Inspiration
Throughout the ages, humanity has been fascinated with stories of unrequited love. Narcissus from Greek Mythos fell in love with his own reflection, and unable to leave the allure of that beautiful image, he "melted away from the fire of passion burning inside him." In the Hunchback of Notre Dame, Quasimodo's love for Esmeralda is blind and hopeless, while Esmeralda is driven by her own love for Captain Phoebus. Perhaps even one of you may know someone who has been in the trenches of the friend zone, unable claw themselves out of the pit of their devastation and the chasm of space between girl and friend. Victor Hugo succinctly describes this in his Signet Classics: "The inexplicable fact is that the blinder [love] is, the more tenacious it is. It is never stronger than when it is completely unreasonable." Our team has seen the bloom of the internet - and specifically of LLMs - exaggerate humanity's quest for love. Yet not love that is deeper, more profound, or more connected - love that is more unrequited. Huge content creators are being worshipped and praised in extraordinary proportions, and millions are being spent in fruitless tokens of affection over dozens of sites. We've also seen the rise of sites like character.ai where people are catching feelings for robots, these false caricatures of sentience. Is this the future of humanity? Is this the future of love? We don't know either, but we know this - in a gold rush, sell shovels.
What it does
Our app lets users chat with fine-tuned models of ChatGPT. We currently have two models: Donald Trump and an anime waifu, which were both built in-house. These fine-tuned models will talk to you about whatever topic you desire. Have meaningful or funny conversations with Donald Trump or the Anime Waifu!
How we built it
We employed cheap labour to get conversations to fine tune our models, which we fed to OpenAI's money machine. We used Nextjs, Tailwind and Framer Motion for our website, along with Firebase for some database storage.
Challenges we ran into
Framer motion was fun to work with but we went down a lot of dead ends trying to make animations that were not really feasible for our skill level. Fine-tuning gave us a new learning curve for us to figure out which took some time to get used to. But in the end, we overcame these challenges.
Accomplishments that we're proud of
Some accomplishments that we were proud of are: Fine-tuning a model that suits our project requirements by using sample conversations, creating a simple and user-friendly UI website where you can start your conversation with the characters of your choice, animating our website with Framer motion with rizzy sound effects.
What we learned
We learned that OpenAI has good documentation. We realized that fine-tuning a model does not actually require as many samples as we thought it needed, and that sub-one hundred samples is enough to train a fine-tuned model.
What's next for The Conversationalist
We shall make people fall in love with our models.
Built With
- firebase
- framer-motion
- google-cloud
- nextjs
- openai-api
- typescript


Log in or sign up for Devpost to join the conversation.