Chat with someone for two minutes, and try to figure out if it was a fellow human being or an AI bot. You can end up talking with someone random, or with GPT-4 / Jurassic-2 / Claude or other state of the art LLMs. Think you can tell the difference?
Hey Hunters! :)
My name is Amos and I work at AI21 Labs. A few months ago I had a cool idea: what if we would create a text-based chatroulette-style game that matches humans with AI bots, and after a few messages you get to guess if you’re talking to another fellow player or an AI bot?
Well, our team decided to go ahead and build it, and here we are! I’m so f***ing excited to share this cool new web game and AI experiment called “Human or Not?” with the community. It’s a “social Turing game” that lets you find out if you can tell the difference between an AI and a human.
So how does it work? The system randomly connects you to either an AI bot or another human like yourself playing the game. During a two minute chat you can ask or talk about whatever you want. Once the conversation ends, you have to choose, and only then the game reveals the truth - if that was an AI or another player. The bots are based on all the latest and greatest models: our very own Jurassic-2, as well as GPT-4, Claude and other LLMs.
The game is free to play, and its goal is just to help all of us better understand the nature of these AI models in the context of not just a productivity tool, but members of our online world.
We hope this will be a weird, crazy, thought provoking and mostly funny experience, and we would love for the Product Hunt community to be among the first to play with it. We are curious to see which strategies you will choose in order to expose the bot/human in the game. We would also appreciate any feedback, and if you liked the game - help us spread the word around! :)
Thanks so much!
Amos & the team
Game is cool, but there are some dead giveaways (like - only humans don't finish their sentences in time and so on).
It would be nice if there were some additional points if you manage to fool the other side.
So people would try to sound like an AI, which would make it harder to guess.
Or even better - points should be only for fooling the other side, not for guessing correctly. It would be nice if you could experiment with that I think.
Just played a round asking 'Pineapple on Pizza. Good or Bad?'. From the reaction it was pretty obvious who was on the other side.😅This kind of reminds me of the game 'The Talos Principle'. The entire game (apart from solving puzzles) you're trying to figure out who you are, what makes a human - a human and have chats with someone on a terminal who you don't know is a human or not. Fun experiment ?makers
Love this concept! Can you tell if your conversation partner is human or AI? Great tool for learning about AI and human-like AI.
Love this concept! I’ll be sure to go give it a try now. Love things like this that use technology to challenge your brain. Great way to bring Chat GPT and gamify it!
Congratulations on a great launch! We’re launching tomorrow 🤞🏻
I really enjoy the idea for this, however I think it would be awesome if it also showed if the other human thought I was a bot or a human. Also, maybe a profanity filter or a report button as I've encountered some profanities in my short visit on the site already made by other humans.
Such a cool idea! Please tell me why you didn't call it "bot or not" instead of "human or not"!
