
False detections?
I'm betting that false positive and false negative rates are above 50%
Having laid bare over half a billion usernames and passwords through meager funding and witless indifference, Yahoo! is putting its faith in artificial intelligence to protect people from bare skin. Yahoo! engineers Jay Mahadeokar and Gerry Pesavento in a blog post on Friday said the company has released an open-source model …
private static Pattern NSFW_MATCHER = Pattern.compile(".*\\.yahoo\\..*", Pattern.CASE_INSENSITIVE);
public static boolean isNsfw (String url) {
return NSFW_MATCHER.matcher(url).matches();
}
It could probably use some more training data but it's working fairly well.
When I worked at a university in the mid/late 90s, a couple of the professors were working on image processing to determine nude pictures. Then there's the question of what is "NSFW", which might not need to be fully nude. Even wearing clothes that are see thru or too tight or a suggestive pose could be NSFW, while a naked person in the right setting might be OK. It is a judgment call even for humans, an algorithm will never master it.
The finest of legal minds gave us this from the US Supreme Court:
I shall not today attempt further to define the kinds of material I understand to be embraced within that shorthand description ["hard-core pornography"], and perhaps I could never succeed in intelligibly doing so. But I know it when I see it...
and that's not for want of trying. So the idea that a bunch of image filters will usefully judge whether Michael Phelps poised on the starting block is more or less unsafe than something plainly salacious but better clad is, shall we say, arousingly provocative...
But as so often it will come down legal tickboxing: run our special script and then bask in the security of having adopted "industry best practice", regardless of actual outcomes.
"Then there's the question of what is "NSFW","
As elastic as there's work. For the algorithm to work it would need to know about the company HR policies of the viewers and the status of the viewers themselves.
One place I worked, an employee was viewing porn. It turned out that he was having to stay late because the collection lorry was frequently late, but being staff, he wasn't being paid overtime. And there was no official HR policy.
We showed him how to delete the browser cache in future.
Then there's the question of what is "NSFW",
Reminds me of the days when I used to stand in for an IT director in a SME.
We acquired a new HR director from one well known (for their reverse Midas touch) UK Telecoms operator. On her second day in the job she understood that I am not running any "NSFW" filters on the web - only AV and malware detection. She came to me to insist to deploy them, because "you know, some people may be looking at stuff".
I told her 1. NO, 2. Enumerate the "Some People". 5 minutes later she came with the CEO in tow.
At that point I took them into the kitchenette, opened the copy of the Sun left in the middle of the breakfast table on page 3 and explained her that _THIS_ _WILL_ _BE_ the golden standard for _ALLOWED_ (this was in the days when there was still page 3) unless she bans it, puts the order which officially bans Sun on the premises in writing and staples it on the kitchenette notice board.
She had the color of a plum tomato towards the end of the discussion, with half of the company laughing their ass off (as they were present and watching).
Let's take this off the computer screen and see how much sense it makes:
"We're going to post a guard at the entrance of the building who will check everybody's person and carried items to ensure that NSFW items do not enter the premises."
If it's senseless in meatspace, then it's senseless on the computer screen.
Ffs.
>How is a company mandating what their computers can and cannot be used for censorship?
BYOD, 4G.
Hmmm, can someone still be sued?
Probably.
The issue isn't really censorship, its "bother, the IT industry has nothing left to sell so we gotta make some snake oil and find a way to flog it." Selling people things they want to in is probably fairly easy. Hence, "A technical solution for X" is an easy sell.
It's a tool. From a technical perspective, there's not much difference other than scale between the corporate filter to keep the employees from facebook and the national firewall to keep the people from reading news about how corrupt and oppressive their government is. I'm sure Saudi Arabia, Iran, Pakistan and (what's left of) Syria are looking at how to plug this technology into their filters to make sure all those corrupting western women with low-cut tops. They might need to retrain it to local standards.
> a classifier for NSFW detection, and provide feedback to us on ways to improve the classifier."
So essentially Yahoo are building a search engine for porn?
I'm surprised it took someone all this time to get a round to that.
All it needs now is a snappy name.
Why is looking at porn at work still a thing? Back in ancient times, porn seekers would use the company Internet connection since "the net" was either unavailable at home or was only available on dial-up. Today we've got 3G, 4G, and LTE on our mobile devices - which are often faster than the company network because of all the damn filtering.
int main(enter the void)
...