I have zero faith in automated moderation. Machine learning is in its infancy and nearly every auomatic system has led to inconsistent and unfair enforcement of a platform's guidelines. Google...
I have zero faith in automated moderation. Machine learning is in its infancy and nearly every auomatic system has led to inconsistent and unfair enforcement of a platform's guidelines.
Google would gain a lot more respect if they actually hired a customer support and moderation team.
I'm not expecting Google to sift through every single video that is uploaded to YouTube. It would be a gargantuan task to moderate the platform to this degree. Copyright detection was probably one...
I'm not expecting Google to sift through every single video that is uploaded to YouTube. It would be a gargantuan task to moderate the platform to this degree. Copyright detection was probably one of the few things YouTube got right in terms of automation (initially, before this Content ID shit came out.)
What I am expecting is a team to moderate and swiftly remove spam comments, which are all over YouTube now.
It would take 720,000 hours to watch every second of new video that's uploaded on YouTube, but likely a small fraction of that time to read through reported comments. We are talking about text here, which takes mere seconds to read, and where humans are better than any current machine learning algorithm at recognising patterns.
Allow me to be thoroughly optimistic here, though not necessarily about the state of ML. Google already has a team for such things. They have to. That team is likely completely overworked with the...
Allow me to be thoroughly optimistic here, though not necessarily about the state of ML.
Google already has a team for such things. They have to. That team is likely completely overworked with the amount of crap they have to deal with. It's not a secret that customer support for millions of viewers and thousands of active channels is not trivial. Even keeping spam to a minimum is probably a team all of its own. Alright, here's the optimism: That team stays in place, but instead of manually moderating the worst offenders, those cases are dealt with automatically now. The team then has time to actually do some real moderation, instead of just acting as a fire brigade.
Also, the powers they give the bots is relatively small. Being muted for 24h isn't the end of the world, for a viewer/commenter.
It's only tangentially related to this article (in that Google will spectacularly fail regular users with its endless automation), but I might as well post about it somewhere since it gave me a...
It's only tangentially related to this article (in that Google will spectacularly fail regular users with its endless automation), but I might as well post about it somewhere since it gave me a terrific headache for a day.
Last week I got an email from the Chrome Web Store about the Fangs listing noting that the screenshot had illegible text and the extension would have "its prominence reduced".
This is the image they included in the email, note its resolution: 512x320. Here is what uploading images looks like on the publishing side, note the listed resolutions: 1280x800 or 640x400. Need I say more?
I'm very respectful of extension developers, but I admittedly do find the text hard to read. I can't tell what's being shown in the screenshot. It looks like a form, or maybe an information...
I'm very respectful of extension developers, but I admittedly do find the text hard to read. I can't tell what's being shown in the screenshot. It looks like a form, or maybe an information display. I think that even if less content were shown overall, a higher font-size would give me a better idea of what the extension is for.
Don't get me wrong, the text is illegible. The real problem is that I couldn't have uploaded that screenshot in the first place. Google must have done some compression and scaling to make it that...
Don't get me wrong, the text is illegible. The real problem is that I couldn't have uploaded that screenshot in the first place. Google must have done some compression and scaling to make it that way. I know for a fact that I uploaded a 1280x800 image. They fucked it up and are punishing me for it, essentially.
If only there was some kind of way of designing things so the website responds to the display. I guess Google has yet to figure it out.
They did a bad job of explaining it, but I think the problem isn’t just the compression, but rather that your screenshot doesn’t look very good in their catalog. They made a choice to go with a...
They did a bad job of explaining it, but I think the problem isn’t just the compression, but rather that your screenshot doesn’t look very good in their catalog. They made a choice to go with a fixed-size image, but probably didn’t show you how your image would be used.
When uploading an avatar to a social media site, there is usually a preview showing your image in context, but in developer tools, nice UI touches like that are often skipped. So they sent you an email a lot later.
I don’t think design feedback like that should be considered punishment even though it was done clumsily.
Ah, I see what you mean. In the original image, I'd bet the text was perfectly legible. That's definitely something they should've spelled out during the original upload process then, and not...
Ah, I see what you mean. In the original image, I'd bet the text was perfectly legible. That's definitely something they should've spelled out during the original upload process then, and not (presumably) months later like this.
A little update to this, after contacting support and telling them it's their own mistake, they have reverted their action. Has anything changed to prevent this silliness from happening again? Who...
A little update to this, after contacting support and telling them it's their own mistake, they have reverted their action. Has anything changed to prevent this silliness from happening again? Who knows.
I have zero faith in automated moderation. Machine learning is in its infancy and nearly every auomatic system has led to inconsistent and unfair enforcement of a platform's guidelines.
Google would gain a lot more respect if they actually hired a customer support and moderation team.
It's basically impossible. 720,000 hours, or 30,000 days worth of video are uploaded to youtube EVERY DAY. It's ML moderation or no moderation.
Supporting creators and communities with better moderation tools would make a difference.
I'm not expecting Google to sift through every single video that is uploaded to YouTube. It would be a gargantuan task to moderate the platform to this degree. Copyright detection was probably one of the few things YouTube got right in terms of automation (initially, before this Content ID shit came out.)
What I am expecting is a team to moderate and swiftly remove spam comments, which are all over YouTube now.
It would take 720,000 hours to watch every second of new video that's uploaded on YouTube, but likely a small fraction of that time to read through reported comments. We are talking about text here, which takes mere seconds to read, and where humans are better than any current machine learning algorithm at recognising patterns.
Allow me to be thoroughly optimistic here, though not necessarily about the state of ML.
Google already has a team for such things. They have to. That team is likely completely overworked with the amount of crap they have to deal with. It's not a secret that customer support for millions of viewers and thousands of active channels is not trivial. Even keeping spam to a minimum is probably a team all of its own. Alright, here's the optimism: That team stays in place, but instead of manually moderating the worst offenders, those cases are dealt with automatically now. The team then has time to actually do some real moderation, instead of just acting as a fire brigade.
Also, the powers they give the bots is relatively small. Being muted for 24h isn't the end of the world, for a viewer/commenter.
It's only tangentially related to this article (in that Google will spectacularly fail regular users with its endless automation), but I might as well post about it somewhere since it gave me a terrific headache for a day.
Last week I got an email from the Chrome Web Store about the Fangs listing noting that the screenshot had illegible text and the extension would have "its prominence reduced".
This is the image they included in the email, note its resolution: 512x320. Here is what uploading images looks like on the publishing side, note the listed resolutions: 1280x800 or 640x400. Need I say more?
I'm very respectful of extension developers, but I admittedly do find the text hard to read. I can't tell what's being shown in the screenshot. It looks like a form, or maybe an information display. I think that even if less content were shown overall, a higher font-size would give me a better idea of what the extension is for.
Don't get me wrong, the text is illegible. The real problem is that I couldn't have uploaded that screenshot in the first place. Google must have done some compression and scaling to make it that way. I know for a fact that I uploaded a 1280x800 image. They fucked it up and are punishing me for it, essentially.
If only there was some kind of way of designing things so the website responds to the display. I guess Google has yet to figure it out.
They did a bad job of explaining it, but I think the problem isn’t just the compression, but rather that your screenshot doesn’t look very good in their catalog. They made a choice to go with a fixed-size image, but probably didn’t show you how your image would be used.
When uploading an avatar to a social media site, there is usually a preview showing your image in context, but in developer tools, nice UI touches like that are often skipped. So they sent you an email a lot later.
I don’t think design feedback like that should be considered punishment even though it was done clumsily.
Ah, I see what you mean. In the original image, I'd bet the text was perfectly legible. That's definitely something they should've spelled out during the original upload process then, and not (presumably) months later like this.
A little update to this, after contacting support and telling them it's their own mistake, they have reverted their action. Has anything changed to prevent this silliness from happening again? Who knows.