alt Hacker NewsFacial recognition is very powerful these days. My friend took a photo of his kid at the top of Twin Peaks in SF, with the city in the background. Unfortunately, due to the angle, you could barely see the eyes and a portion of the nose of the kid. Android was still able to tag the kid.
I feel like Facebook really dropped the ball here. It is obvious that Squire and colleagues are working for the Law Enforcement. If FB was concerned about privacy, they could have asked them to get a judicial warrant to perform a broad search.
But they didn't. And Lucy continued to be abused for months after that.
I hope when Zuck is lying on his death bed, he gets to think about these choices that he has made.
Replies
The EU AI act activates this year. Facial recognition is in the restrictive list. You don't want to give auditors ammunition before it goes live as top fine would cost FB around $4B, and wouldn't be a one time fine.
Even if only law enforcement can use it, having that feature is highly regulated.
[edit] I see this is from years ago. I should read the articles first. :)
> I feel like Facebook really dropped the ball here
This story was from more than a decade ago.
Facebook had facial recognition after that, but they deleted it all in response to public outcry. It’s sad to see HN now getting angry at Facebook for not doing facial recognition.
> I hope when Zuck is lying on his death bed, he gets to think about these choices that he has made.
Are we supposed to be angry at Zuckerberg now for making the privacy conscious decision to drop facial recognition? Or is everyone just determined to be angry regardless of what they do?
I would hazard a guess that the facial recognition will limit the search scope to people associated (to some degree) with your friends account and some threshold of metrics gathered from the image. I doubt it is using a broad search.
With billions of accounts, the false positive rate of facial recognition when matching against every account would likely make the result difficult to use. Even limiting to a single country like UK the number could be extremely large.
Let say there is a 0.5% false positive rate and some amount of false negatives. With 40 million users, that would be 200 000 false positives.
> I feel like Facebook really dropped the ball here
This case began being investigated on January 2014 [0], which means abuse began (shudder) in 2012-13 if not earlier.
Facebook/Meta only began rolling out DeepFace [1] in June 2015 [2]
Heck, VGG-Face wasn't released until 2015 [3] and Image-Based Crowd Counting only began becoming solvable in 2015-16.
> Facial recognition is very powerful these days.
Yes. But it is 2026, not 2014.
> I hope when Zuck is lying on his death bed, he gets to think about these choices that he has made
I'm sure there are plenty of amoral choices he can think about, but not solving facial detection until 2015 is probably not one of them.
---
While it feels like mass digital surveillance, social media, and mass penetration of smartphones has been around forever it only really began in earnest just 12 years ago. The past approximately 20 years (iPhone was first released on June 2007 and Facebook only took off in early 2009 after smartphones and mobile internet became normalized) have been one of the biggest leaps in technology in the past century. The only other comparable decades were probably 1917-1937 and 1945-1965.
---
[0] - https://www.bbc.co.uk/mediacentre/2026/bbc-eye-documentary-t...
[1] - https://research.facebook.com/publications/deepface-closing-...
[2] - https://www.cbsnews.com/news/facebook-can-recognize-you-just...
Facebook rightly retired their facial recognition system in 2021 over concerns about user privacy. Facebook is a social media site, they are not the government or police.
When people on hacker News talk about requiring cops to do traditional police work instead of doing wide ranging trawls using technology, this is exactly what they meant. I hope you don't complain when the future you want becomes reality and the three letter agencies come knocking down your door just because you happened to be in the same building as a crime in progress and the machine learning algorithms determined your location via cellular logs and labelled you as a criminal.
[dead]
Google photos has the advantage of a limited search space. Any photo you take is overwhelmingly likely to be one of the few faces already in the library. Not to say facebook couldn't solve the problem. But the ability of Google to do facial recognition with such poor inputs is that it's searching on 40~ faces rather than x billion faces.