For example, Facebook or Twitter will want to train the neural network to recognize the faces of people at different ages. What will help better than the thousands of old photos that users will send themselves?Actress Lily Collins in a # 10yearchallenge flashmob
In January 2019, a flash mob returned to the social network with comparisons of their photos from the past and present. This is a traditional format: users show how they have changed in a year, in two or five. In 2018, there was already a challenge # 2012vs2018 . Now it’s the turn of the # 10yearchallenge (or # 2009vs2019challenge ) – as the name implies, it is dedicated to images taken 10 years ago.
The flash mob was picked up by celebrities and brands, and can be found on Twitter, Instagram, Facebook or VKontakte. Against this background, Wired columnist Kate O’Neill (Kate O’Neill) asked a question: what if the new Challenge was “cunning” of large social networks to collect relevant data users?
— Reese Witherspoon (@RWitherspoon) January 15, 2019
2009 vs 2019 pic.twitter.com/I7xH8P4qtb
— Marques Brownlee (@MKBHD) January 13, 2019
2009 vs 2019 I’ve come a long way pic.twitter.com/hovGtfk60o
— Jarrod Alonge (@JarrodAlonge) January 13, 2019
The journalist stressed that she does not claim that the flash mob appeared that way. But, in her opinion, formats like # 10yearchallenge would be very useful for Facebook or Twitter to train face recognition algorithms.
[perfectpullquote align=”full” bordertop=”false” cite=”” link=”” color=”” class=”” size=””]Imagine that you want to train an algorithm for recognizing faces based on age characteristics and, more precisely, age progression (that is, finding out how people will look with age). Ideally, you need a dataset with a bunch of photos of people. Even better, most of the pictures are made with the same difference. For example, ten years.[/perfectpullquote] [perfectpullquote align=”full” bordertop=”false” cite=”” link=”” color=”” class=”” size=””]Of course, you can just take a photo from the pages on Facebook, but there [is] stored a lot of useless. And here – a huge, but at the same time already “cleared” and labeled data set.[/perfectpullquote]
In other words, O’Neill writes, the flashmob turns into a great foundation for training the algorithm. And any humorous options (like the tweet “My cat in 2009 / my cat in 2019”) will be eliminated by the neural network itself. “Even if this particular meme is not an example of social engineering, then companies have already done this in a similar way – remember Facebook and Cambridge Analytica ,” the author added.
The journalist asked readers in the future to think about how to interact with technology, what data to share in social networks and how they can be used on a larger scale. O’Neill called three ways how to use the photo from the flashmob “2009 vs 2019”: acceptable, ordinary and risky.
- The algorithm of recognition of persons in the age progression will help to find the missing children after a few years. Neural network will be able to predict how they will look matured;
- The algorithm will help companies show different advertisements to people of different ages and adapt with time. The date of birth will not be needed – just a photo;
- The algorithm can be embedded in the insurance and health care system. If the neural network decides that you are aging faster than an ordinary person, then you may be denied health insurance.
— Сергей Д (@sd0107) January 15, 2019
— Одна тут отдыхаешь? (@Livotovas) January 15, 2019
— таня (@kotanya) January 15, 2019
— Rio Ferdinand (@rioferdy5) January 15, 2019