond human capacity

Dig Deeper on AI case studies