When people talk about AI today, two reactions are common:
- excitement
- fatigue from hearing about it everywhere
But there is one well‑known weakness of generative models that everyone recognizes:
hallucinations.
Large language models can produce answers that sound perfectly plausible — even when they are completely wrong.
Yesterday I realized something funny.
Human LLMs have existed for decades.
I recently started watching Cheers again, the classic sitcom set in a Boston bar. And suddenly it hit me.
One of the regulars — Cliff Clavin, the postal worker — behaves exactly like a hallucinating chatbot.
Cliff Clavin:
- can answer any question asked
- the answer sounds authoritative and detailed
- fact‑checking usually makes it collapse quickly
- he is absolutely confident while explaining it
- and the explanation tends to be… very long
In other words, Cliff behaves like an LLM running at high temperature.
And of course, he never says:
"I don't know."
Once you see it, it is hard to unsee.
Every workplace probably has at least one human LLM.
