---
AI Answers and Gell-Mann Amnesia
2026-05-31
---

When I get a google search  result that's AI generated about a topic
I know,  it's usually pretty accurate  -- I'll admit. But  when it's
not, I  know enough about the  topic I'm searching for  to know when
it's feeding me bullshit. Even if  I forget the exact details of the
thing I'm looking for, I know what /wrong/ looks like!

It makes me  think about Gell-Mann amnesia: how we  can quickly spot
bad news  reporting on a topic  we're deeply familiar with,  yet for
topics  where we  aren't  experts, we'd  trust  the same  reporter's
accuracy. Can  the same be  true for AI? If  I ask it  a programming
question and get a  spotty answer, I know it's spotty.   If I ask it
something  about gardening  am I  willing  to trust  it's output?  I
should  be applying  the same  level of  scrutiny regardless  of the
search topic.

I try to  avoid generative AI whenever I can,  but sometimes I admit
to  going with  the first  result given  to me  by google  (I really
should look at  finding a different search  engine...) which happens
to be the generated one.

What a time to be alive, eh...