The following is selection of articles and videos from different sources around the topic of facial recognition. Essentially, facial recognition is a technology that can identify an individual based on a comparison with a database of facial images. The technology is typically studied as part of the computer vision field, which is a sub-field of artificial intelligence. Currently used facial recognition systems mostly rely on deep learning (please refer to my visualization “What is a Neural Network” for more details on this) and, more specifically, on convolutional neural networks. Vision is an extremely difficult task to teach a computer. Only a few years ago, a computer could not easily identify objects, or even tell apart a person from an animal or an inanimate object. Today, the technology has advanced to the point of identifying specific individuals with a high degree of accuracy. This is a remarkable achievement and a testament of human persistence and creativity. Furthermore, deep learning algorithms are characterized by their ability to improve their performance (or “learn’”) the more times they perform a task. Therefore, the more a facial recognition algorithm is used, the more it learns from its mistakes, and the better it gets at making accurate predictions.
The technology is however far from perfect and, as it is the case with any technology, there are trade-offs to be made with regards to its use. Most recent articles have focused on a particular aspect of facial recognition and that is its use for surveillance and policing. These articles point mainly to two critical issues: (i) a bias towards racial minorities that are not accurately represented in the image databases, and (ii) concerns around privacy around how the images that populate the databases are obtained and used. These are indeed issues that need to be discussed and addressed now, and where appropriate regulation is needed.
As a result of concerns around privacy and civil liberties, there has been some recent emergence of different types of attacks that either aim to prevent detection from facial recognition systems (adversarial attacks), or to populate the image databases with bad data (poison attacks) thus rendering the system more error prone. This has given rise to the creation of some curious new products such as black clothing covered with printed license plates (an example of poison attack, designed to trigger automatic license plate readers) or invisibility cloaks, which are in reality shirts printed with patterns in bright colors (an example of adversarial attack, which relies on a deep-learning optical illusion aimed at preventing the system from seeing the person).
The topic of facial recognition extends however to many applications beyond surveillance and a serious discussion on the topic needs to look at its many applications (both current and potential). One of the most interesting applications of facial recognition is in the field of emotional artificial intelligence. In this case, the technology also relies on an underlying database of facial images, but the main purpose is not the identification of an individual but on the detection on non-verbal cues. This, for example, has proven extremely useful to improve the understanding of and communication with people in the autistic spectrum. Another promising application is in the development of self-driving cars, where facial recognition is critical for the car to be able to identify the existence of a cyclist or pedestrian nearby.
Furthermore, the technology is not only exclusively applied to humans. For example, in farming and agriculture there are uses of facial recognition systems applied to cows. These systems rely on the use of surveillance cameras, computer vision, and predictive imaging to track animals and analyze their behavior.
In the words of Demis Hassabis (AI researcher and founder of DeepMind) “as with any powerful technology, and AI could be especially powerful because it’s so general, the technology itself is neutral. It depends on how we as humans decide to design it and deploy it, what we decide to use it for, and how we decide to distribute the gains”.
The Bias in the Machine, Why facial recognition has led to false arrests
By Sidney Perkowitz (Nautilus – August 19, 2020)
http://nautil.us/issue/89/the-dark-side/the-bias-in-the-machine
Very comprehensive and well researched article that provides a well-balanced discussion on the biases existing in facial recognition when used for policing.
How Facial Recognition Makes You Safer
By James O’Neill (The New York Times – June 9, 2019)
https://www.nytimes.com/2019/06/09/opinion/facial-recognition-police-new-york-city.html
An interesting article written by New York’s police commissioner explaining how facial recognition is used in practice by the New York Police Department.
Facial Recognition: Last Week Tonight with John Oliver
By John Oliver (HBO – June 15, 2020)
https://www.youtube.com/watch?v=jZjmlJPJgug
With its usual mix of research and humor, John Oliver exposes in this video several of the dangers of facial recognition in regard to privacy and civil rights. He also argues for the need for deeper reflection around these topics and quick action to pass appropriate legislation.
Who is using your face? The ugly truth about facial recognition
By Madhumita Murgia (Financial Times – September 18, 2019)
https://www.ft.com/content/cf19b956-60a2-11e9-b285-3acd5d43599e
An in-depth investigative article on the different image databases that exist today, how they were populated, who can access them, and how they are used.
Dressing for the Surveillance Age
By John Seabrook (The New Yorker – March 9, 2020)
https://www.newyorker.com/magazine/2020/03/16/dressing-for-the-surveillance-age
This excellent article focuses on new innovations, mostly around the fashion industry (in a broad sense), to defeat facial recognition systems (either as adversarial or poison attacks). The article also provides a very good explanation of the theoretical underpinnings of facial recognition.
When your tech knows you better than you know yourself
By Ephrat Livni (Quartz – March 14, 2019)
https://qz.com/1570414/facial-recognition-helps-this-tech-tool-read-human-emotion/
The article discusses the work of Dr. Rana el Kaliouby, co-founder of Affectiva, a company that applies facial recognition to emotion recognition.
Should we be worried about computerized facial recognition?
By David Owen (The New Yorker – December 10, 2018)
This piece focuses on some transformative applications of facial recognition in fields such as agriculture, farming, medicine, and sports. These applications are carefully balanced with the risks involved in facial recognition.
Photo credit: @heyerlein Unsplash