When asked if two unfamiliar photos of faces depict the same person, a human respondent will answer correctly 97.53 percent of the time. New software developed by Facebook’s artificial intelligence research group scores a close 97.25 percent correctly on the same challenge, regardless of variations in lighting, or whether the person in the picture is directly facing the camera.
This is a significant improvement over previous face-matching software, demonstrating the power of a new approach to artificial intelligence known as deep learning, something Facebook and its competitors have invested in heavily during the past year. This area of AI involves software that uses networks of simulated neurons to learn to recognize patterns in large amounts of data.
Basically, deep learning software attempts to mimic the activity in layers of neurons in the neocortex, which represents 80 percent of the brain where thinking occurs. The software literally learns to recognize patterns in digital representations of sounds, images and other data. The idea is that software can simulate the neocortex’s large array of neurons in an artificial “neural network.” This is actually decades old, and it has led to as many dead ends as breakthroughs.
But thanks to improvements in mathematical formulas and increasingly powerful computers, scientists can now model many more layers of virtual neurons than ever before. This has led to some pretty impressive advancements in speech and image recognition.
According to Yaniv Taigman, a member of the AI research team at Facebook, “You normally don’t see that sort of improvement. We closely approach human performance.” The error rate has been reduced by over a quarter, compared to earlier software designed to do the same task.
Facebook’s latest software, referred to as DeepFace, does what researchers call facial verification. That is, it is able to recognize that two images show the same face. This is not to be confused with facial recognition, which is putting a name to a face. Some of the underlying techniques could also be used to resolve that issue, according to Taigman. This might contribute to improving Facebook’s accuracy at suggesting whom users should tag in a newly uploaded photo.
For the time being, DeepFace is a research project. Facebook released a research paper on the project recently, and the researchers will present the work at the IEEE Conference on Computer Vision and Pattern Recognition in June 2014.
DeepFace processes images of faces in two steps. First, it corrects the angle of a face so that the person in the picture faces forward, using a 3-D model of an “average” forward-looking face. Then, the deep learning comes into play as a simulated neural network works out a numerical description of the reoriented face. If DeepFace comes up with similar enough descriptions from two different images, it decides they must show the same face.
The deep-learning part of DeepFace consists of nine layers of simple simulated neurons, with more than 120 million connections between them. To train that network, Facebook’s researchers tapped a tiny slice of data from their company’s hoard of user images – four million photos of faces belonging to almost 4,000 people.
Although Facebook has reassured that DeepFace is merely a research-driven project, and will not affect the 1.23 billion people who regularly use Facebook, it’s clear that there’s more to come. Facebook CEO Mark Zuckerberg has expressed deep interest in building out Facebook’s artificial intelligence capabilities when speaking to investors in the past. His ambition actually goes beyond facial recognition to analyzing the text of status updates and comments to decipher mood and context.
It’s not just an intellectual pursuit; understanding all the information we post on the social network is central to Facebook’s business model, which uses data to personalize ads so users are more likely to click on them. The company’s growing ability to recognize users from photos posted has caught the attention of privacy advocates and government officials. Not too long ago, privacy-conscious governments in Europe have demanded that Facebook delete all of its facial recognition data.
This article discusses Facebook’s latest development on the deep learning and facial verification technology front: DeepFace. Although it tows the line on user privacy, it remains purely a research project.
CIPP Exam Preparation
In preparation for the Certified Information Privacy Professional/Information Technology (CIPP/IT), a privacy professional should be comfortable with topics related to this post, including:
- Purposes and uses of PII (I.C.c.)
- Privacy expectations (II.A.)
- Personalization – end user benefits and privacy concerns (II.C.a.; II.C.b.)