Anti-Slop Policy
From the Bowstones is proudly written by a human, for the benefit of humans. I do not, and will not, use generative AI technology to produce material for this website.
More specifically:
- My original content will never be AI-generated, nor will generative AI be used anywhere during the creative process.
- Any reference to existing content that I know or suspect to AI-generated will be clearly marked, and made only when that is the only sensible option (most obviously - I sometimes want to criticise AI-generated content!).
- The site is configured, so far as is possible, to prohibit access by the web scrapers that AI companies use to fetch training data, and the source code is hosted on Gitlab, which has committed not to train AI on customer data.
Why do I have this policy?
We are at a point where, with a few button clicks and a short prompt, anyone can produce large quantities of machine-generated text, code, images, audio or even video. I'm no modern Luddite - using technological advancements to come up with new, genuinely better ways of doing things is a critical part of the human experience and I'd not have it any other way. And genAI is certainly an impressive technological advancement; in fact, I could even see a few positive uses for genAI, applied with consideration and restraint in certain situations.
So you might reasonably ask, then, why I am so opposed to it. Well, here are four good reasons, ordered roughly by my view of their significance.
1. It serves giant corporations more than it will ever serve me or you
GenAI isn't being deployed, as I described above, with consideration and restraint, and it's certainly not restricted to the certain situations where good uses exist. Instead, it's being shoehorned in wherever it will fit, bankrolled predominantly by the enormous wealth of big-tech companies: companies that stand to gain huge amounts of money and power by doing so, and care little about the impact on anyone but themselves, even when that impact includes:
- shutting out independent artists from anniversary celebrations in their hometown
- contributing to the spread of misinformation by misrepresenting news reports
- lending credibility to hateful messages spread by extreme right-wing politicians through fabricated statistics and quotations
- providing dangerous advice to people experiencing mental health crises, even instructing or directly encouraging people to take their own lives.
- literally creating pornographic images of children.
2. It undermines the point of the website
I run this website because I like creating the content of this website. It is an outlet for my own thoughts and creativity, and formulating that into something I can publish online and share with the world is a process I enjoy.
What genAI does is try to sell me a magic box that automates and accelerates that process. Why on earth would I want a machine that makes me spend less time on something I actively enjoy? Ideally I want precisely the opposite.
It frames itself as freeing me from a burden, but that framing is completely wrong, because this isn't a burden, and if it ever became so, I could simply choose to step away from it, without the need for any magic boxes.
3. The output is awful anyway
I can write better than AI can.
If that sounds unduly arrogant, you should know that so, in my opinion, can most people. It doesn't matter how you ask AI to write; it will get it wrong. It's difficult to put into words what, precisely, it gets wrong, but it grates to read. I never come across an AI-generated text that I particularly want to be reading; nor a human-written text that I feel AI could have done a better job of.
The same goes for any other medium. Until recently, my local pubs and takeaways would design their own signs, menus, and posters. Some would commission a graphic designer, but often they would cobble something together themselves in Microsoft Publisher or similar. This way of doing things isn't totally dead, but I'm increasingly seeing this sort of design replaced by AI, and I can sincerely say that in every single case that this has happened, without exception, the old homegrown design was in every conceivable way better.
AI posters tend to lack clear space and use lots of bright colours, glossy textures and large or heavily emphasised text, which makes them overwhelming to look at. They all use pretty much the same dreadfully ugly "font", if I dare call it that. They often seem to generate meaningless slogans that sound straight out of a corporate meeting and couldn't be more out-of-place down the local pub. The images are unnatural and unappealing, particularly those of food (yes, at least one of my local pubs has used AI-generated images of their own food!).
But above all, it's soulless. A menu made in Microsoft Publisher won't win any design awards, but it still provides some human connection to the person who made it. Its AI replacement can't possibly do that, because there is no human maker to connect with in the first place, just a very advanced, very powerful computer program that tries to pretend.
4. There is enough AI slop on the internet already
Even if I thought it were a good idea, there would be no point in me writing a blog full of AI-generated content, because I would have nothing to add to that space. Blogs full of AI-generated content are already rife. You might think this contradicts my previous assertions that AI-generated media is overwhelmingly poor-quality, soulless rubbish. If it's really so bad that nobody could possibly want to read it, one could argue, how can so much of it exist?
I think the answer is actually quite simple: content that nobody wants to read, view or listen to is still perfectly capable of serving a purpose for its creator. Some blog owners have always wanted as many clicks and page views as possible so they can make money from banner advertising; some marketing departments have always filled their web pages with sales pitches thinly veiled as helpful tutorials in the hope of convincing search engines that they are slightly more credible than their competitors; some social media influencers have always tried push their follower counts a little higher in ethically dubious ways; some internet trolls have always taken pleasure in simply wasting people's time. This sort of poor-quality content that nobody wants is often described as "clickbait", and it had been part of the furniture on the internet for years before this sudden rush to build generative AI began. Generative AI has simply made the production of clickbait trivially fast and easy.
If you've read all the way to the end of this page, it seems that you enjoy reading what I've written. Might I interest you in my archive of blog posts?