The AI post

All anyone wants to talk or write about is AI, and I’ve been avoiding it because I didn’t feel like my voice could add much to the cacophony of other, smarter, and more experienced ones. But I do have my opinions, and I haven’t come across this angle yet, so here we go.

I recently read a piece that described a post-AI world as being “hollowed by machines,” and that mental image stuck with me. Instead of filling the world with the infinite creativity of humankind, we are entering a period of accelerating AI vomit.

Not that people aren’t creating things anymore, but the vast international corporate machinery that is constantly scraping for another shred of profitability is employing AI to drown out anything unique with redigested, regurgitated, recomposed slop.

It’s easy to feel buried in that slop, but we’ve always had a signal-to-noise problem. The incentives to create vacuous marketing drivel will always vastly overpower every other creative impulse. AI is only making it cheaper for marketers to drown us.

The bigger problem with AI—especially when we set aside the emotionally and culturally charged topics of art and literature and focus on software—is that AI can only regurgitate. So where does it get its raw materials? What source has it digested to spew out its generated programs?

As we sink deeper into this AI pit, the answer becomes clear: it’s consuming its own vomit.

As the human puppeteers direct their fleets of AI slaves to generate millions of vacuous listicles explaining in far too simple terms and with far too many words how to install a Ubuntu package, those same LLMs are constatly crawling the web, drinking from a firehose of that same vomit.

You can photocopy a photocopy maybe one or two times before it becomes unusable for any purpose. How long will it take before 80% of the information found on the web is a fuzzy simulacrum of anything real? Listicles based on listicles. Stack Overflow answers written by robots trained on other robots’ over-digested excretions.

We might be headed toward something that resembles an information apocalypse.

This isn’t really new news, either. Ever since the Google search cat-and-mouse game began back in the ’90s, we’ve been unwitting victims of this technological power play. Marketers analyze Google’s behavior to figure out how to move their content up in the search results, so Google responds by modifying its behavior to temper that “optimization.” Lather, rinse, repeat.

Before you know it, “trust” on the internet is a fantasy; PageRank was built for an internet of humans acting in good faith, but on today’s internet only a sliver of what you encounter was made by a person, from their own inspiration or knowledge or heart.

Forbes wants to be at the top of your search results page for “best dehumidifier,” despite having no qualifications whatsoever to answer that question for you. Backed by armies of unsleeping, unquestioning robot writers, they’ll work to convince Google that they do.

When you ask ChatGPT what the best dehumidifier is, it will happily reference 19 different worthless listicles, some of which it probably wrote, and give you a very confident and partly hallucinated answer.

The AI chat bot is a lot like a child wandering the library, following footnotes forever, assuming the citations to be true without reading any of the source material, blissfully unaware of the increasing absence of source material.

Comments