4 ms·

I still kinda fail to see how this would work in practice. For instance I'd set it up so that a child could see full nudity (let's say greek art or primitive wa

I still kinda fail to see how this would work in practice. For instance I'd set it up so that a child could see full nudity (let's say greek art or primitive warriors are fine) but no sex (no masturbation, no intercourse) A site would then have to tag each page with the relevant metadata, including user generated contents, and decide if any specific picture is nudity but not sex. But then my definition of what is sex will probably not match the sites'. Someone fully clothed rubbing it against a table should be banned for me but would probably pass withtout a tag, etc. Basically, the very concept of having tags to accurately convey the nuances of what's happening in a page or media is complicated, and have that happen at the internet level looks to me like an unsolvable problem. Some would say "better something than nothing", but I think we quickly fall into problematic cases when muslim libraries start to rely on it to block profane pictures for instance, where a decently maintained white list would work better.

flangola7 3y ago
>Basically, the very concept of having tags to accurately convey the nuances of what's happening in a page or media is complicated, and have that happen at the internet level looks to me like an unsolvable problem. SOTA computer vision systems are capable of handling this automatically. I've implemented an augmented GPT-4 solution that categorizes nude material into several categories and it outperforms the average human tagger.
Kwpolska 3y ago
No ratings system can be perfect. But with precisely defined categories, it could work. User-generated content could be specifically tagged as such, with the parent being able to choose between “allow all”, “allow if has rating and it matches the standard allowed levels”, or “block all”.
- makeitdouble 3y ago
  Yes, this is a popular approach. To me the intractable part is we still rely on either - A: content creators to properly/genuinely tag their content - B: a central entity to tag each content To clear the simple case, the B route is basically akin to the white listing efforts we have now, and going PICS or not doesn't look to me bring much to the table. If anything, PICS needs more management on the individual side than the current MDM approach where for instance, the school admins have a whitelist of sites available from within the school network. Going the A route, we still hit every moderation issue ever met by a user generated content system has. The question becomes, is the PICS system better than a bespoke filtering system ? I'm not sure, because: Depending on the site I'll probably want different options (I might want exceptions to the "no sex" filtering on medical discussion sites for instance). The site would be offering very tailored and specific options that would be set to reasonable values by default; site specific apparently could be implemented in PICS, but then I'd have to manage them in the PICS options of my browser, and I might have no idea what half of them even mean, when the site would have in context information on exactly what these options do. As funny as it is, having the filtering options bound to accounts instead of the browser could yield better results. I could have "personas" depending on what I intend to look for, which would be a pain with browsers (especially site by site). All in all, I have the feeling it's complicated either way, and PICS trying to be clever and simple brings different pitfalls by being the intermediary.

flangola7 3y ago

>Basically, the very concept of having tags to accurately convey the nuances of what's happening in a page or media is complicated, and have that happen at the internet level looks to me like an unsolvable problem. SOTA computer vision systems are capable of handling this automatically. I've implemented an augmented GPT-4 solution that categorizes nude material into several categories and it outperforms the average human tagger.

Kwpolska 3y ago

No ratings system can be perfect. But with precisely defined categories, it could work. User-generated content could be specifically tagged as such, with the parent being able to choose between “allow all”, “allow if has rating and it matches the standard allowed levels”, or “block all”.

makeitdouble 3y ago
Yes, this is a popular approach. To me the intractable part is we still rely on either - A: content creators to properly/genuinely tag their content - B: a central entity to tag each content To clear the simple case, the B route is basically akin to the white listing efforts we have now, and going PICS or not doesn't look to me bring much to the table. If anything, PICS needs more management on the individual side than the current MDM approach where for instance, the school admins have a whitelist of sites available from within the school network. Going the A route, we still hit every moderation issue ever met by a user generated content system has. The question becomes, is the PICS system better than a bespoke filtering system ? I'm not sure, because: Depending on the site I'll probably want different options (I might want exceptions to the "no sex" filtering on medical discussion sites for instance). The site would be offering very tailored and specific options that would be set to reasonable values by default; site specific apparently could be implemented in PICS, but then I'd have to manage them in the PICS options of my browser, and I might have no idea what half of them even mean, when the site would have in context information on exactly what these options do. As funny as it is, having the filtering options bound to accounts instead of the browser could yield better results. I could have "personas" depending on what I intend to look for, which would be a pain with browsers (especially site by site). All in all, I have the feeling it's complicated either way, and PICS trying to be clever and simple brings different pitfalls by being the intermediary.