Nos últimos tempos o Facebook e o Instagram têm sido ‘tomados de assalto’ por publicações onde os autores mostram duas fotografias, uma atual e outra tirada há dez anos. Trata-se do #10yearchallenge, um desafio que pode servir para ensinar à Inteligência Artificial (IA) do Facebook mais sobre o processo de envelhecimento dos seus utilizadores.
Ainda que nada exista que possa suportar esta teoria, a hipótese é levantada por uma autora e oradora, Kate O’Neill, especializada em tecnologia na sua página de Twitter. Como pode ler nos ‘twets’ abaixo, O’Neill coloca a hipótese do desafio ser usado para treinar algoritmos e IA, colocando de forma simples duas fotografias da mesma pessoa uma ao lado da outra.
Novamente, não existe nada que suporte esta teoria mas o facto de ter sido criada atesta bem à forma como o Facebook deixou de ser confiável aos olhos dos seus utilizadores.
“Não estou a dizer que ninguém deva entrar em pânico ou sentir-se mal. Simplesmente é bom termos noção de como os nossos dados podem ser usados. Não precisamos estar atentos a tudo: só precisamos de pensar criticamente e aprender mais sobre o potencial que os nossos dados tem a esta escala. Ainda estamos todos a aprender”, escreve O’Neill.
Leia mais: Facebook tem 262 milhões de euros para projetos ligados ao jornalismo
Me 10 years ago: probably would have played along with the profile picture aging meme going around on Facebook and InstagramMe now: ponders how all this data could be mined to train facial recognition algorithms on age progression and age recognition
— Kate O'Neill (@kateo) 12 de janeiro de 2019
Most common rebuttal in my mentions: "That data is already available. Facebook's already got all the profile pictures."Of course. And I'm not trying to say this is a crisis or that it's inherently dangerous. But just for fun, let's play this out.
— Kate O'Neill (@kateo) 13 de janeiro de 2019
Let's just imagine that you wanted to, say, train a facial recognition algorithm on age-related characteristics. You'd ideally want a broad and rigorous data set with lots of people's pictures. It'd help if you knew they were taken a fixed number of years apart — say 10 years.
— Kate O'Neill (@kateo) 13 de janeiro de 2019
Sure, you could mine Facebook for profile pictures and look at posting dates or EXIF data. But that's a lot of noise; it'd help if you had a clean then-and-now. What's more, the photo posting date and even EXIF data wouldn't always be reliable for when the pic was actually taken.
— Kate O'Neill (@kateo) 13 de janeiro de 2019
Why? People could have scanned offline photos. People might have uploaded pictures multiple times over years. Some platforms strip EXIF data for privacy; people's captions are helpfully adding that context back, as well as other context about where and how the pic was taken.
— Kate O'Neill (@kateo) 13 de janeiro de 2019
Thanks to this meme, there's now a very large data set of carefully curated photos of people from ~10 years ago and now. Is it bad that someone could use it to train a facial recognition algorithm? Not necessarily. It could help with finding missing kids, to cite one benign use.
— Kate O'Neill (@kateo) 13 de janeiro de 2019
Like most emerging technology, facial recognition's potential is mostly mundane: age recognition is probably most useful for targeted advertising. But also like most tech, there are chances of fraught consequences: it could someday factor into insurance assessment and healthcare.
— Kate O'Neill (@kateo) 13 de janeiro de 2019
I'm not saying anyone should panic or feel bad. It's simply worth becoming more mindful of how our data can be used. We don't need to be wary of everything; we just need to think critically, and learn more about the potential our data has at scale. We're all still learning.
— Kate O'Neill (@kateo) 13 de janeiro de 2019
