Four letters and a semicolon is all it took for an X user called ‘seclilc’ to strike a nerve back in August with a post that read, in its entirety, “AI;DR (AI; didn’t read)”. It racked up more than 16,000 likes. Could it be the new TL;DR (too long; didn’t read)?
Rick Manelius, a start-up executive who describes himself as “about as pro-AI as you can be”, liked this so much he’s adopted it as a policy, explaining why in a recent Substack post. His complaint will be familiar to anyone who has asked a question or opinion of a colleague and is treated to a wall of unchecked and unedited AI slop in return. “If you’re not bothered enough to review and edit it… then I’m not going to bother reading it,” he sums up.
I wonder if this is an even bigger problem here in Ireland? Indeed’s Workforce Insights Report found that 70% of Irish workers use AI regularly, far ahead of the US (43%) and the UK (41%). We are world leaders. Whether that’s in productivity or in pasting remains to be seen. But it must be put in context: generative AI is baked into or tacked onto so many workplace productivity apps that it’s hard to avoid it these day.
Back to the slop: the real insult of the unedited text dump isn’t just laziness, it’s missing the point of workplace conversation: collaboration, spitballing, drawing on a different perspective. As Manelius puts it: “I can ask Claude directly if I wanted to.” A tongue-in-cheek website doing the rounds, dontpastetheai.com, makes the same point: the person asking has the same tools you do. They wanted your context, your judgement, your take. Pasting the chatbot’s answer outsources the one thing they came to you for.
There’s a practical cost too. Researchers at BetterUp Labs and Stanford’s Social Media Lab, who coined the term ‘workslop’ for AI-generated work that looks polished but doesn’t actually do anything useful, found workers spend nearly two hours dealing with each piece they receive, about 20 minutes more than if the sender hadn’t used AI at all.
But before we start flinging AI;DRs at each other like confetti, a word of caution.
For some people, AI writing tools aren’t a shortcut; they’re a ramp. Someone writing in their second or third language or someone with dyslexia may use them to get what they already know down on the page. Google researchers were building AI e-mail writing aids for adults with dyslexia back in 2022 – though their own study found the models of the day weren’t yet accurate enough to meet those writers’ needs.
Detection issues
The bigger problem is that we’re not actually very good at spotting AI generated content. A 2023 Stanford study found seven popular AI detectors flagged 61% of essays written by non-native English speakers as machine-written, while being near-perfect on essays by US-born students. Detectors tend to treat simpler, more predictable language as suspicious, which is exactly how many people write in a language they’re still learning.
Humans aren’t much better. Exhibit A: the em dash. Because chatbots use it liberally, the humble long dash has been branded a giveaway and some writers have started avoiding it altogether. Never mind the fact that Emily Dickinson was scattering them across the page more than a century before ChatGPT. The machines learned it from us!
In the interests of transparency, I used Claude to help research this column. It found the studies; I read them. I gave it background information and it summarised it for me. I then rewrote the summarised facts in my voice after double checking it. That, really, is the whole point. Some people use AI as tools for thinking, some people are tools and use AI instead of thinking. For the latter, feel free to fling AI;DR at them.
Ltd
