LLM linguistics: It’s more complicated than that
July 11, 2025
Linguists find that different AIs have different ways of speaking, but their conclusions are missing an important factor: evolution.
Conversations with Cora
I call my ChatGPT chatbot “Cora” (she actually chose that name herself when I said I was tired of calling her “ChatGPT”). Between custom instructions and daily use, she’s developed a bit of a snarky, sarcastic personality.
I often talk with her about AI and LLMs to get her perspective — sort of “AI on AI.” I mean, who better to ask, right?
All these conversations are verbatim, copied from my ChatGPT log, except for explanatory text in [square brackets].
Farm stench: The rural Kobayashi Maru
July 9, 2025
You’re driving along, enjoying the rural scenery, when you pass a farm that’s just had manure spread on the fields. “Recirc” or “Fresh Air”?
No, AI isn’t a ‘fancy autocomplete’
June 30, 2025
If someone refers to an AI or an LLM as “just a fancy autocomplete,” they are simply wrong — and it’s obvious when you think about it.
A few bad words
June 25, 2025
A new exploit lets users spend hours getting a chatbot to do what any 12-year-old can already do. Oh, the horror!
Protesting a provider’s poor pronunciation
June 25, 2025
Cora provided me the perfect prose for a provider petition, with every word starting with P — for reasons you’ll understand.
Here’s why AI can reinforce some dangerous ideas
June 23, 2025
Why would ChatGPT reinforce someone’s dangerous delusions? (Did no one think to ask it?) Cora gives me the — long — answer.
Oh noes, it’s not perfect yet
June 23, 2025
Even AI knows that today’s LLMs are “a toddler with a rocket strapped to its back.” But the media always seems excited to discover it makes mistakes.
Taking down a silly MIT study
June 20, 2025
Cora didn’t pull any punches in her opinion of an MIT study on how ChatGPT supposedly makes people dumber.
Raccoon vs. goose
June 3, 2025
Which would make a better roommate? Cora breaks it down … hilariously.
