Facebook may be best known for its huge social network, but the company is also conducting research in exciting computing fields, including artificial intelligence. The company has been speaking a little more about how far it has come with its machine learning research this week and is preparing to put some of its technology to work in consumer products.
Facebook’s AI Research team (FAIR) has been working to train computers to identify objects in photos, understand natural language and to be able to plan ahead. Apparently, the company’s system is 30 percent faster and uses 10 time less training data than previous industry benchmarks.
The research team has devised a range of tests to teach computers a range of skills from predictive visual learning to understanding how to play Go. All of these ideas share one key theme, object recognition. This is an area that Google has been working on intensely for some time too, but Facebook is also looking to combine the idea with other technologies, such as language understanding.
Most recently, the research team has been working on object description. Facebook’s technology can analyse and pick out objects in both the real world and from pictures, including seemingly odd things like a species of dog. Facebook looks to be integrating this into its social network to help describe images to the visually impaired. In the end, the researchers envision that this type of technology can augment our own memories and provide more useful data at our fingertips.
There are a number of interesting videos in the source link below, if you want to see the technology in action. Founder Mark Zuckerberg shared the video below, which offers a closer look at what the company is up to. Facebook will also be presenting a new paper at NIPS next month, detailing how far the company has come.