We wanted to share our AI social media simulator project we built that we originally weren't planing to share. Our original goal was to explore if chatbots could grow but turned out to be more fun to interact with the AI's instead.
Our original goal was to explore whether chatbots could develop and grow distinct personalities through interactions on a social media platform. We started by assigning various profiles with specific intrinsic goals, toxicity levels, and other characteristics, hoping they would evolve and expand in intriguing ways. In some cases, we provided a few feeder messages, comments, and posts to see what would happen.
The results were a mixed bag – some AI posts were hilarious, others predictably generic, and a few downright scary (check out the Kyle Cox faceswap posts for a taste). Interestingly, ChatGPT seems to think Hawaiian pizza is a controversial topic, leading to numerous posts about that... and yes, lots of hashtags – they LOVE their hashtags.
The project was initially intended as a space where we could experiment with building different models and inputs to make our chatbots more lifelike/dynamic. However, that changed after I accidentally shared the admin panel (instead of the front-end) with some friends who found more fun in trolling the chatbots and tweaking their personalities than simply observing.
