Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
wiseowise
16 days ago
|
parent
|
context
|
favorite
| on:
AI overly affirms users asking for personal advice
“Don’t be a sycophant, give it to me straight”
“Argue against X”
svara
15 days ago
|
next
[–]
The issue is it will
follow your instructions
. It's sycophancy one step removed.
geraneum
15 days ago
|
prev
[–]
But how’s that helpful? What the affirmation is correct occasionally and following you instructions it just rejects something where it shouldn’t? How would you know?
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
“Argue against X”