ufal's log
2025-07-09-03
Lemme drop some references to spot if some text online is written by AI or not, because recently there is a Facebook fanpage sharing info about racehorses that's too obvious that the admins are using LLM model to write the caption.
I Can Spot AI Writing Instantly — Here’s How You Can Too by Evan Edinger
The rise of Whatever by fuzzy notepad
This one is not really about AI writings, but you can get the glimpse about what LLM doing when "helping" user to code, and the other stuff
Everything around LLMs is still magical and wishful thinking by A Place Where Even Mammoths Fly
Why A.I. Struggles with Negative Words by Otherwords
This is just a bonus, but you will understand why LLM model is unable to understand what is a "no"
For myself personally, I have been experimenting with GPT 4o-mini model for help me troubleshooting on coding and help me setting up stuff on my Linux machines, it sucks, totally. What it did is just bunch of hallucinated answer, giving non-existent repository on GitHub, always keep the positive vibe even though as a newbie I know it's not even possible. Dude, just spit the fact without sugarcoat it!