Facebook has further stuffed its AI toolbox with the acquisition of retail artificial intelligence developer GrokStyle.
Facebook has not disclosed any details of the deal, and GrokStyle’s web page has miraculously shrunk to a single landing page, saying “thank you” to anyone that stumbles across it.
It claims that it is “incredibly proud of what we have built – and believe we have only scratched the surface of what is possible with computer vision.
But, it continues, “Today, we are excited to share that we are moving on as a team. Our team and technology will live on, and we will continue using our AI to build great visual search experiences for retail. As part of this next chapter, we are winding down our business.”
A rummage around the Wayback Machine throws up some cached pages which throw some further light on GrokStyle. It lists the founders as Prof Kavita Bala and Dr Sean Bell. According to LinkedIn, Bala is now chair of computer science at Cornell, having left GrokStyle last July. Bell is listed as a research scientist at Facebook as of January.
GrokStyle pitched its technology as combining visual search and product recommendations, and just under a year ago, apparently scored a deal with Ikea to provide visual search tech for the furniture giant’s augmented reality app.
Thus you could snap a picture of, say, a chair, and the platform will find similar items, and show you what they’d look like plonked in your own living space.
So it’s not hard to envision other ways Facebook could exploit that technology. Yep, you won’t be restricted to looking lustfully at your “friends’” lifestyles. You could immediately buy them – or at least the furniture, clothes, and accessories.
Last year, Facebook bought UK-based Bloomsbury AI, apparently to use its NLP technology to bolster its fight against fake news, amongst other things.
The traffic hasn’t been all one way. Facebook has thrown a fair amount of AI and ML technology into the open source area. Just last week it shared code and models for a cross lingual model pre-training. Last October it shipped the first release candidate for v1 of its PyTorch project, a machine learning library for Python. Last December it opened up both its PyText NLP framework, and the DeepFocus framework used in its Half Dome headset.