I ALWAYS fake my age in age verification systems. If even by a day sometimes. Also give the wrong phone number when I register places. No company needs my data. I’m really against age verification systems. Most of it is just parents who can’t be bothered to help their children.
As someone who worked in machine learning for about a decade, I would absolutely *never* trust tech like this. It will work "fine" for maybe 95–98% of people (whose backgrounds are represented in the training dataset, and whose skin tone allows for sufficient contrast in the lighting, which are both major limitations). It will then will miscategorize the remaining people (or fail to recognize their faces as faces entirely)—but the fact that it won't work for such a small percentage of folks and would lock them out of their games or whatever. This will mean that companies will have little incentive to improve the technology for folks like Patrick who look perpetually sixteen or whatever. (lol I'm right there with you buddy, that's one of the reasons I grew a massive beard)
Add to that the privacy concerns (which remain, in my mind, despite the ESRB's assurances) and the fact that once the tech is deployed, companies will have increased incentive to expand its use-case. ("Sure, we said we wouldn't use it to age-gate games, but now that everyone's used to it...") Just a no-go for me.
But like a lot of former software devs I'm leaning hard into Luddism as I age. My reflex take on stuff like this is always cynical. 🤷
I don’t know if I would trust an automated system to accurately guess whether or not someone is a child based on their face. Especially anyone who isn’t white - there’s plenty of research showing how bad existing systems are at recognizing people of color!
I ALWAYS fake my age in age verification systems. If even by a day sometimes. Also give the wrong phone number when I register places. No company needs my data. I’m really against age verification systems. Most of it is just parents who can’t be bothered to help their children.
The system that exists right now—child accounts managed by adult accounts—seems to work pretty well, imo, absent more layers of control.
As someone who worked in machine learning for about a decade, I would absolutely *never* trust tech like this. It will work "fine" for maybe 95–98% of people (whose backgrounds are represented in the training dataset, and whose skin tone allows for sufficient contrast in the lighting, which are both major limitations). It will then will miscategorize the remaining people (or fail to recognize their faces as faces entirely)—but the fact that it won't work for such a small percentage of folks and would lock them out of their games or whatever. This will mean that companies will have little incentive to improve the technology for folks like Patrick who look perpetually sixteen or whatever. (lol I'm right there with you buddy, that's one of the reasons I grew a massive beard)
Add to that the privacy concerns (which remain, in my mind, despite the ESRB's assurances) and the fact that once the tech is deployed, companies will have increased incentive to expand its use-case. ("Sure, we said we wouldn't use it to age-gate games, but now that everyone's used to it...") Just a no-go for me.
But like a lot of former software devs I'm leaning hard into Luddism as I age. My reflex take on stuff like this is always cynical. 🤷
I wish I, too, could grow a massive beard. I just look...off. LOL.
But yes, I agree with you here and suspect it's why it'll keep getting (rightful) pushback.
I don’t know if I would trust an automated system to accurately guess whether or not someone is a child based on their face. Especially anyone who isn’t white - there’s plenty of research showing how bad existing systems are at recognizing people of color!
Technically speaking, I think it could probably work "good enough." Whether it's necessary is a whole other question entirely.