Facebook introduced an AI-based algorithm that can distinguish objects on its own. In the future, it can be used to improve social media or diagnose diseases.
The researchers explained that much of artificial intelligence (AI) still learns and works only with the help of humans. In some cases, this interference is less, but often they require algorithms that are created and curated by living scientists to work.
Now Facebook has shown how some AI algorithms can learn to do useful work without human assistance. The company has built an algorithm that has learned to recognize objects in images.
The Facebook Seer algorithm was taught based on more than a billion images from Instagram, while it decides for itself which objects are similar. For example, it can combine mustache, fur, and pointed ears images into one folder. Then the algorithm was sent a small number of tagged images – so it was able to recognize the images and understand how it worked.
“These results are impressive. Learning algorithms for independent work is a very difficult task, and breakthroughs in this area have important consequences for improving visual perception, ”the researchers noted.
Notably, the Instagram images were not hand-picked to facilitate self-study. Facebook’s research is a milestone for an approach to AI known as “self-learning,” added Facebook Chief Scientist Yang Lekun.
For example, in the future, this technology could be used to train an algorithm to interpret medical images without the need to label scans and X-rays. This approach is already being used to automatically generate hashtags for Instagram images.