Lots of discussion about AI bringing about the end of the world at the moment, so I’m running in the other direction (not because I don’t think AI is an existential threat, but because the noise to signal ratio is a tad too high right now). My other direction is a simple question: what damage are we causing ourselves, if any, by allowing AI to shape our messaging — researching, writing, content-creating that carries our name, as individuals, as companies, as institutions?
This is in some ways a more existential question than the existential one. Because we can do our credibility tremendous damage if we are perceived to be outsourcing any part of the creation of content — an email, a blog post, an ad — to artificial intelligence. More on this to come.
First off, there are several cognitive biases we have when it comes to consuming AI, and when we use AI to create something ourselves. This is because we have different antennae for each process: if we’re reading or watching or hearing content, we are increasingly skeptical of the provenance of everything. We tend to forget that we’ve been at this whole GPT thing for more than three years now, so a pattern has set in. AI is definitely getting better, but so are we at detecting it — or at least suspecting it, which is tantamount to the same thing.
If we think something might be AI, we’ll not look at/read/consume it in the same way. It becomes slop in our minds, even if it’s not. And when it feels like slop, we run a mile.
When we use AI to create something, our antennae are still a few steps behind. We feel that we didn’t really use AI to create something, so much as edit, tweak something, provide a second opinion on something, help research something. We feel we are still in the driving seat, and so AI’s role is limited and therefore “this being done my name is not slop.” We overlook its resemblance to AI, and indeed, we might quickly forget which bits we did and which bits were AI.
This is the fundamental problem — the credibility gap between how we perceive our own work and the work of others.
It’s a sort of bias blind spot, an asymmetry of authorship, but there’s something more at play here. Once we establish ownership over something then that ‘thing’ remains ours, whatever else is added to it by others, human or not. Call it the Every Breath You Take Effect. Sting’s song would be fairly forgettable — indeed it was initially considered a reject from The Police’s final album Synchronicity (see David Harley’s excellent The world’s most played song shouldn’t be this simple). It’s the deceptively simple guitar part, composed by Andy Summers, with the inclusion of a 9th chord, which lifts the song into the heavens (and saved it from the trash). But the song is credited to Sting, and Sting alone.
We’re all Sting, in a way. If something originates with us, even if it’s just the concept, or in AI speak the prompt, then it’s ours. If we see something in it that feels unlike us, we mostly call it growth and a sign of our burgeoning talent. We overlook the warning signs of slop ahead.
This is the only explanation I can give for why I see so many good writers, X tweeters, YouTube creators, who allow AI to creep into their work. I see it in companies’ material too, including op-ed pieces, blog posts, press releases, everything. I’m not going to name them here, because I believe the stakes are too high.
The stakes are high because once you sense AI, you can’t unsense it. You can’t reason that, well, that bit might be AI, but the rest of it isn’t. You can’t afford that luxury. Something that is not cut from whole cloth is suspect, not because it is entirely fabricated, but because you can’t tell which bit is and which bit isn’t whole cloth.
And, perhaps more important, you can’t trust the judgement of the person creating it. If the argument they present, as evidence of their wisdom and subject-mastery, is suspected to have involved AI, then not only is that argument suspect, but the reader/listener/viewer must question the creator’s mastery of the topic as a whole to be suspect. There’s just no point in doing anything otherwise. If they knew their topic well, they wouldn’t need AI.
And that’s the point. The point of content that wants to be taken seriously. It’s supposed to be evidence of experience, knowledge, thinking, creativity, uniqueness. It’s supposed to be us.