YouTube Thinks I’m a Child | Hacker News Reader

YouTube Thinks I’m a Child

youtube.com

21 pointstomcam9 comments

I watched all 30 minutes of that, and it was just painful to live vicariously through the blatant ignorance and impotence of Google Support who kept copy-pasting the same answers while ignoring the poor lady's pleas, questions, and assertions. I get the same feels from Google Support all the time, and most other Tech Support desks. I get it from my employer's Tech Support. Interestingly, though we depend on Google Workspaces as mission-critical for our day-to-day operations, we don't even have recourse to Google Support for it. At all. If I discover some sort of bug or issue at work, I have to reproduce it in my personal account and try my luck at Google One support.

She does have a really salient point, in that if anyone could prove racial or gender bias of some kind that is hurting people monetarily, Google could have a real problem. She didn't say the words "class action lawsuit" but that's what I'm thinking.

You can find great Google Workspace admins on Upwork.com

Saved my bacon a few times.

Are you telling me to hire a Workspaces admin? For what reason?

My employer has their own Workspaces administrators. They wouldn't allow a stranger, non-employee to put hands on the dashboard like that.

Are you telling me to hire a Workspaces admin for my personal account? I don't have a business-class Workspaces account; there is nothing to administer.

A freelance admin won't help me if I find a bug that I need to report to Google. I've found a couple bona fide bugs that required me to work with One Support for an extended period before Tier-1 escalated to an internal team and they finally all agreed that it wasn't my fault and I'd better just report it as a bug.

I already pay Google One for support; hiring someone on UpWork seems like an outrageous expense for run-of-the-mill support questions.

So I'm curious what kind of AI or algorithm could reliably detect "minors" and that laundry list of scenarios she listed that could indicate abuse or exploitation.

I mean, she's got a very young look and so visually, could pass for a teenager, but listen to her voice and she's obviously mature.

Any sort of automated "minor detection" has got to have a wide margin for error, or it will simply be sweeping up every 18-20 year old in sight, and mostly women, yes.

Face recognition technology is really great these days. The fact that only a portion of her videos get flagged seems to indicate that Youtube is not employing facial recognition when it searches for minors; it just looks at every single video in a vacuum.

My guess is that the shorts don't provide enough data for the algorithm to make an accurate assessment, whereas her longer videos probably show more of her face, voice, etc. If the algorithm assessed her videos collectively (or at least the most recent ones), it would probably not return all these false positives.

Recognition is one thing, but age assessment is another (and determining those abstract situations still more unlikely).

If humans can misjudge a face by five years or more, what sort of algorithm can reliably differentiate a 17.5-year-old from a 19-year-old, and have such a small error margin that Creator Support doesn't even know what she's on about?

If their system is only going by her face, then they're doing a disservice, because a person's age is not merely in their face.

Is this 100% AI-driven, or are there humans in the loop as well, reviewing flagged content?

Yes, it would not be perfect. I think it would help in her case because it does identify her as an adult in most of her videos.

However, if you consistently get flagged as a minor, the algorithm will need to take into account more data than just your physical appearance. For example, if it could recognize your appearance and connect it to your profile (including your verified age), then it could just ignore you and only guess at the ages of people in your video who don't have a profile with a verified age.

Admittedly, this does get into creepy privacy territory, but if it were an opt-in system, I think YouTubers who are having this issue would really appreciate it.

Youtube keeps putting community notes on the videos I watch, often citing wikipedia as the source of truth. Despite the video being from a Nobel prize winner on the subject.

I used to see these, but I haven't seen them in a long time. I think they're covered by one of my uBlock lists (probably "annoyances" or something like that).

Reply on news.ycombinator.com