Comment

Not to disagree with the sentiment, but the screenshot is fake. I just tried the same prompt and it solved the “puzzle” without issue.

Sort:hotnew top

brbposting@sh.itjust.works ⁨1⁩ ⁨month⁩ ago
The system’s not deterministic, I think they say, try 10 times and maybe get five or 10 different answers.

source
- JohnyRocket@discuss.tchncs.de ⁨1⁩ ⁨month⁩ ago
  Sure, but the fact that the top of the first prompt is cut off is a little sus.
  
  source
  - brbposting@sh.itjust.works ⁨1⁩ ⁨month⁩ ago
    Perhaps, yes.
    
    I’d already scrolled the thread by this point and it’s actually surprising just how many people tried it themselves. Tons of confirmation and comparisons to competing models.
    
    source
The_Dark_Knight@lemmy.sdf.org ⁨1⁩ ⁨month⁩ ago
No it isn’t I had made a comment trying out four LLM in this thread and none of them got it right and the answers were more or less similar .

source
31337@sh.itjust.works ⁨1⁩ ⁨month⁩ ago
I tried it once with GPT-4o, GPT-4, and GPT-3.5, Meta AI, and Gemini. They all failed. Pretty interesting.

source
NikkiDimes@lemmy.world ⁨1⁩ ⁨month⁩ ago
I just tried with 4o and got the same response as OP.

source