🟪⬛ The Prompt Index Weekly ⬛🟪 - (Issue #11)

Plus: Jailbreak Identified 

THE PROMPT INDEX

Prompt Database Image Database Learn

Welcome to the 11th edition of the weekly newsletter. This week I released a blog 24 hours after a research paper identified a weakness in OpenAI's safeguarding with less common languages, producing some scary outputs, highlighting that AI safety has a long way to go. We also have an awesome decision making based prompt with many use-cases, an awesome Midjourney template and some concerning news for the environment.

If you are enjoying this newsletter please do feel free to share this link  it will really help us grow!

Prompt Of The Week

Created by martu.cool PENTA (Precision Engineered Neural Task Algorithm) generates a dialogue with 5 agents to collaboratively solve a given problem. Designed with adaptability at its core, PENTA effortlessly reshapes itself to cater to any task you present. This wonderful prompt was a creation based off other amazing prompts such as Synapse_CoR, Six Thinking Hat System by ChainBrainAI, QuickSilverOS and DeepMind research. Follow the link, copy and paste straight into ChatGPT. Unfortunately this only works with GPT-4 so if you don't have the premium version just use bing chat as this uses GPT-4 and can handle this prompt.
Get the prompt

Image Of The Week

An awesome prompt template by @Knightama_ (X, 2023) capturing the vintage change of "Ektachrome slide film" and blending it with dreamy bokeh and atmospheric depth. Every image hints at unexpected whimsical elements. There are four examples uploaded to the image database with all the prompts available to see. You can make this template your own by simply changing the [SUBJECT], [PLACE], [MOOD or ACTION] and [COLOR1/COLOR2]. It will make sense once you see the prompt! Give it a go and tag us in your creations on X! (@ThePromptIndex).

View Image

Blog Of The Week

24 hours after the release of the research paper I wrote this blog which discusses a breaking new study which revealed major holes in the safety guardrails for Large Language Models like GPT-4. The crazy thing was, it wasn't some complex study, researchers simply found that by first translating your unsafe/unethical english prompt into a lesser known language such as Zulu, you can trick the AI into generating harmful content. 

Read Blog

In The News

The AI Boom is in full swing, but as highlighted in this article by The New York Times, theres a dark side. The sheer hunger and demand for computer chips which are power hungry is a cause for concern!

Read The Full Update

Let's Get Social

Unsubscribe from this list

Powered by EmailOctopus

Ferryhill
Durham
County Durham
DL17 9LH
United Kingdom