MatchGPT: Train your GPT

Can OpenAI’s GPT model, fine-tuned with our collective preferences, help identify the perfect match for us? Consider this: If a thousand participants label their most and least favored answers for 100 intriguing questions, and we then use this labeled data to fine-tune OpenAI's GPT, would the model be adept enough to identify ten individuals who could be potential close companions for you? Could one of these ten even become a significant figure in your life?

Our dataset comprises 2,000 distinct questions, with a staggering 70,000+ voice-recorded answers. Rest assured, you won't be matched with anyone from the initial dataset. After all, you wouldn't need GPT to connect you with voices you've already heard. Instead, the GPT's task is to introduce you to one of the 1,000 experiment participants. This is our way to gauge GPT's prowess (or lack thereof) in this particular realm. As you rank answers for your selected questions, you'll have the chance to record your own response and, if you wish, share it with the community behind the dataset's creation to potentially start new conversations.

Are you intrigued by the potential impact of GPT fine-tuned on our collective data? If so, we welcome you to be a part of this exciting journey. As soon as we reach 1,000 participants, we'll initiate the fine-tuning of GPT with this fresh data. We'll then run your personalized GPT version against the voice recordings from our 1,000-member group. You'll have access to the answers from your matches, offering you a unique opportunity to listen, connect, and perhaps forge new bonds.