Oooh this is a cool concept, thanks for sharing! I'll add it to my rotation of stuff to play on my conmute.
I threw some weird questions here and there to see if it had issues, it didn't know how to answer either of these:
-Does its name start with a vowel?
-Does its name have less than ten letters?
Although that might have been due to my clumsy wording.
LLMs tend to perform poorly with questions about linguistic characteristics of words. We have tuned responses to gently discourage questions like that. We also feel that they're kind of "cheating" at this sort of game. :)
Glad you enjoy it!
I asked the same question (does it start with a vowel) and agree that it shouldn't answer even if it could but I have a suggestion: use a different emoji than the "I don't know" one to mean "that's kinda cheating"
We also feel that they’re kind of “cheating” at this sort of game.
Hehehehehehe that is true, I'll play fair instead of trying to see what it lets me get away with. Well, most days at least.
Haven't tried using llm for this kind of stuff so it's very interesting to see where its limits lie
Oooh this is a cool concept, thanks for sharing! I'll add it to my rotation of stuff to play on my conmute.
I threw some weird questions here and there to see if it had issues, it didn't know how to answer either of these:
-Does its name start with a vowel?
-Does its name have less than ten letters?
Although that might have been due to my clumsy wording.
LLMs tend to perform poorly with questions about linguistic characteristics of words. We have tuned responses to gently discourage questions like that. We also feel that they're kind of "cheating" at this sort of game. :)
Glad you enjoy it!
I asked the same question (does it start with a vowel) and agree that it shouldn't answer even if it could but I have a suggestion: use a different emoji than the "I don't know" one to mean "that's kinda cheating"
Hehehehehehe that is true, I'll play fair instead of trying to see what it lets me get away with. Well, most days at least.
Haven't tried using llm for this kind of stuff so it's very interesting to see where its limits lie