It may not be a name you associate with machine learning, but Pinterest is increasingly pouring resources into AI in order to surface relevant ideas for its 150 million users.
What happens when you add machine learning to a database of 100 billion image-rich objects and ideas? Pinterest is already scratching the surface of a potential answer to that question with its AI-powered tools, including visual search and Pinterest Lens, but now it wants to dig deeper.
The company wants to join the ranks of industry giants Facebook and Google in accelerating the growth of artificial intelligence through open research and collaboration. To help it achieve that goal, it is launching a new group — dubbed “Pinterest Labs” — comprised of machine learning experts, whose investigations could help transform the way users discover ideas.
In the words of Pinterest chief scientist and Stanford associate professor Jure Leskovec: “As much as we’ve done, we still have far to go — most of Pinterest hasn’t been built yet.”
By working with the research community and universities — such as the Berkeley Artificial Intelligence Research Lab, University of California San Diego, and Stanford University — Pinterest Labs is hoping to build the AI systems for some of its integral features. These include the “taste graph,” the technique used by the company to map the connections between pins, people, and boards in order to surface relevant ideas for users. The company is also hoping machine learning can help it to provide personalized recommendations faster.
Those interested in its work can keep up with the research group on its dedicated website, and by attending its public tech talks, the first of which took place on Tuesday at the company’s headquarters in San Francisco. Pinterest Labs will also share its findings with the academic community by publishing research papers and releasing its data to researchers.
Leskovec claims that Pinterest’s systems now rank more than 300 billion objects per day. In the last year, the platform has increased the number of recommendations it serves by 200 percent, while making them 30 percent more engaging.
Pinterest took a big leap in to machine learning with the launch of its Pinterest Lens tool at the start of this month. The machine learning system that powers Lens can recognize objects in photos, along with identifying their features, such as color, allowing users to snap images with their smartphone camera in order to discover and purchase related items on Pinterest.