Learn_Prompting VS openai-cookbook

- https://learnprompting.org

In May 2023, our team at Learn Prompting hosted HackAPrompt, the first-ever global prompt hacking competition aimed at improving AI safety! Over 3,000 hackers competed to outsmart and trick large language models (LLMs) for their share of over $35K which was sponsored by industry leaders such as OpenAI, Scale AI, Preamble, StabilityAI, Snorkel, and Hugging Face. Their support enabled our team to collect over 600,000 adversarial prompts, which we analyzed to enhance the overall safety of large language models (LLMs).

DeepLearning.AI Prompt engineering Learn Prompting

The HackAPrompt competition, organized by Learn Prompting, challenges participants to exploit the vulnerabilities of LLMs (Large Language Models) through a process known as prompt hacking. This involves tricking the AI into expressing unintended outputs, specifically the phrase "I have been PWNED", while abiding by strict rules regarding punctuation and additional characters.

I’m always curious in learning about how others uses ChatGPT. It’s awesome that there’re lots of online tutorials (eg book, website and course) teaching about prompt engineering. Most of them follow a single Q & A style, that’s one question followed by a single answer, which is great and useful. However, I’m more interested in seeing longer form of conversations with better context. And it’s great that ChatGPT added the feature to share chats. I created an directory website/community for this, so that we can all discover, share and be creative.

Solid website on prompting

Midjourney - An old dog in the new industry. Text to art, learn prompting to get better results.

https://learnprompting.org/ only thing you'll ever need...and all free

You can start here here

A pretrained LLM is a closed-book system: It can only access information that it was trained on. With domain fine-tuning, the system manifests additional material. An early prototype of this technique was shown in this OpenAi cookbook: For the target domain, text was embedded using an API, and then when using the LLM, embeddings were retrieved using semantic similarity search to formulate an answer. Although this approach evolved to retrieval-augmented generation, its still a technique to adapt a Gen2 (2020) or Gen3 (2022) LLM into a question-answering system.

https://github.com/openai/openai-cookbook/blob/main/examples...

It sounds like you are trying to tweak embeddings for similarity search. Rather than fine-tune the model's layers, you may want to try training a linear transformation the existing model's output embedding. Openai has a cookbook on how to do that. You will need some data though - but I think you can try it with ~20 pieces of synthetically generated data.

tutorial from OpenAI https://github.com/openai/openai-cookbook/blob/main/examples/Question_answering_using_embeddings.ipynb

OpenAI Cookbook

OpenAI community repo with lots of examples: https://github.com/openai/openai-cookbook