Discussion
I am lurking in a slack instance full of tech executives and just saw a comment go by about how somebody felt like Claude was gaslighting them and so they told it not to and instructed it to set up a framework to prevent it from gaslighting them again and I just...
What are we doing here. What are we doing here. What are we doing here. What are we doing here.
Healthy relationships, reliable tools, what's our model here of how anything does or should work.
What are we doing here.
@mhoye imagine instructing a chatbot to devise a PIP for itself and then adhere to the PIP instead of doing literally anything else in the whole world
@jenniferplusplus @mhoye And imagine the human equivalent: "one of my team members needs to be specifically told to not delete the production database, every time I give them a task."
The human wouldn't have a job for very long.
@jenniferplusplus @mhoye That's actually a thing. Projects have a CLAUDE.md file in them now to do something just like that.
@mhoye @grmpyprogrammer I wish everyone could go through the AI training I was forced to go through a few weeks ago. In it, the instructor meticulously showed us exactly how LLMs (and the additional tools around them that we collectively call “AI”) work, and it’s very clear from the math and everything that these are very sophisticated auto-complete machines. That’s all. There is no intent in anything they do, and the sooner we stop treating them as if there is, the better for all of society.
Install bonfire.cafe
Get the full app experience
