Facebook's brain is getting smarter.
The social network has achieved "new milestones...in Facebook's long-term artificial intelligence research," the company said this week. That includes a "state-of-the-art system that segments, or distinguishes between, objects in a photo" 30 percent faster and using 10 times less training data than previous industry benchmarks.
"Our AI can now look at a photo, figure out what's in it and help explain it to you," Facebook chief Mark Zuckerberg wrote in a post. "We see AI as helping computers better understand the world—so they can be more helpful to people."
The Facebook AI Research (FAIR) team has been training systems to recognize patterns in pixels, so your computer or smartphone can distinguish and identify objects in a photo. "This is especially helpful if you're blind or can't see the photo," Zuckerberg said.
So when a friend posts a photo of their cat sunbathing on the sofa, Facebook's AI can describe the snapshot to you in enough detail for you to form a mental image.
But don't expect to turn your News Feed into an audiobook any time soon; according to Zuckerberg, the technology is still in the early stages. The FAIR team will present its newest work at the NIPS artificial intelligence conference next month.
Facebook also detailed new developments in natural language understanding via its Memory Networks (MemNets) technology.
"MemNets add a type of short-term memory to the convolutional neural networks that power our deep-learning systems, allowing those systems to understand language more like a human would," Chief Technology Officer Mike Schroepfer wrote in a blog post. "We've scaled this system from being able to read and answer questions on tens of lines of text to being able to perform the same task on data sets exceeding 100K questions."
Combine natural language understanding with computer vision, and you get a system Facebook calls VQA, or visual Q&A, which lets people ask questions about what's in a photo.
"Think of what this might mean to the hundreds of millions of people around the world who are visually impaired in some way," Schroepfer said. "Instead of being left out of the experience when friends share photos, they'll be able to participate."
The FAIR team is also working on longer-term challenges, including unsupervised and predictive learning, in which systems can learn through observation and make predictions.
Individually, these programs are impressive and offer hope for the future. But put them together, and you get Facebook's new AI assistant, M. Introduced in August, the new Messenger-based feature can help you purchase items, find birthday gifts, and book restaurants, travel arrangements, appointments, and more.
"Our AI research … is a long-term endeavor," Schroepfer wrote. "It will take a lot of years of hard work to see all this through, but if we can get these new technologies right, we will be that much closer to connecting the world."