Tumblr has been known for two things: pornography and social justice blogs. After December 17th it will only be known for social justice blogs. The service announced that it was going to commit corporate suicide by remove all pornography from the site. But Tumblr isn’t taking the easy way out. Instead it has opted to prolong its misery, to commit corporate seppuku if you will, by using machine learning to remove pornography from its site:
For some reason, the blogging site hopes that people running porn blogs will continue to use the site after the December 17 ban but restrict their postings to the non-pornographic. As such, the company isn’t just banning or closing blogs that are currently used for porn; instead, it’s analyzing each image and marking those it deems to be pornographic as “explicit.” The display of explicit content will be suppressed, leaving behind a wasteland of effectively empty porn blogs.
This would be bad enough for Tumblr users if it were being done effectively, but naturally, it isn’t. No doubt using the wonderful power of machine learning—a thing companies often do to distance themselves from any responsibility for the actions taken by their algorithms—Tumblr is flagging non-adult content as adult content, and vice versa. Twitter is filling with complaints about the poor job the algorithm is doing.
Machine learning has become the go-to solution for companies that want to make it appear as though they’re “doing something” without taking on the responsibilities. We’re already seeing the benefits of this decision. A lot of non-porn material is being removed by whatever algorithms they’re using and when users complain Tumblr can say, “Don’t blame us! The machine screwed up!” Thus Tumblr absolves itself of responsibility. Of course the three people who post non-pornographic content to Tumblr are likely to flee after tiring of playing Russian roulette with the porn scanning algorithm but I’m fairly certain Verizon, which owns Tumblr now, just wants to shutdown the service without listening to a bunch of people who still use the platform whine.