I've seen people trying to break the community guidelines by using inappropriate icons, however I think there should be a system for blocking out the aforementioned (i.e. pictures of faces, 18+ content, e.t.c.) which would make Scratch a more kid-friendly site

How would the filter know when a profile picture was 18+?

Just report the profile pictures.

starhero5697 wrote:

Just report the profile pictures.

Agreed, yet people can just put a new one of the same thing up

Miiraaya wrote:

Agreed, yet people can just put a new one of the same thing up

Then report them again, if they do it repeatedly, they could face a bigger punishment.

TheAnimalKingdom42 wrote:

How would the filter know when a profile picture was 18+?

Yeah. And even then, how accurate would the filter be?
It might censor a completely innocent PFP, while it might allow a completely inappropriate PFP.

TheAnimalKingdom42 wrote:

How would the filter know when a profile picture was 18+?

It would probably know by using a few telltale signs, so the ST can block the icon being used

Miiraaya wrote:

TheAnimalKingdom42 wrote:

How would the filter know when a profile picture was 18+?

It would probably know by using a few telltale signs, so the ST can block the icon being used

What are the telltale signs? How would the filter work?

Miiraaya wrote:

TheAnimalKingdom42 wrote:

How would the filter know when a profile picture was 18+?

It would probably know by using a few telltale signs, so the ST can block the icon being used

What telltale signs could AI determine?
Also, AI can recognize faces, but what if someone wanted to make their profile picture a famous person or a movie character? I've fooled many different face recognition AIs with my art. That would happen in this case, too, resulting in many false positives.

starhero5697 wrote:

Miiraaya wrote:

Agreed, yet people can just put a new one of the same thing up

Then report them again, if they do it repeatedly, they could face a bigger punishment.

Yeah but the ST won't take immediate action against the user with the inappropriate profile picture. If there were multiple reports, the ST will see it at some point.

As for the suggestion, this sounds like a good idea but there are some problems.
1. How would the filter detect profile pictures?
- What I meant was, what would the filter detect in the profile picture?

2. There will be a lot of false positives.

Miiraaya wrote:

(i.e. pictures of faces, 18+ content, e.t.c.)

If it filters pictures of faces, Griffpatch would have to change his profile picture because it has his face in it. In fact, some faces may not be real, but could be detected by the filter.

Miiraaya wrote:

(i.e. pictures of faces, 18+ content, e.t.c.)

If it filters pictures of faces, Griffpatch would have to change his profile picture because it has his face in it. In fact, some faces may not be real, but could be detected by the filter.

When I said pictures of faces, I meant photographs ^^

Miiraaya wrote:

Tunde123 wrote:

Miiraaya wrote:

(i.e. pictures of faces, 18+ content, e.t.c.)

If it filters pictures of faces, Griffpatch would have to change his profile picture because it has his face in it. In fact, some faces may not be real, but could be detected by the filter.

When I said pictures of faces, I meant photographs ^^

Even if it is photographs, some people can change the look of a face using Photoshop.

Nearly any AI can be fooled, and with the 512x512 resolution max of Scratch pfps it's even harder to figure out if an image would be marked inappropriate, and with faces it's even harder to moderate (a drawing of a face would get moderated, an image that slightly resembles a face would get moderated, etc.)

Reporting people would be (and is) much more effective than any AI would be on Scratch.

stuppypanda wrote:

Nearly any AI can be fooled, and with the 512x512 resolution max of Scratch pfps it's even harder to figure out if an image would be marked inappropriate, and with faces it's even harder to moderate (a drawing of a face would get moderated, an image that slightly resembles a face would get moderated, etc.)

Reporting people would be (and is) much more effective than any AI would be on Scratch.

That's correct.

Imagine, for example, someone set their PFP to the famous picture of the crater that looks like a face:

Because the AI for the filter isn't smart enough to recognise that that's not a face, but rather a crater that looks like a face, the filter would remove that PFP, even though it isn't revealing anyone's face.

TheAnimalKingdom42 wrote:

https://u.cubeupload.com/bebegriggs_/why.png

What's the URL for this??

RED-001-alt wrote:

TheAnimalKingdom42 wrote:

https://u.cubeupload.com/bebegriggs_/why.png

What's the URL for this??

Please don't go off-topic. That's in my signature. I don't know if Quora links are allowed but you can look it up.

Anyways, OP, you said this would detect faces, as well as other “telltale signs”. What “telltale signs” are you referring to? How would the AI recognize it?

its real…
Anyway, no support, it would be too complicated, and if the filter bot doesn't even work half the time on Scratch then there for sure isn't gonna be any face detection or any kind of image detection at all.

If this were to be implemented, people would try to find out how it works and when they do, try to workaround it.

TheAnimalKingdom42 wrote:

How would the filter know when a profile picture was 18+?

I'd imagine something like this. Your text to link here…

Miiraaya wrote:

I've seen people trying to break the community guidelines by using inappropriate icons, however I think there should be a system for blocking out the aforementioned (i.e. pictures of faces, 18+ content, e.t.c.) which would make Scratch a more kid-friendly site

Scratch could use an AI that could get smarter by each pfp that gets into scratch but idk