Slop is not a new problem. It is a scaled problem that we had before AI too. I’m convinced that the problem is our failure to define what good is. We’re sloppy enough with our definitions that we can’t really say what anti-slop is, and that’s why we have so much of what is now popularly called “AI slop”.
As I shared in my recent essay and Signals & Subscriptions episode, way too much about what Good is gets chocked up to taste, when it isn’t taste at all. Good is something we can decide, we can define, and we can measure. But Good is not an input metric, where it originated doesn’t actually count for much most of the time. So the question isn’t “was AI involved?” it’s “is this thing any good?” Because if it’s good as an output, generally nobody slows down enough to care about where it came from. It’s good, it’s useful, we can trust it, and we move on.
What makes something good? Now THAT is a conversation worth having! I’d welcome more of that. Let’s focus on making more of what we do want, rather than less of what we don’t. There’s been a lot of discussion in the last few weeks around this, but I’ve yet to find anyone who wants to have this discussion. It seems to be more of an Unfounded Opinionfest. We can do better than that.
While there is room for many opinions, this I know: Authorship is clearly not the answer. I can use the perfectly aligned author process and still crank out bad work that wastes everyone’s time. Likewise, I can use a well-tuned AI workflow to crank out solid work at incredible speed. Note that I’m not saying most people do this today, I’m saying it is absolutely possible to do. I know because I’ve done it, and I’m not the only one.
Ultimately, what usually happens next in these kinds of situations is the cat-and-mouse escalation. I would expect that in the near future many people end up using AI to filter and mediate the communications they receive, to counter those that they send. But this is not a good answer by any measure. That approach simply won’t work at AI speeds.
I’ve been trying for over a decade to quantify what are generally regarded as qualitative measures. In my 2017 interview of Douglas Hubbard for the Doable Change podcast, we covered this in detail. It’s very possible more often than people like to admit. Again, I know because I’ve done it, and I’m not the only one.
The next time you’re tempted to sort out what is from AI from what is human, remember that’s not really what you want to know. What you’re looking for is whether something is worth your time, or wasting your time. Get specific on what that means to you, and you escape the cat and mouse of AI detection.
And I’ll bet you’ll actually get more of what you want. Ask me how I know.