What you need to know
- Human or Not? is a web-based chat game that pairs you with a random user that may or may not be an AI bot.
- Throughout the conversation, it's up to you to decipher whether or not your chat person is real before being randomly connected to another.
- AI21 Labs built the game, a company focused on building and understanding start-of-the-art AI language models.
Exciting new forms of technology usually follow a similar trend. Some manage to progress from a patched-together concept to mainstream tools more successfully than others, but there's no denying that ChatGPT had a hand in the recent explosion of language learning models (LLMs) making their way into everyday life. It seems the natural progression from there is to turn everything into a game, and this latest one would have either impressed Alan Turing or freaked him out. You decide.
AI21 Labs developed Human or Not? and it's exactly what it sounds like. Start the game and a random chat partner is selected. The catch is, you don't know for sure whether that partner is a human being or an AI-powered chatbot. It's up to you to decipher the answer with a short series of messages. If you ever thought Google's reCAPTCHA two-factor authentication images were getting accurate to the point of concern, wait until you see how chatbots can imitate the poor spelling and unusual grammar of regular people.
My first couple of encounters seemed like a laughably poor effort from a real person trying to sound like a machine, but it turns out that the robots behind this game are pretty clever at appearing to be a little dim. Curious inconsistencies with capital letters and even littering sentences with slang had me second-guessing early on. Still, the big giveaway is usually when a conversation ends immediately because the other player has quit.
It's an interesting break from talking to Bing Chat, which feels a little more pedestrian compared to its more unhinged days of wearing the 'Sydney' codename and playing Human or Not? quickly brought me back to how I felt during the early days of ChatGPT when it would provide answers ranging from inaccurate to insane. Watching AI assistants morph into something more accessible and balanced has me forgetting how convincing they could be if the restraints were off.
Head to the Human or Not? website and try it out for yourself. You might even run into me as I try to imitate a well-meaning chatbot, but so far, I've mostly only convinced my fellow humans that I am also a fleshy, brain-having member of the species. Removing a search-focused element like Prometheus on Bing Chat leaves you with nothing but the layer responsible for impersonating a genuine assistant, but this is purely for fun.
Still, those movies about sentient robot butlers are starting to feel a little more realistic. Maybe we should stop asking them stupid or dangerous things just in case a GPT-powered helper remembers all the times I asked it silly questions and requested highbrow topics explained to me like we're monkeys. I hope not. Unless it found that stuff as funny as I did. Fingers crossed.
