@rch
I spent the last few weeks building a few simple agents. One realization had been how similar they can be in some ways to humans.
an experience that reinforced it today. was on our daily evening walk with W and I was talking about a problem that I discovered recently by auditing an ops centric MS Teams channel with claude. Half way through the conversation she got distracted by a conversation of a bunch of teens sitting close by.
As we walked past them, I quizzed her on what I was talking about as a jest.
To her credit, she was able to recall most of it. But also added more stuffs like I had claude create a dashboard and a system to prevent the errors.
Agents hallucinate and so do humans when their knowledge is incomplete.
0 reply
0 recast
2 reactions