3 ms·
This is not a complaint. But as someone who uses AI heavily for tailwind, I found that GPT-4 couldn't update the syntax even after the cutoff is updated. Likely
by YetAnotherNick 3y ago
This is not a complaint. But as someone who uses AI heavily for tailwind, I found that GPT-4 couldn't update the syntax even after the cutoff is updated. Likely it gives equal weightage to documents regardless of the date. And even with 2023 cutoff it just can't work with new syntax of many libraries I tried.
I had hopes with google in this as they already likely have things like official docs, updation etc labelled so they can give different weights to every document. But then there is Gemini.
- duckmysick 3y agoHave you experimented with Retrieval-Augmented Generation (RAG)? Here's a good primer on it: https://aws.amazon.com/what-is/retrieval-augmented-generation/ https://aws.amazon.com/what-is/retrieval-augmented-generatio... And here's a recent HN thread about RAG on PostgreSQL: https://news.ycombinator.com/item?id=39613669 https://news.ycombinator.com/item?id=39613669
- YetAnotherNick 3y agoYes, it's not as good as the base knowledge of the model. And in a way even chatgpt/bing web search is RAG.