Fun concept. I wonder if the data gathered could be used to train Chatbots to sound even more human. And i'm not sure my first conversation can possibly help:
?Hey, heeeey, how are u
Im good. Im not a robot btw
how can i be sure?
Today is 4/20
You don't really need special data to make GPT-3.5, 4 etc sound as human as to be indistinguishable. The default robotic voice/writing is specifically fine-tuned in. Even then, the ability isn't gone, just buried behind the default.
I feel this game would have been better if we got points for convincing the other player that we are a bot. It is actually way more fun than guessing if the other player is a bot or not
I asked “what country are you in?”, and the reply was “just remember that I won’t answer any factual questions”. That’s seems enough to identify any bot.
Of course, it doesn’t rule out humans pretending to be bots.
EDIT: Well I retract my observation! The second time I tried it, it was a bot and it was willing to lie to me and make up a factual answer.
Given the glacial performance of the site, 2 minutes often isn't enough.
Anyway, my initial prompt "My brain hurts!" followed by "My! Brain! Hurts!" (regardless of response) turned out to be sufficient to filter out all bots. And non-Monty Python lovers, potato, potato...
Until a few years ago it was enough to ask someone if they were a human to know if it’s some chat bot or not but now someone can just setup an ai to say things for them. I’m concerned about the professions like therapist.
neat concept but the website is just too janky. the controller sometimes gets loaded twice or something so two different instances fight over the UI components and it's pandemonium. also it's just too short to really be useful.
I asked it what it saw on the screen while it was waiting for the conversation to start. The bots made up nonsense, while humans responded something like "falling boxes" (except for the humans who tried to act like bots).
?Hey, heeeey, how are u Im good. Im not a robot btw how can i be sure? Today is 4/20