Enslopification

And how thoughtless creation is making the internet unusable

“Slop” is frequently thrown around in AI circles. People use it broadly and multimodally to describe text and images (and soon audio) on the basis that they are, at once, generic, low-quality, garish, tasteless, and effusive.

Enslopification is how a large percentage of the internet is becoming AI-generated and mean-revertive, both in form and function. It describes how it has become easier than ever to express one’s thoughts and somehow also easier to express these thoughts without actually thinking.

Some examples of slop:

· Ad copies and event marketers using AI-generated images, without regard for obvious visual discontinuity or misspellings

· The increasing share of product reviews written by AI

· >10% of images on certain searches being one-shot generations, created in recognizable “AI style”

We’re currently in this weird period where only very clued-in populations can actually discern AI-generated content, and the rest cannot. So in the short term, in the same way that traditional advertising had its heyday as Americans got televisions in the 1960s, slop is and will be, at scale, similarly effective in influencing very young and very old populations. In the medium term, enslopification will have a more profound impact, and determining what makes content “good” or “bad” will make us all cynics first, and consumers second.

Some thoughts on the downstream impacts of enslopification, on the basis that it will make much of the internet unusable:

· Widespread agent use will probably occur faster than most people think

· Backlinks as a metric for quality will die, except from brands that had value pre-tipping point

· People will use increasingly esoteric proxies as signals to other humans to indicate quality. As college is a proxy for intelligence today, a [insert proof-of-personhood via a culturally-dense artifact] will be a proxy for intelligence in the future

The longer term is when AI enters a recursive fast takeoff, at which point the content AI produces will actually be better than what humans can produce. But we will already have had the content combinatorial explosion, and so this content will only be useful insofar that it is parse-able by other AIs.