Current Micro-Season

Loading...

Loading...

Loading...

Loading...

March 22, 2026

Polly Wants a Better Argument

First, anyone repeating the argument to assert that LLMs are never useful discredits themselves with anyone who has access to the internet and enough curiosity to use an LLM for any length of time.

Second, asserting that LLMs do not and cannot serve any useful purpose actively prevents addressing the harms they can cause specifically because they do work.

SE Gyges engages with the "stochastic parrots" criticism of LLMs in the most rigorous and good-faith fashion that I've seen. Frankly, I'm completely done with the argument and don't think it merits nearly the amount of attention it continues to command, but I also recognize that simply dismissing an argument is unsatisfactory from a scientific perspective.

My biggest concern with the argument is the second one that Gyges lists. So many people have been taken in by stronger forms of this argument that it detracts from the ability to actually do something about the myriad ways that LLMs can negatively affect the real lives of real people.

I know this likely won’t convince anyone who has already made their mind up (welcome to 2026!) but if this convinces even one fence-sitter, it’s worth sharing.

← All notes

Send your thoughts

Name and email are optional.

ESC
Type to search...
↑↓ to navigate to select