AI uses something like a Pinterest board of our faces to detect, analyze and identify us. Is it an invasion of privacy or a desperate solution for our safety? Read on.
We are all unique. No, this is not pep talk; it’s a fact. We are all physically unique. Humans come with identifiers that are impossible to be found in others. Our fingerprints, DNA and even faces for the most part (one in every 42 children is born a twin). There are distinguishing features in all of us that have proven beneficial for security purposes across industries. You can unlock your phone screen using facial recognition and look into a camera to get through immigration quickly. But how does this technology work, and is it all that great? Find out.
Coming face to face with facial recognition technology
Using specific software, AI-powered face recognition technology gathers facial data to confirm someone’s identity. It identifies and measures facial features (your eyes, lips, chin and more), and then compares them with an existing database.
Largely, it employs three steps: detect, analyze and recognize. The first stage—detection—uses computer vision to detect people. It gathers facial information using complex AI tech.
The second step is analysis. After it has gathered extensive data, such as the distance between your eyes, the shape of your cheekbones and such, it assigns the face a numerical code. Much like fingerprint tech, it uses unique facial landmarks to analyze the person on the other side of the camera.
Finally, in the recognition stage, it matches the facial features to existing images in the database.
How does AI face recognition work exactly?
It primarily relies on machine learning and deep learning algorithms. Machine learning is an AI extension that imitates the way humans learn. In doing so, it becomes more and more accurate over time. A common algorithm used for facial recognition technology is deep learning Convolutional Neural Network (CNN). CNN takes unique features from images or videos and then puts them in different categories.
When the technology is provided with a photo or video, its first task is to find a face. It does so by focusing on about 68 distinct features that typically make up a face. It’s looking for texture, edges and the like. For a machine to be able to do so, it has to be taught what a face is. That’s its creator’s job. The creator informs the machine how to tell a face from a wall. This feat is “learned” by the machine.
Then, the technology turns its findings into mathematical code and hunts for similar codes in its existing database. Where does this database come from? It varies. When law enforcement uses it, it could be comparing existing mugshots and surveillance footage. For immigration, the government’s collection of our images is enough. Now, owing to social media, technology has access to a greater database.
Ultimately, if all goes well, it confirms the person’s identity accurately.
What is the accuracy of face recognition technology?
It is so-so. Many factors contribute to the technology, including lighting, colors, camera quality and more. If these are not optimum, the technology’s results won’t be either.
Moreover, studies have found that face recognition technology has a race problem. A 2018 report found that AI facial recognition tech is accurate on white men almost 99 per cent of the time. However, it is far less accurate on people with darker skin, successful only for about 35 per cent of dark-skinned women. Features like these have landed the tech in deep waters. For instance, in 2015, Google’s photo app labeled African Americans as “gorillas”. It issued an apology after that, but it also brought to light the long way AI-powered face recognition tech has to go.
Some people believe that facial recognition technology is “socially toxic” by design, in that it further propagates racial, gender and other such disparities. Plus, it can raise red flags for privacy.
How can you protect your privacy when someone is using facial recognition?
If you’re worried about your facial data being leaked through your own devices, consider opting out of facial recognition settings. In the wrong hands, this data could be used to access and abuse private information.
That said, the average citizen has little control over the use of this technology in public. With surveillance cameras everywhere and social media inundated with images of people, you cannot escape being assigned a numerical code in the recognition technology’s system.
Some researchers are counting on face masks to confuse facial recognition technology. Even factors like aging or wearing sunglasses can hamper the technology’s effectiveness. However, sometimes, you could give certain organizations the benefit of the doubt, given that they might genuinely be looking out for safety.
- What’s the Difference Between Facial Recognition and Facial Detection?
- Personal.ai: Your Personal AI to Pick up the Slack
- Facial Recognition Is No More on Facebook
Header Image by Pixabay