Amazon's facial recognition technology can already detect fear

Rekognition, the controversial and hitherto inaccurate facial recognition technology from Amazon Web Services, is now able to detect fear, the company announced this week in its blog.

Earlier this week, Amazon Web Services (AWS) enthusiastically announced the arrival of various updates to its biometric facial recognition technology, introducing improvements in accuracy and functionality.

First, the facial analysis of the seven different emotions on people's faces is improved and optimized: "happiness", "sadness", "anger", "surprise", "disgust", "calm" and "confusion" . An eighth emotion was also added: fear, previously undetectable by this software.

It should be stressed that Amazon Web Services technology generates metadata regarding gender, age, emotions, attributes such as smile, facial expression, facial image quality and face reference points.

One of the improvements is the increase in Rekognition's ability to estimate a person's gender, as well as their age, with technology now providing a narrower range of possibilities for most age groups.
The announcement of the updates came peppered with controversy and caused a furious wave of online criticism, with cloud computing economist Corey Quinn pointing out that the movement seems incredibly timely given the recent controversy surrounding Rekognition.

"Amazon Web Services is criticized for Rekognition's sales to the federal government, which in turn is building concentration camps for children, and the answer is to improve the estimate of age range and fear detection" was expressed on Twitter.

Countless Rekognition tests have already revealed the shortcomings of technology, and the day after Amazon published the new blog post, the American Civil Liberties Union of Northern California held a press conference to announce a major error: the software Facial recognition mistakenly identified one in five California lawmakers as criminals.

"If they falsely accuse you of an arrest, what happens?" Said Phil Ting, one of the falsely matched lawmakers, during the press conference. “It could affect the ability to get a job and it absolutely impacts the ability to get housing. There are real people who could have real impacts, ”he denounced.

Despite the current controversy, improved facial analysis models are now available for Amazon Rekognition Image and Amazon Rekognition Video, and are the new default option for customers in all supported AWS regions. Machine learning can be used without previous experience.