Who knew that when AI started to assert itself, it would turn out to be a prude? It is a Handmaid’s Tale level prude. It is so prudish, it is impossible to predict what might have it blushing and wafting its virtual face with a CGI fan.
I seem to have developed the knack of being able to offend AI without even trying.
The latest example came when trying to update my author profile on Amazon. When I hit the save button, this message flashed up: ‘There is language within your submission that is in violation of our Content Guidelines. Please remove this content to create your submission.’
Admittedly, the text did reference murder, animal cruelty, and domestic abuse. But as this was part of a synopsis for a novel, I’d have thought there wouldn’t have been a problem with mentioning hard-hitting topics. I mean, how can an author describe what their novel is about if they have to stick to language befitting a Disney children’s movie? An ‘outraged’ message was whizzed off. Turns out it wasn’t anything about domestic violence or kicking dogs that had upset the bot, it was the mention of a website address, so I can’t really point the finger of prudishness in that direction.
Google, on the other hand, is unambiguously guilty.
Google AdSense publisher policy violation reports regularly pop into my inbox, informing me of violations on my websites. And what are these violations? Sexual content apparently.
Sexual content on travel websites. Maybe there’s a niche market there. Maybe it’s already occupied. But not by me, as far as I know.
Anyway, thankfully Google’s policy centre has more details than the rather vague emails they send, so I’m able to see what pages have offended Google with their ‘sexual content.’
One was a post about travel and toilets. It may not be in the best possible taste – it’s not easy to write a post about toilets and travel that is – but only a deviant (and maybe some bots are, we don’t know yet) could consider content that includes ‘Only then did I realise there was no paper, no means for washing… nothing except the hole in the ground. All I had in the pockets of my snow-white shorts was a wad of rupees. It was the most expensive toilet visit I’ve had to this day…’ as sexual.
Another web page reprimanded for sexual content contained an article about ‘Ten Alternative Things to see in Lisbon.’ The sexual bit being a brief snippet about a self-proclaimed ‘sexiest toilet in the world.’ That reference seemed enough to have Google bots reaching for the AI equivalent of heart pills.
The third page that caused offense wasn’t because of anything to do with the sexualising of conveniences, it was because the content mentioned tobacco. Unsurprising really as it was an article about the women who still hand-roll cigars on the Canary Island of La Palma; thankfully not on their thighs as that would have no doubt earned a ‘sexual content’ violation as well.
So, what do I do with these warnings? Not a thing. All they mean is that some companies might not be happy with their Google ads appearing on such pages, and I don’t really care about that.
The more worrying problem is that of indirect censorship increasingly creeping in. These violations are clearly nonsense. But if Google bots are attempting to censor such trivial posts as these, what else is being quietly squashed? Thin edge of the wedge and all that. It’s ironic given the amount of truly offensive material out there.
Those AI bots have got to get a hell of a lot smarter before they’re in a position to intelligently sift through the messy and complicated sea of human-produced jetsam and flotsam on the internet.