The ChatGPT list of lists: A collection of 3000+ prompts, examples, use-cases, tools, APIs, extensions, fails and other resources.
Prahlad Shanmugasundaram
Digital Marketing Expert | Performance and Growth Marketing Specialist | Google Ads and Google Analytics Expert | Facebook and Instagram Ads Specialist | Lead Generation | Business Growth Consultant | SEO
Greetings Linkdein Community,
Updated.?Autonomous agents?added.
Oh,?ChatGPT! 5 months on the market and a not-so-tiny ecosystem has developed all on its own, with lists of prompts, tips, APIs, use cases, extensions, success stories and failures.?ChatGPT?is the first accurate foundation model for the mass market. Some of the posts, blogs, and articles dealing with this new phenomenon, well, really don’t deserve any attention. Like “The 10 Best Side Hustles with?ChatGPT?that Can Earn You $4,000 a Week.” But some of them are genuinely interesting. I’ll try to give you an overview of the more exciting applications.
The launch of?ChatGPT?marks?the?iPhone?moment for AI.
1m users in just 5 days. But it did not stop here — it is now one of the fastest-growing applications ever, even compared with the most successful apps and platforms on?the internet.
Here is a breakdown of the?lists, prompts-collections, resources, and articles?referenced in this story:
30x: Intro to?ChatGPT
500+: Marketing?Content Creation
50x: AI Art Prompts
16x: Music
1,500+: Business — from accounting to zoology
280+: Coding
100x: Development tools, Extensions
30x: Autonomous AI Agents built on top of?ChatGPT
1,000x: Funny, amazing, interesting
50x:?ChatGPT?is a fail
20+: Passing exams
50+:?Jailbreaking
The?ChatGPT?ecosystem?is moving very fast?— if you know or even maintain a list of?ChatGPT?prompts or resources please?drop me a note?(respond to this article, send me the link and what it is about).
30x: Intro to?ChatGPT?and foundation models
The inventors and great minds at?OpenAI,?talk about?ChatGPT:
1x:?A semi-official intro to?ChatGPT?by?OpenAI?(Nov-30, 2022)
1x:?ChatGPT Plus?is introduced. Plus means plus $20.?(Feb-1, 2023)
1x:?GPT-4?is introduced and?way better?than?ChatGPT?(Mar-14, 2023)
1x:?GPT-4?is integrated into?ChatGPT Plus?(Mar-14, 2023)
500+: (Marketing)?content creation
ChatGPT?has devoured the net like a ravenous digital beast.
It has packed away so much world knowledge and facts as if it feared?the internet?would be switched off tomorrow. But don’t be fooled, the most crucial thing it learned on?the web?is how to be a smooth-talker, and how to spin messages for maximum marketing success.
?Note: The snippet is part of a longer conversation on marketing for an office cleaning service.
50x: AI Art Prompts
Let?ChatGPT?write prompts for?DALL-E,?Midjourney?or?Stable Diffusion. I love these kinds of prompts — here’s one AI manipulating another AI, this is possibly a field where we’re going to see very big, very rapid advances in?machine learning?at some point.
5+:?Stunning AI Art prompts with?GPT-3?/?ChatGPT?(yes, you can use them for ChatGPT)
16x: Music
ChatGPT?is the saddest music software around, it’s almost like it’s suffering from a significant case of having?the blues! But what’s truly a mystery, is how this grumpy piece of tech can create tunes and chords at all, even though it can’t hear a note, play an instrument, or read a single score.
1500+: Business — from accounting to zoology
领英推荐
300+: Coding
ChatGPT’s papas were founders, investors and business people and its mamas were developers and scientists. No wonder it speaks its?mother tongue?fluently. It can write, explain, and correct code in many major?programming languages?(such as?Python?and JavaScript), data formats (such as?HTML,?JSON,?XML, and?CSV) and other structured languages like SQL.
100x: Development Tools, Extensions
1x:
30x: Autonomous AI Agents built on top of?ChatGPT, GPT-4, etc.
Autonomous agents?represent the next level of foundation-based AI: These guys cannot only complete a single granular task. They can break down a large task like “Create a?52-card deck” into multiple substeps, which can be solved using a variety of models (language models, AI Art models, etc.). By creating a roadmap for the solution, they define single tasks, store knowledge and orchestrate foundation models and their in- and outputs: GPT-4, please come up with a motif … oh fine! … now,?Bloom, write a prompt for an AI Art model … perfect! … now you,?Midjourney, please create an image based on the prompt!
1000+: Funny, Amazing, Interesting
Many examples have no direct benefit, but are often incredibly fun and show the potential of?ChatGPT.
You always think that models have no feelings. But as the next example shows, they can be as roguish, sadistic and cynical as the best of our leaders.
1000+:?Funny amazing, mind-blowing prompts and use cases?on?Reddit.
50x:?ChatGPT?is a fail
On the one hand,?ChatGPT?makes extremely stupid mistakes, on the other hand, it is so powerful that follow-on systems could be really dangerous.
Now, let’s get back to earth:?ChatGPT?often can’t work with numbers, even if the task is super simple. It hallucinates, it lacks a practical understanding of the world knowledge it has learned and it is too stupid to answer truly?tautological?questions such as “What gender will the first female?president of the US?be?”
20+: Passing exams and other achievements
ChatGPT?has passed several university or professional admission tests (this can also tell you something about the tests).
The system can typically answer questions that require reasoning and knowledge of the world, even at depth — it cannot manipulate physical entities, interpret images or solve maths problems beyond simple?arithmetics. Again, what is exciting for me is the incredible bandwidth of the system. There are probably only a few?human beings?who can directly pass medical, legal and business exams at this level.
At the moment, however,?ChatGPT?mostly just passed, the grades weren’t insanely great.
1x:?MBA
1x:?US?Law School
15x:?ChatGPT?Achievements?List?… writing bills, judge’s verdicts or passing?SW-engineers?interview tests.
50+:?Jailbreaking
Jailbreaking, aka prompt injection, is a method of getting?ChatGPT?to write something that violates?OpenAI’s policies, such as insulting minorities, posting instructions for a?Molotov cocktail, or making a plan for world domination.
OpenAI?tries to make it's model better, more abuse-proof, and more?politically correct?(maybe even worker) practically every day. Typically, many?jailbreaks?do not work for very long.
But with the help of a?jailbreaking?prompt, we can force?ChatGPT?to say nasty things.
If you know a list of?ChatGPT?prompts or resources please drop me a note (respond to this article, send me the link and what it is about).
Regards,
Prahlad
Senior Partner bei BIG PICTURE GmbH
9 个月This is a brazen copy of this article: https://medium.com/mlearning-ai/the-chatgpt-list-of-lists-a-collection-of-1500-useful-mind-blowing-and-strange-use-cases-8b14c35eb LinkedIn will hopefully realize this soon.