Show HN: FAQify – Algolia for FAQs using GPT-3

[fatihcelikbas]

Hey HN! My name is Fatih and I’d like to share FAQify - a project I’ve been working on this week. FAQify is a tool that answers questions about Amazon products using GPT-3. Simply paste the link to an Amazon product and ask your question! The app will generate a good answer based on the textual data on the product webpage.

The current version of FAQify is a simple POC, but I hope to expand it to be something like Algolia for question-answering. The idea is to build a widget that you can integrate with any website. You give it both text on your page but also external data like user manuals. Once it’s trained on these texts, it will be able to help your customers get answers to their questions 10x faster. This can be useful in many domains such as e-commerce websites, code documentations, and blogs.

If you have any feedback or other use cases in mind, feel free to email me at [email protected].

Thank you!

[fatihcelikbas]

Here is the Reddit post on FAQify. It includes a demo video as well: https://www.reddit.com/r/GPT3/comments/yrrpnc/i_created_a_fa...

[nicoburns]

This seems pretty useless, because won’t GPT3 happily synthesise false information? What good is it answering my question if the answer isn’t correct? Google search’s “question answer” section has a similar issue.

[fatihcelikbas]

You can condition the prompt to make it say 'I don't know' etc. which is what we actually did too. Give it a try with a random question.

P.S. This is meant to be integrated to individual vendors' websites so the data we will get from will be much more detailed, making answers much more to the point

[zicxor]

Good project, keep going!

[fatihcelikbas]

Thank you!