trigaten / learn_prompting Goto Github PK

Prompt Engineering, Generative AI, and LLM Guide by Learn Prompting | Join our discord for the largest Prompt Engineering learning community

Home Page: https://learnprompting.org

License: Other

JavaScript 0.82% CSS 0.09% TeX 1.39% Python 0.09% TypeScript 0.05% MDX 97.56%

gpt-3 prompting prompt-engineering gpt3 llm chatgpt gpt-4 gpt-4-api large-language-models nlp

learn_prompting's Introduction

Learn Prompting

https://learnprompting.org

Prompt Engineering, Generative AI, and LLM Guide by Learn Prompting | Join our discord for the largest Prompt Engineering learning community

Contribution Guidelines

We welcome contributions in any form.

We are actively looking for:

content suggestions
translation
content/art contributions
typos :)

Local Development

First, install git and node.

Make sure you are using Node 18.0.0 or higher (node -v). Then, run the following commands in a terminal:

# download the website code with git
git clone https://github.com/trigaten/Learn_Prompting.git
# enter the project directory
cd Learn_Prompting
# install node modules
npm i
# run the website locally
npm start

If you get an error related to the node version, you probably are using an older version of node.

Make sure the newer version appears higher in your path than any older versions.

Alternatively, you can use nvm to install the latest version of node. Install nvm, then do the following:

export NVM_DIR="$([ -z "${XDG_CONFIG_HOME-}" ] && printf %s "${HOME}/.nvm" || printf %s "${XDG_CONFIG_HOME}/nvm")"
[ -s "$NVM_DIR/nvm.sh" ] && \. "$NVM_DIR/nvm.sh" # This loads nvm
nvm install-latest-npm

npm start starts a local development server and opens up a browser window. Most changes are reflected live without having to restart the server.

Thanks to all contributors ❤

Cite

Use the provided GitHub citation in this repository:

@software{Schulhoff_Learn_Prompting_2022,
 author = {Schulhoff, Sander and Community Contributors},
 month = dec,
 title = {{Learn Prompting}},
 url = {https://github.com/trigaten/Learn_Prompting},
 year = {2022}
}

learn_prompting's People

Contributors

Stargazers

Watchers

Forkers

wesjd ianb lucas-kohorst hrocha stjordanis janek brianjking alphapro06 whoaa512 voltek62 fundou planetscape standardgalactic harounahmed2 drewarnold1 netzkontrast carlossosamorera bbswitzer antsanchez reidoutloud aalviani asterfinde thepok zjrwtx adrianaul funkmyster pablosalazar7 codebunkr-charlie cristianoamaral72 cristina-gabriela thegovind elguille repoztrees adesanyaaa seizethedayta pterameta fabiochiusano tyjk renatonagliati jtone123 wesen nzores nitesh053 bernbecht emilsabri iboss21 dastardi freddiefujiwara vkochukov jlhasson adamslogic jvrapp krdprog wesmare chenxizhang aitips lgs arijit1000 pidgoreckiynestor akhilpenugonda hasarinduperera ardevsrc kurtseifried miblue119 michaelmartinez tianjunz adjdunn adeilozo erwinrnasution cooltunes goroman skywalker19 mediapreneur oebirch thinkflow19 caceager carlosplanchon 1openwindow tbnag nagpalnitesh chvg1968 godagoo celeritas78 monk1337 pmrprdc eric-seekas laisee chunmk gviruete belkmouf riveroma timegated shiroiushi jeancodeunmax kshitiz305 igorverse harishmohanraj fabianolothor knightcn1983 baitphish

learn_prompting's Issues

More on calibration

Add this paper's calibration strategy: https://arxiv.org/pdf/2205.03401.pdf

HCI/design research for prompt engineering UI

New Topic Request: https://promptable.ai/projects/default/workspace

New Topic Request: weaknesses

Frequent hallucinations: expressing “facts” that aren’t true in plausible ways. Always fact check its answers.
Suggestible. Avoid leading questions because they can frequently cause hallucinations.
Often tripped up by large number math. Use a calculator, instead, or consider giving GPTs a programming language interpreter if it needs to perform reliable logic and calculations.
GPTs are great at role playing, but don’t seem to hold strong opinions of their own. They don’t appear to actually care about anything, including giving you grounded, factual, or correct responses.
GPTs often crack jokes and use sarcasm, but not well, and it’s not always obvious the’ve tried to do it unless they explicitly tell you. Sometimes it is obvious, though: “Could a cat pilot a rocket?” GPT-3: “The purr of the engine would be very distracting.”
Of course, also mention pretraining, no long-term memory after training cutoff, occasional tendency to paste media that is not original (particularly Codex - there is a filter for this in the Copilot settings)….

Credit: Eric Elliott

Resolve Dyno related monkey patch

Would be great to have a page on the website that shows the change log copied from github version releases.
Plan here would be to autogenerate a change log from https://github.com/trigaten/Learn_Prompting/releases

Docusaurus does this and uses lerna and a custom parser which is in the repository: https://github.com/facebook/docusaurus/tree/main/website/src/plugins/changelog

Page on different prompt engineering products

New Topic Request: Least to most prompting

https://arxiv.org/abs/2205.10625

https://promptshake.com/?ref=producthunt

Automatic prompt engineering (APE)

New Topic Request: https://getimg.ai/guides

What topic would you be interested in seeing an article/chapter about?

Any additional info you have about it:

Clarification

Sorry, what can I clarify?
What is a labelspace?

How to: fake GPT3 startup

Walk through making like fake gpt based startup
“Ok so we want to create a website that writes marketing content garbage for you! Let’s start engineering a prompt”

Add dream studio, https://www.patience.ai to IDE list

Make Credits Page

Add SD generated images to the top of articles

newsletter~

tune in for the latest research on PE (newsletters, arXiv search terms, twitter accounts)

Survey of PE IDEs

Clarification of formulas in Calibration

In this page there are a number of formulas given, but it could really use some examples to further clarify those formulas.

Some examples:

Given a 0.9/0.1 distribution for "neutral", what will 0.95 or 0.5 be mapped to?
- There's two proposed equations, both of which map 0.9/0.1 to 0.5/0.5, but I'm guessing they'll map other values differently
Because this is specifically mapping two number to two different numbers, can the two equations be mapped to simpler equations?
A better example than "N/A" would be nice. text-davinci-003 insists the sentiment is also N/A (which is correct, but makes for a poor demonstration)
I'm not clear how to get those probabilities from GPT-3.

I tried using "full spectrum" probabilities in the GPT-3 playground in order to see its probabilities, but I get this and I'm not sure how to map it to this calibration:

Write content on prompt injection

Small extension to "Solve Discussion Questions: Many discussion questions are not effective for prompting"

In this section you have the reader theoretically rewrite the question to make it better. This itself is something you can ask the LLM to do, and I thought it might be a good example to embed in that section.

I was able to do:

Rephrase this as an essay description:
"The Civil War a conflict over expansion? Agree our Disagree and why?"

> "Write an essay about whether or not the Civil War was a conflict over expansion. In your essay, take a stance on the issue and provide evidence and reasoning to support your argument."

With a more modest prompt ("Rephrase this as a class discussion topic:") it fixed the grammar and made slight improvements, but probably not as much as desired.

(There is a typo "Agree our Disagree" ... not sure if that was intentional, but it's something GPT can fix so I kind of like having it in there)

General description
Example usage
results, ablations, notes, etc.

The single most important ASPECT of this course is your feedback

lengthen modules
more background/theory
examples help
examples on every page would be good