2 ms·
Show HN: I built a tool to crowdsource and share LLM experiments
Hi, I'm Diego, the creator of Will AI Do This?
As part of my job, I need to be news related to GenAI and LLMs in particular.
I often found myself uploading promtp + answer + image to posts on linkedin and twitter, only to never find them again when I wanted to share them elsewhere.
Sometimes I would create a blogpost for a collection of prompts, but it often felt like an overkill.
Will AI Do This allows you to upload prompts (with images), renders code and it supports basic metadata (was the prompt successful?, what model was used?, etc).
Filters and custom pages coming soon!
Quick summary of the website: https://youtu.be/Dw2MqYO8WkM?si=fsahGG_dOYDVMWeq https://youtu.be/Dw2MqYO8WkM?si=fsahGG_dOYDVMWeq
How to upload a prompt: https://youtu.be/ESne3HhQXQQ?si=nq6NzsAiC7Maj5WN https://youtu.be/ESne3HhQXQQ?si=nq6NzsAiC7Maj5WN
- nomilk 3y agoThis is cool (love the demo video; straight to the point) Do you have a sense for how exhibits can be tagged (e.g. what category they belong to). For example, if I was curious to see how others had experienced LLM hallucinations, dangerous answers, incongruent compounds of otherwise correct answers, and any of 100's of other LLM quirks, then I could perhaps click on a category and see dozens/hundreds of submitted examples? (but it's hard if they're not categorized) One other thing, I've seen people edit the page HTML in chatGPT to trick others into thinking it said things it didn't (I think even a NYT article famously misled readers because it omitted the start of a conversation where the user told the LLM to act like a psychopath). But tl;dr is there a way you can verify submissions are genuine?
- kiedanski 3y agoHi, thanks for the interest. Indeed, having categories and filtering by those is one of the features I am currently working on. Regarding the verification, I don't think it would be possible now (there is no way to share conversations AFAIK).