General Intelligence
400 Million Indians Might Soon Need To Use Facial Recognition To Access Their Bank Accounts
Four banks are currently testing facial recognition systems for two-week trial periods
Welcome to General Intelligence, OneZero’s weekly dive into the A.I. news and research that matters.
More than 400 million people in India who have a government-sponsored bank account or receive financial assistance from the government may soon have to look into a facial recognition camera to get their money.
The Indian national government is testing a change to its digital identification program, Aadhaar, that would make facial and iris recognition necessary to receive benefits, according to the Economic Times’ tech site ETtech. Aadhaar allows people access to a range of services, including scholarships, pensions, and some welfare programs.
Currently, recipients of these funds access them through ATMs equipped with fingerprint sensors, which would be replaced by iris scanners or mobile devices outfitted for facial recognition. Four banks are currently testing facial recognition systems for two-week trial periods.
The main driver of this shift in biometrics is the coronavirus pandemic, according to the Indian government, which says shared fingerprint sensors are a public health risk. ETtech also cites an error rate of up to 20% for the fingerprint scanners, as many users have warped or eroded fingerprints due to manual labor.
The Aadhaar authentication program facilitated more than 1 billion transactions in June alone.
The two programs that will use the facial and iris recognition systems are Pradhan Mantri Jan-Dhan Yojana, a financial inclusion initiative that has given more than 400 million Indians access to affordable banking, and the Direct Benefit Transfer program, which is how the government deposits money into citizens’ bank accounts for various benefits.
These programs account for a massive amount of money being transferred. The Aadhaar authentication program facilitated more than 1 billion transactions in June alone, according to the Reserve Bank of India.
While the change wouldn’t extract more data from Indians — the Indian government already has 1.2 billion citizens’ data as a part of Aadhaar — the move would entrench facial and iris recognition devices into the country’s infrastructure.
The use of Aadhaar facial recognition data also has surveillance implications. Public officials weighed mobilizing Aadhaar’s enormous cache of face images to fight the coronavirus by connecting it to fever cameras or even drones in public places, and then running facial recognition on those whose temperature measured above 100 degrees.
Collecting and using this data at all cements the technology as an inevitable part of how the government interacts with its citizens. It represents another step in India’s push toward normalizing facial recognition, alongside past efforts to use facial recognition in policing and even to grant entry into places of work.
This may be just the beginning. Banks could theoretically use images taken of customers for all sorts of analysis, including emotion detection, to get more data on their customers.
And now, here are some of the most interesting A.I. papers of the week.
An open source Boston Dynamics–esque robot dog
Roboticists Maurice Rahme and Adham Elarabawy made a tiny open source version of Boston Dynamics’ Spot Mini, which they call Open Quadruped (also sometimes Spot Mini Mini). Just like its scary older sibling, Open Quadruped can navigate rough terrain, and it seems that the robot has headroom for even more improvement to its movement algorithms.
Reprogram an image generation algorithm with no code
This brilliant piece of research from MIT and Adobe shows how highly advanced algorithms can be paired with simple user interfaces. The team figured out how to get an image generation algorithm to rewrite itself based on a user selecting a few images of subjects they wanted to see combined. For example, if you have pictures of horses being ridden by jockeys wearing hats, you can click on the horses’ heads and then a jockey’s hat, and the algorithm will generate images with horses wearing hats. The example, which you can see in action here, is simplistic, but the core tech is powerful.
Microsoft makes the HoloLens 2 a research device
The HoloLens 2 is packed full of sensors aimed to measure everything about its surroundings and the person wearing it. Now, Microsoft is giving researchers tools to access those sensors more easily. This research paper outlining the headset’s hardware also has a bunch of nuggets for nerds, like the fact that the HoloLens sensor for detecting hand movements has a 45 Hz refresh rate.
