Porn-detecting bots might not be able to tell the difference between grot and art

Content filters are often the first, last and only line of defence between innocent users and copping an eyeful of sexually explicit images in the last place you expect it.

Sometimes you can really do without an unexpected glimpse at some porn, and so many sites are using algorithms to stamp out grot before you see it. Unfortunately, it seems some of these porn-blocking bots aren’t working as well as you would hope, and they’re bravely saving several users from high art, as well as the adult stuff.

What’s to blame here? Machine learning.

Explained by Panda Security, they have explained the best way to teach a porn-blocking bot is to show it a whole bunch of pornographic images. This is similar to how anti-virus and anti-malware system works, and similar to those tools, there are false positives.

Related: Best PC Games

Thing is, there is a big difference between a photo of someone in lingerie (that would be bad) and someone in a bikini at the beach (generally ok) and don’t even get started on how similar something like breastfeeding (again, ok) can appear to a bunch of things that are definitely bad.

We’ve seen this happen a few times recently. Panda Security mentions Tumblr’s contentious porn filter that activated in December, which has been flagging up a whole bunch of innocent images as sexually explicit.

Related: Samsung Galaxy S10

“Ultimately, intelligent systems built using machine learning are flawed because the people programming them are unable to specify exactly where the line between ‘ok’ and ‘porn’ is crossed,” claims the Panda Security blog post. “The algorithms may be able to block 99.9% of questionable content, but the 0.1% that remains will always be an issue.”

“These nuances are irrelevant to anti-malware detection because a file can only be one of two states: ‘virus’ or ‘not virus’. Images on the other hand have three states: ‘porn’, ‘not porn’ and ‘maybe porn’. ‘Maybe porn’ is where machine learning can (and does) fail. It is also where the majority of investment in automated systems will take place in the coming years.”