The rise of AI zombies

preface

This article is a collection of my thoughts on general-purpose LLMs. These I’ve come to realize cause more problems than they solve. At least in my eyes. I must emphasize that this article is geared specifically towards the general purpose LLMs marketed to the public as “AI”. I’m not an anti-AI fearmonger that thinks it’s going to take over the world (I’m more convinced that a way sadder and less exciting dystopian future is awaiting us). And I do recognize that there are some good use cases for AI, though these are rather limited and not nearly as great as advertised. Still, the little camera that detects whether the label is positioned correctly on your soda bottle has been doing a solid job for the past 20 years and will do so for the foreseeable future. Same technology in essence, but applied in a way where it makes sense.

Let’s start at the beginning.

So somewhere at the end of 2022, ChatGPT got introduced to the world. This started a honeymoon phase for many that started using the service in the beginning for hypothetical tasks, later for real-life tasks.

This is also exactly how I got pulled into the AI maelstrom. I started including it in my workflows, seeing where I could gain the most efficiency. And of course inevitably, to skip the hard work of reading actual documentation to hit results faster.

AI works great, until it doesn’t

At first, this worked great; I spent less time on average completing tasks, and the things I produced were of seemingly higher quality than before. Life was good, until I hit that (one) problem AI couldn’t solve. To get rid of this problem, only one option remained: RTFM. And so that’s what I did, discovering in reading that AI had just made up some things that didn’t make sense at all and did things in a quick, unmaintainable internet solution way that looked good but was unsafe and unsustainable. Cue the full rewrite of everything to fix the actual problem.

At this point I realized the problem with embedding AI in my workflow: it makes me feel more productive, but in the long run, it’s costing me big time. This all whilst I was using AI in a “responsible” manner. I didn’t ask things I did not really know the answer beforehand, and I was careful to read its output before implementing it. Still, it was slowing me down in a subtle but sure way by making my brain lazy and introducing problems that could have been foreseen by just going slow and reading documentation first.

The rise of the AI zombies

While going through my own AI journey, others were starting to discover it as well. It took the world by storm in a scary way; meetings were held on how to embed AI in our workflows, what we should allow, what not, etc. What we didn’t realize at that point was that this would be a way bigger problem than any of us could ever have foreseen.

Not only were people using it to gain some speed in generating code and text. A large portion of people started to use it as a search engine. This evolved into the current state: people just continuously have an AI window of sorts running so they can interact with it. These people have apparently lost any ability to make decisions or solve problems without consulting AI, they’ve essentially become AI zombies.

Now what?

If you’d believe popular media and the internet at large, AI is going to take over the world and replace us all at our jobs. This idea has mainly been put forward by some “big AI” CEOs that fared well with promoting the “AI can do everything” concept. These days even those CEOs have slowly been coming back on their statements of “General AI is within reach” and are promising less. Still, the promises vastly outweigh the probable future for general-purpose LLMs.

Personally, I’m more convinced that general-purpose AI will go down the same route as social media did. It’ll stay but become a crappier shadow version of itself, and it’ll be infected with advertising in a last-ditch effort to recover some of its cost. Answers will probably be limited, and it’ll be trained increasingly to follow your thoughts to keep you hooked and serve you more advertisements.

And so we enter an era of AI zombies. No big AI takeover, no big war between men and machine. Just a bunch of stupid people becoming more stupid and enslaved to the device in their pocket. It’s boring and sad.

Can we fix it?

Maybe, probably not. Some might escape it like some have already escaped the social media addiction. But my fear is that a majority of mostly younger people will be lost to AI for the foreseeable future.