Thanks for the idea! At this time, we don’t “certify” individual projects, so to speak, but we do our best to support projects and help them get through the approval process.
@m-a.schenk the biggest challenge is matching reviews to products and then parsing that data at scale. We’re using public information the same way search engines do. The only difference is that we’re using language models to offer people summaries and other tools to better navigate what’s out there.
Good questions @m-a.schenk. From a copyright perspective, we will cite every source and trim the content that we make available on our platforms to be compliant with “fair use”. We’re still figuring out the details of what’s required, which is why this demo we’ve shared has limited information.
Out of curiosity, you use Product Pictures to understand the product correct? I am trying to use gtp-3 to identify if a product is X or Y, and can’t wrap my head around it as yet since gtp-3 works mainly on text-based inputs.
Sorry, a real newbie here, hence would love to get your feedback on this.
We don’t use GPT-3 for image recognition. Our process starts with the product name and then finds everything accordingly (reviews, images, prices, specs, etc.) using various methods. We only use GPT-3 for review analysis and summarization.