ChatVerse

Created by team DeJaNi on December 08, 2023

Idea: We plan to the better use the outputs generated by current LLMs by letting the users' to make the chat with LLM public. This chat once public will become visible to all users' within the platform who can then like, share, comment on the chats. Based on the Data liked by each user, we can personalize their own chats by leveraging the user traits or interests. Given the chats are now backed by human vote, their is an increased chance for less hallucination. We also plan to enable SEO for the chats so that if any user generated a good output for a question, the same output can be later used for other folks. Impact: We think that the Idea which we're proposing here can be helpful to any LLM creator and can help them create a business model. As of today, AI Startups need a lot of funding to build LLMs' and we think out Idea can help them generate model in some form. Additionally companies like HuggingFace who work on open source can also leverage this Idea to generate a revenue which they can spend into their open source initiatives. We also think that HuggingFace can create an open source social media so that users' can also understand what Data is being collected and on what basis is the recommendation being done. Prototype done as part of Hackathon: We've build a website which displays the outputs generated by the LLM. We also added features such as Like, Comment. Given personalization is an ongoing active research problem, for now to show how personalization can be done, we've Engineered a custom prompt where the user's traits will considered and based on this, the user's output is personalized. For better understanding, we've mentioned everything within our Demo

Category tags: