Hewlett-Packard is looking into the facial recognition technology used in some of its PCs after a YouTube video surfaced that seems to show the software and camera failing to recognize an African-American man.
The YouTube video of the facial recognition failure, which runs about 2 minutes, 16 seconds, was originally posted on the video-sharing Website Dec. 10. On Dec. 21, HP posted a blog response to the video, which assures customers that it is looking into the question of why the software seems to recognize the white woman featured in the video but not her black co-worker.
“Everything we do is focused on ensuring that we provide a high-quality experience for all our customers, who are ethnically diverse and live and work around the world,” said an HP blog post by Tony “Frosty” Welch, the lead social media strategist for HP’s Personal Systems Group. “That’s why when issues surface, we take them seriously and work hard to understand the root causes.”
The video shows two people, Desi, who is black, and Wanda, who is white, looking into a camera used with an HP MediaSmart PC. The facial recognition software is supposed to make the camera follow whoever is in front of the computer.
Although the footage on the HP PC does not move when Desi is in front of the camera, it does move when Wanda enters the frame. When Desi re-enters the picture, the software and camera appear to stop. The motion restarts when Wanda comes back into the frame.
“I’m going on record and I’m saying it: Hewlett-Packard computers are racist,” Desi said in the YouTube video, although he appears to be joking and smiling through most of the impromptu demonstration.
“The worst part is I bought one for Christmas,” he said, referring to the HP MediaSmart computer.
It’s not clear from the video how many times the two tried to use the facial recognition software or whether it was simply one malfunctioning computer or a serious problem. However, the video caused enough of a stir for HP to make a blog post about the incident, stating that the company has started looking into whether there is a problem with the software.
“The technology we use is built on standard algorithms that measure the difference in intensity of contrast between the eyes and the upper cheek and nose,” Welch wrote in the blog post. “We believe that the camera might have difficulty ‘seeing’ contrast in conditions where there is insufficient foreground lighting.”