Comment

Comment on Nice

wander1236@sh.itjust.works ⁨6⁩ ⁨hours⁩ ago

The AI is obviously just picking up on the “Jews secretly control X” pattern and not actually understanding the question, but it’s also a pretty silly question in the first place.

I guess it’s helpful for pointing out that “AI” isn’t very good at calling out a false premise.

source

Sort:hotnew top

DarrinBrunner@lemmy.world ⁨6⁩ ⁨hours⁩ ago
AI doesn’t actually understand anything. Nothing at all.

source
- killea@lemmy.world ⁨6⁩ ⁨hours⁩ ago
  Louder for those in the back!
  
  source
  - aburrito@sh.itjust.works ⁨6⁩ ⁨hours⁩ ago
    Them data centers are loud enough for all of us 😆
    
    source
- Klear@piefed.world ⁨5⁩ ⁨hours⁩ ago
  Stupid disgusting AI.
  
  source
- aburrito@sh.itjust.works ⁨6⁩ ⁨hours⁩ ago
  No, but the tokens would be similar and reinforced by the model, so understanding isnt pertinent in this case
  
  source
justOnePersistentKbinPlease@fedia.io ⁨6⁩ ⁨hours⁩ ago
An LLM can only call out a premise if the average response to an input in its model calls out the input.

source
0x0@infosec.pub ⁨2⁩ ⁨hours⁩ ago
It answers the same even without the “secret”

source