Accuracy of facial recognition technologies from different companies. Retrieved from: https://sitn.hms.harvard.edu/flash/2020/racial-discrimination-in-face-recognition-technology/
As of April 2020, the National Institute of Standards and Technology (NIST) reported that the leading face identification algorithm had an error rate of only 0.08%, compared to 4.1% in 2014.
Previously, questions were raised about the accuracy of facial recognition regarding demographic bias. One of NIST's 2019 test results found false match rates of 0.1% for black women and 0.025% for black males. These results were respectively 10 and 2.5 times higher than those for white participants.
However, more recent research by NIST refuted previous assumptions of significant demographic bias in leading facial recognition technology systems. Seventeen of the best facial recognition systems demonstrated false-negative rates of 0.49% or less for black females and no more than 0.85% for white males.
The statistics on accuracy have clearly shown a marked improvement over time. It demonstrates a significant advantage over humans in learning and processing information.
This is because the facial recognition system's AI learns from its mistakes and self-corrects based on the data it receives. Whenever the AI successfully matches facial images, it remembers the process and improves its accuracy.