On the forum for language learners where I'm active to get help learning Spanish, the other users are begging the app developers to update the grammar breakdowns generated by the built-in ChatGPT feature with the newer and better ChatGPT models … for tens of thousands of sentences.
First of all, I've come to hate interacting with other forum users. Whenever someone asks about my native language and I, a native speaker, take the time to answer, I always get, "You're wrong. ChatGPT says something else."
Second, the LLM-generated grammar breakdowns are absolutely worthless and *always* wrong. Sometimes just a little, sometimes completely. Even for such constrained problems like grammar breakdowns, when there's a fixed and known set of rules and the LLM ingested all grammar books.
Third, I dared to explain in the forum the correlation between the use of "AI", the demand for data centers, the environmental destruction, and climate change. I asked the other forum users to please believe native speakers more than the "AI". For context, I also mentioned that I'm not some random know-nothing but that I have a background in computer science and understand how ChatGPT works under the hood. And I laid out why the "AI" isn't some infallible oracle but an unreliable source of information. (The forum is supposed to be kid-friendly, thus no swear words wanted.)
I was immediately attacked by addicted users in denial, like, "Oh, Mr. Smarty Pants who went to university thinks he's smarter than all non-academics among us." Some called me a Luddite. It was then I who was scorned by the admin of the forum and lead developer of the app that it's unproductive behavior to attack his user base. If the users want "AI", they'll get "AI", the forum admin and lead developer said.
A month ago, I posted a link to a news article that reported on how France, Spain, and other countries are literally burning right now. That's the cost of LLM usage.
🧵⬇️
#GenAI