Artificial intelligence researchers at Facebook are set to unveil a new system that can identify objects in photographs. While this is not an entirely new idea, Facebook's AI Research (FAIR) team says that it has reached a new milestone, meaning that recognition is now much faster and requires less training.
Any AI-driven recognition system is built on sample data which can be used as a reference point. FAIR's new system needs just a tenth of the amount of training data than other systems, and operates 30 percent faster. But the team's progress doesn’t end there -- great strides have also been made in natural language understanding and predictive learning.
Image recognition and natural language understanding have been married in a way that enables individuals to ask questions about photographs and be supplied with a response. In a post on the Facebook blog, you can see a video that shows the system's ability to correctly answer questions such as 'is there a baby in this photo?'.
It's easy to see how these features could be integrated into Facebook at some point in the future, but it's less clear how the sample uses of predictive learning fit into the social network. Another video on the blog shows the system learning from experience:
Unsupervised, or predictive learning, is the ability to understand what will happen in the future by learning from observation. To try to give computers this ability, the FAIR team has developed a system that can "watch" a series of visual tests -- in this case, sets of precariously stacked blocks that may or may not fall -- and predict the outcome. After just a few months' work the system can now predict correctly 90 percent of the time, which is better than most humans.
It's more likely that this technology, and an AI-powered game-playing bot showcased by the team, would feature in Facebook's future ventures into virtual reality.