Is the FBI no match for Facebook?
|
I saw a foreign media report that the facial recognition system deployed by the FBI was challenged, and that the FBI's system was not as good as Facebook's, and that it needed its help to improve the accuracy of facial recognition. Is this true? The FBI plans to fully deploy the Next Generation Identification (NGI) facial recognition system this summer. By the end of this year, the NGI facial recognition system will enter a large amount of image information involving citizens in all 50 states into the central federal database. The NGI facial recognition system is not perfect. The Electronic Frontier Foundation (EFF) seems to be deliberately pouring cold water on the FBI's NGI facial recognition system: EFF believes that NGI is not perfect. Assuming there is a photo of a suspect's face, the NGI system will display a suspicious list of up to 50 people, and the probability of successfully locking the suspect's name information is only 85%. Compared with the DeepFace facial recognition system just announced by Facebook earlier this month, NGI pales in comparison. The DeepFace system can tell whether the people in two pictures are the same person. The accuracy of this technology is as high as 97%, which is even comparable to human recognition ability. There is no doubt that Facebook's advantage lies in the data value of the network. Under the same circumstances, it can perform facial recognition in a smaller range, and the success rate is naturally higher. Just based on this point, even the most powerful law enforcement agencies in the United States can only sigh in front of Facebook. Some people pour cold water on facial recognition technology. Although many people are happy to see this "close to human level" facial recognition technology born in the world, the real facial recognition technology is far from as easy as the industry promises. Shahar Belkin, chief technology officer of FST Biometrics, a well-known facial recognition technology development company, said: "Those commercial companies that claim to have powerful facial recognition technology are lying, because the gap between the human brain and the computer is huge." FST Biometrics' customer base is willing to sit in front of the camera to cooperate with the work, and its recognition technology has multiple layers of verification links, such as height and gait tracking technology, which can eliminate active false information as much as possible. The FBI's NGI facial recognition system does not have these advantages. What it calls facial information is like a fingerprint, one fingerprint corresponds to one identity, and is more suitable for use in court. But human facial information is different. People can easily distinguish and identify facial information, but machines cannot. FBI's disadvantages and Facebook's advantages What's worse is that the quality of the pictures used by the FBI's facial recognition technology is worrying. Belkin said that the facial recognition system requires frontal pictures and picture information from no less than 15 angles with the face as the central axis. Such pictures are often presented consciously. Pictures taken by cameras in public areas are not suitable for facial recognition systems. For the FBI, this is a very tricky problem. Because surveillance cameras are generally placed on ceilings or street lights, even if the suspect is photographed, it is difficult for the facial recognition system to distinguish such pictures due to problems such as angle and picture quality. Nowadays, the surveillance camera systems of various countries were not initially established to work in coordination with facial recognition systems. However, Facebook can avoid this problem because Facebook knows who your friends are, which is crucial to facial recognition systems. Moreover, Facebook can find a large number of relevant photos. In terms of the total amount of picture information, Facebook can use up to 250 billion pictures, while the FBI only has 50 million pictures, which is completely different. The benefits of a large number of pictures are obvious - there is more room for choice, and a large number means more relevant information, which is of great benefit to the accuracy of facial recognition technology. The FBI may only need a piece of paper to get Facebook's help. The FBI can still build an excellent facial recognition system. What it needs is more pictures, more names, and a smarter network to organize them together. In short, the FBI needs Facebook's help, and the cost of such help is also very low - only a court order is needed. At present, Facebook is in a lawsuit because of user privacy allegations. Once it loses the lawsuit, law enforcement agencies are likely to "obtain" more user data, which is exactly what the law enforcement agencies dream of. The facial recognition system is necessary. Like other cutting-edge technology competitions, whether it is Mac PK.PC or Android VS.iOS, first-come-first-served will take the initiative, but "first-served" cannot ensure a competitive advantage. If Facebook can make its facial recognition system work perfectly, the FBI can also do this through various means. The question now is how well the FBI's GNI compares to Facebook's facial recognition system. This is an open question. But either way, the FBI may lose to cameras and computers, but it will still win in court. As a winner of Toutiao's Qingyun Plan and Baijiahao's Bai+ Plan, the 2019 Baidu Digital Author of the Year, the Baijiahao's Most Popular Author in the Technology Field, the 2019 Sogou Technology and Culture Author, and the 2021 Baijiahao Quarterly Influential Creator, he has won many awards, including the 2013 Sohu Best Industry Media Person, the 2015 China New Media Entrepreneurship Competition Beijing Third Place, the 2015 Guangmang Experience Award, the 2015 China New Media Entrepreneurship Competition Finals Third Place, and the 2018 Baidu Dynamic Annual Powerful Celebrity. |
>>: Domestic smart watches: groping in the dark
Recommend
How to choose safer soy sauce? What are the cognitive misunderstandings? Learn more in one article →
Recently, reports that cadmium and arsenic were d...
What kind of data analysis should be done from the start-up to the maturity of APP?
DAU, MAU, retention rate , frequency, duration......
Breaking news! Today, all Meituan’s apps were removed from the App Store! What is the reason?
Just now, the top APP promotion company suddenly ...
Learn the best practices of modern Android development in one article
Author: Wang Peng, Sun Yongsheng What is MAD? h...
How much does it cost to be a tea mini program agent in Hexian market?
Is it easy to be an agent of Hexian Tea Mini Prog...
There is no duck in the duck blood for hot pot? 15 types of duck blood were tested, and the results were unexpected!
After testing the authenticity of mutton skewers ...
The whole process of information flow data analysis, with examples~
I heard that everyone has been tortured by data a...
2020 Huamanlou Finance original full set of tactics video tutorial
Introduction to the 2020 Huamanlou Finance origin...
This "lotus leaf" is not actually a plant!
Speaking of "ceramics", there is no dou...
Is it expensive to produce the Jixi snack mini program? Jixi snacks applet production cost and process
There are two types of Jixi snack WeChat applets ...
I'm so tired because I'm poisoned!!
Audit expert: Yin Tielun Unlike the energetic &qu...
TV games can also be played like this. LeTV game console brings arcade pleasure
Speaking of fighting games, if you tell me that y...
Do children only have stuffy noses, runny noses, and sneezing when they have a cold? That may not be the case...
gossip In spring, if children repeatedly experien...