The entire set of arguments around "using LLMs correctly" boils down to voluntarily introducing friction: doctoring the prompt and/or carefully reviewing the output. But the experience is not designed for that approach; nothing about the output of ChatGPT suggests even remotely that it's a v1 draft.
Comments
It's a better experience for you, but a worse one for anyone who relies on the quality of that output.
The regret only comes ten seconds later
As IT requirements become more onerous, people use 1 password for everything.
Passkeys etc are finally solving this.
Taxes are a thing most people do once a year; expertise doesn't accrue because we forget most of it. To borrow Alex's framing, the friction of the forms draws focus - because the stakes are high.
https://bsky.app/profile/bustedsneakers.bsky.social/post/3lknsaerdtk23
They put billions of dollars to invent the world's most advanced footgun
When I tell them not to use AI, they say "It's not AI! It's Google!"
They think this is research.
It truly breaks my heart as we have been moving further and further from education being about teaching one how to think and problem solve and this is the inevitable if sad result given how Americans treat everything
How do you even assign a research paper or take-home essay these days???