Facial Recognition Technology -- a very important aspect of Artificial Intelligence -- that can enable you to identify or verify a person or more from a video frame or digital image, has the potential of bringing very appreciable improvements in the quality of work and living. In fact, it is already proving itself useful in various areas related to business as well as domestic purposes.

Corporations such as Google, Adobe Systems, Uber, and Tesla are making extensive use of Image Recognition, and as per an Adoriasoft report, researchers have made a prediction that the global market of Image Recognition is going to reach $38.92 billion by 2021.

However, the adoption of this technology has had its fair share of controversy which has brought up the question in many people’s minds whether the technology is stable or accurate enough for mass usage, or whether the application of such technology amounts to invasion of privacy, and the repercussions of a mistaken identity by the Recognition System.

An Unfortunate Case In Point:

Amazon’s Facial Recognition Technology, for example, had erroneously identified 28 members of Congress as people who had been arrested for crimes.

In May 2018, when it was revealed that Amazon had been marketing and selling its software Rekognition to police agencies, the American Civil Liberties Union (ACLU) of Northern California tested Rekognition. It was then found that people of color had been disproportionately misidentified in a database of mugshots. This raised concerns such as loopholes for racial biases, as well as chances of abuse by law enforcement. As a result, Amazon’s CEO Jeff Bezos was urged by the people standing for privacy to discontinue providing the government with Rekognition.

Jacob Snow, a technology and civil liberties attorney at Northern California’s ACLU Foundation stated that their test emphasizes the fact that face surveillance can be unsafe for government use. It can be used to give strength to surveillance that is full of discrimination. It can also be used to strengthen policing that could be targeting activists, immigrants, or communities of color. Mr. Snow opined that such damage, if and when done, would be impossible to mend.

The ACLU Test:

ACLU had decided to use the same facial recognition system that Amazon presents before the public, for the purpose of scanning for matches between the various images of faces in a database. 25,000 public arrest photos were used to form a search tool and face database. These 25,000 photos were cross-referenced with public photographs of each and every member of the US House and Senate.

11 out of the 28 who got misidentified in the test were people of color.

Even though people of color form only 20% of the strength of Congress, 11 out of 28 meant almost 40% of those who were erroneously matched by the software. Civil Rights Leader John Lewis was included in the false matches too, along with 5 other members of the Congressional Black Caucus.

Following this test of Amazon by the ACLU, the executive director of the Center For Media Justice, Malkia Cyril said that now it had been doubtlessly established that Amazon’s facial recognition tool could be dangerous to democracy as it is discriminatory, and therefore needed to be rolled back at the earliest.

The Defense:

In defense of its technology, Amazon stated that such a situation might not have arisen at all. The ACLU had used an 80% confidence threshold, whereas, as per Amazon, customers are advised to set a threshold of 95% or higher, when facial recognition is used for law enforcement activities.

ACLU, in its turn, said that it had used the default match settings set by Amazon for Rekognition, and that Amazon, on its own website, references an 80% confidence metric for facial recognition.

This could well mean that Amazon’s technology could not be expected to work soundly in its default mode. It was noted by Jacob Snow that there was no evidence of Amazon guiding the police to make use of the higher threshold, just as there was no guarantee that the police were abiding by any such instruction. Moreover, the technology allows thresholds to be changed by the law enforcement.

Cyril opined that if the police used Amazon’s technology, thousands of people would be at the risk of being misidentified. This could especially have an impact on people of color, and black people who were already being discriminated against in the US system of policing. Cyril also went on to say that Amazon was “playing politics with people’s lives.”

On the other hand, a spokesperson from Amazon spoke of Rekognition’s beneficial aspects such as

• Curbing Child Exploitation

and

• The Prevention of Human Trafficking.

The San Francisco Situation:

On 29-1-2019 appeared the news in sfexaminer.com that San Francisco Supervisor Aaron Peskin had proposed that Facial Recognition Technology be banned in the city.

In Peskin’s opinion, “…facial recognition technology, which has the biases of the people who developed it, disproportionately misidentifies people of color and women…”

People belonging to Civil Rights groups have raised concerns related to

• Accuracy
• Privacy
• Safety

when it comes to Facial Recognition Technology.

Former BART board member Nick Josefowitz had expressed an interest, some time ago, in facial recognition software that could help in identifying certain individuals, like those with arrest warrants. That was when BART officials had come under fire.

The Board of Supervisors may give their approval for the new surveillance technology only if

• Civil Rights shall remain Protected
• No Group or Community shall suffer a discordant Impact
• Benefits will outweigh the Costs

Peskin put his measure forth as an extension of the policy of putting Privacy ahead of everything. This could be setting transparency requirements and limits on the collection of personal data by companies doing business in the city, as well as city agencies.

While the technology is far from perfect and is still being improved upon, one cannot overlook the obvious benefits it may provide – not as the primary decision-making tool, but more as a technology augmentation to an existing process with proper safeguards in place to handle situations where the Recognition Engine may not work as desired.

For Example:

• Attendance Systems with Manual Override
• Identification of Objects or People within Video or Image
• Image Based Searching
• Usage in Law Enforcement but with Human Override

and

• As a Support Tool with Stricter Thresholds.

To Conclude:

Facial Recognition Technology is not a silver bullet. It has its strong and weak points, advantages, and limitations. Therefore, while drafting a solution around this technology, it is important to develop a balanced solution, keeping all these factors in mind.

Are you considering harnessing the benefits of Facial Recognition Technology to make work easier at your business, factory, or residence?

• To back-up Face Matching for Security/ Attendance
• To Identify Celebs for your newspaper/magazine
• Sundry other purposes

Do feel free to contact us any time at all. We would be delighted to design and customize a solution based on your idea, so that you can enjoy all the benefits of Facial Recognition Technology, and profit from the process.