Boiling the Oceans for Nothing

“AI” has been a silly little buzzword for a while now. What big tech marketing calls “AI” is a more interactive spin on machine learning called a Large Language Model (LLM), and there’s nothing intelligent about it.

While I was hunkered down to finish the first draft of Skybreaker, LLMs blew up in the cultural discourse. Nvidia and affiliated tech companies are keeping the stock market afloat while this fake nothing-burger of a tech service is infiltrating its way into everything we use in a modern, connected world. Now, I’m not a luddite, but I am a skeptic, and so far the early returns on LLM utility are mixed at best. 

• Products with integrated LLMs, like Copilot in Microsoft Office, literally cost more, whether you want it or not (Zdnet)
• “AI” search results are slowly ruining the economy of the internet by depriving websites of click-throughs and ad revenue (Pew)
• “AI” data centers are raising consumer household electricity prices (Tom’s Hardware)
• ChatGPT is making us lonelier (Engadget) and hurting our ability to critically think (Time)
• Health insurers are using “AI” to deny necessary medical care (AMA)

 I won’t confidently declare that all of the above means that LLMs are a useless technology—I’ve used Gemini to fix several of my own poorly written SQL queries. “Help me fix this broken code” and “help me rewrite this sentence to sound more professional” are reasonable enough use cases, so long as the prompter has a basic idea of what they are doing and how to spot hallucinations. 

I will confidently say that, on an individual level, the price is awfully high just to save ourselves a few clicks on Google or in Microsoft Office. On a societal level, these tools are being used to boost the profit margins of corporations while making us dumber and poorer. 

So, no, what we call “AI” today will never be the “Artificial Intelligence” teased by Asimov, nor will it ever truly be able to think or reason. 

At best, we’ll continue to throw more GPUs and training data at LLMs to arrive at a more functional technological assistant that can eliminate some of our more menial day-to-day tasks. 

At worst, we’ll bake the Earth and boil the oceans only to train a hallucinating, sycophantic chatbot that makes the rich even richer while putting real humans out of work. 

P.S. 

“help me rewrite this sentence”

I believe this is a reasonable use case for LLMs at my place of work. However, I want to be clear that for the hundreds of thousands of words that went into writing Spacewalker and Skybreaker, exactly 0 of them were brainstormed with, written by, or rewritten by LLMs, “AI,” or any other buzzwordy ML-assisted tools. 

My bad ideas and bad sentences are wholly my own.