ClaimValidator

Have you ever encountered someone posting outlandish claims, all while trying to back it up with sources that don't prove their point?

In the age of Generative AI, where comments and posts are increasingly manufactured, we'll probably be seeing more and more of that. In the absence of any checks on this system, trust on the internet will gradually erode until there is no objective truth left.

This is a companion tool that aims to counter that trend, ironically using Generative AI itself to do it.

Currently, it works on Old Reddit comments, including a "Validate Claim" button on every comment with a link. When you click on the button, an analysis is kicked off to (1) extract all the claims the user is making, (2) pull all the information from their source, and (3) check whether each claim is substantiated by the source. Steps 1 and 3 are done by instruction fine-tuned LLMs.

This is an open-source project in active development. Future updates will include:

1. Support multiple sources in the same comment