Breaking News Miscellaneous Social Media

Pinterest Forms Labs Unit to Explore AI, Machine Learning

Pinterest is placing more focus on machine learning and artificial intelligence with the start-up of a new research unit.

The new Pinterest Labs unit is run by the company’s chief scientist Jure Leskovec, whose team consists of top researchers, scientists and engineers that work throughout the various divisions of Pinterest.

The goal, Leskovec said in a blog post, is to “tackle the most challenging problems in machine learning and artificial intelligence” while working on “image recognition, user modeling, recommender systems and big data analytics.”

Pinterest LabsLabs will partner with the Berkeley Artificial Intelligence Research Lab, the University of California San Diego and Stanford University on the project. Findings will be shared on its website, as well as through research papers, releasing datasets to advance academic research, coming up with challenge problems for research communities to work on and giving tech talks, which will be open to the public.

“With more than 100 billion objects, we’re working with the world’s largest image-rich data set that mixes technology and human curation,” Leskovec said. “This unique opportunity enables us to analyze trends, understand intent and predict consumer behavior. We’ve just scratched the surface of what’s possible to achieve by developing cutting edge algorithms.”

Working with experts from outside of the company as well as inside should translate into a speedy formation of the artificial intelligence for the taste graph as well as quicker personalized recommendations.

“There’s a lot of talk about the future of AI and machine learning and where it can take us in the next 10 years,” Leskovec said. “While we’re committed to shaping the future of AI, we’re also inspired by working on technology that’s reaching everyday people today. We’re not waiting on the future of machine learning, we’re bringing it to more than 150 million people on Pinterest. For example, our systems rank more than 300 billion objects per day, and in just the last year we’ve increased the number of recommendations we serve by 200 percent, while making them 30 percent more engaging.”

Those who are interested in the project can stay abreast of new developments at


About the author


Jennifer Cowan

Jennifer Cowan is the Managing Editor for SiteProNews.

1 Comment

Click here to post a comment
  • The challenge for Pinterest will not be in the data they generate or even mapping it to a certain few points. But recommender systems don’t understand emotional data, and I’m not talking about moods, say something like nostalgia.. they don’t understand that style profile, fashion products vary from one geography to another, or even dialects and languages. I’m sure they want to consider it all and are probably working towards them, but the semantics are too vast. For e.g. In our line in India in the north Silk Sarees are called Resham, while southern India is about Pattu Sarees. Figuring this out is easy, but to show a pattu saree to someone who’s originally from south of India but accessing the site from the north.. how does one solve this bit?