10 Levels of ChatGPT Prompting: Beginner to Award Winning

I've spent countless hours researching and testing different prompts to see what gets the best results in this video I've distilled it all down to 10 levels prompt engineering we will go from the basics all the way to the expert techniques that recently won Singapore's prompt engineering competition so let's start now level one just telling you what you want at this level you're just basically mindlessly telling chat gbt what you want out of it no real thought behind it sometimes you get good with results sometimes you don't this is me just basically telling chat GPT to summarize a Wikipedia article a pretty easy task and it does fine but there are a lot of ways we can improve on this level two basic formatting small things like formatting can make a surprisingly big difference for instance just adding some dashes to separate your prompt can greatly help chat GPT understand the different sections of your prompt right now it might not make that big of a difference but as our prompts get more complex it's huge other simple formatting esque things that make a big difference are being nice avoiding negatives and even begging and pleading some Research indicates that being polite in your prompts helps large language models with their accuracy even if being nice didn't improve responses I'd probably still do it anyway just cuz I don't want to get in the habit of being a demanding jerk no one's quite sure why but large language models seem to perform better when you tell it to do something rather than to not do something something my pet theory is that it's just like our brains for instance don't think of an elephant and you thought of an elephant I use this one sparingly but it's been shown that appealing to intense emotions can improve the responses of LGE language models some examples of this are saying something like this is really important to my job do you get this right or I'm going to show your response to my grandma and I want her to be proud of me so please get it right similarly you can threat in chat GPT this works too personally I don't do it just in case AI takes over the world I want to be on the friendly side of the robots level three Focus requests here's where we really start to see an uptick and quality responses level three is all about being clear and focused with what you want out of the chatbot for instance you don't want to say make the response pretty instead write please respond with heading subheadings and tables for instance or maybe you only want certain information tell it what you do and don't want in the response so this right here is an example of a bad request there so much ambiguity in it and honestly this one so ingrained in me I kind of had a hard time making a bad request you see I told it to do columns right here and that guided it towards making a table so take a look at this second Super specific here is some data put it in a table I only want the columns class CL level name major order it by class level and then at the very end tell me what Carl's major is boom got exactly what I wanted level four give examples this is our first somewhat Advanced prompting technique and if you've heard the term fuse shot this is what they're talking about basically you want to give chat GPT some example inputs and some example outs let's take a look at an example here I give a quick description of what I want extract the info from the text of this LinkedIn page as an example I'm going to use my CEO's LinkedIn page and I'm going to put the example output in the format that I want lastly I'm going to copy and paste a LinkedIn page here's mine for example and without any specifics in just one example it responds perfectly and in the right format if you have supervision you may have noticed it missed my last job where I briefly worked in marketing there are a handful of techniques to make sure it doesn't miss any information like that but one of the easiest is level five self-reflection hey chat gbt did you miss anything this one is almost too easy large language models are statistically better at evaluating than generating so by asking this question you're really playing to chat tpt's strengths level six nail the system prom this is a big one I'm not going to dive too deep into it cuz I'm working on a video right now that's all about the system prompt but the gist of it is a special set of instructions that guides chat GPT to answer how you want it to answer I use the system prompt a little bit differently than most people I've seen and I'll I'll leave a link to the format in the description below but the gist of it is I try to give as much context to chat gbt about me and what I want out of it I tell that I'm a software engineer I prefer X programming language and I prefer short answers with follow-up questions rather than huge walls of text etc etc so make sure to add this in it'll only Take 5 minutes and it will improve all the responses from there on out level seven use personas say you have a simple riddle I see a glass door with push written on it in Mirror writing should I push or pull the door most of us intuitively know that because it's in Mirror writing we should probably pull the door chat GPT however confidently gets this wrong well sometimes it's as easy as telling chat GPT to just be an expert in whatever we're asking about a few studies I looked at showed that just by using personas large language models improved 6 to 20% in the accuracy of their response and it just blows my mind that you can get that much better results just by saying act as whatever level eight Chain of Thought another way to get chat gbt to handle tough problems is to ask it to explain its thought process I use this in almost all of my prompted works so well there has been a lot of research around this one it seems like the most bang for your buck is just to add the phrase let's think step by step I find telling chat gbt to look at all angles of the problem also provides better outcomes level nine self- prompting so you've learned all these techniques and you're still with me now throw it all out it turns out large language models are better at prompting themselves than humans are so we just need to ask chat gbt to make a prompt to get us the answer that we want here you can see it write a very detailed prompt on how to solve a given riddle I'm just going to copy this open a new chat paste it in replace the riddle and see how it does it did some awesome reasoning and then got it right level 10 co-star framework this is the final dots there are many ways to organize a prompt but in my opinion the best is co-star this is the exact framework Sheil L used to win Singapore's gbt 4 prompt engineering company here's how it works each letter of co-star stands for a specific part of the prompt let's go through each one C stands for context this is where you give it any relevant background information for instance information about you or the task that you want it to do O is for objective this is where you give Crystal Clear instructions on what you want chat gbt to do do we learned this way back in level three this is for style in this section we tell Chad gbt the style of writing that we want could be something silly like we want it in the style of a Snoop Dog rap or something like write it in the style of top CEOs T is tone what tone do you want the response in funny emotional threatening you decide a is for audience this is where we tell Chad gbt who the audience is for instance if the audience is 5-year-olds it will be wildly different than if the audience is World leading physicists and lastly R response here is where we tell Chad gbt the format of the response that we want do we want a detailed report do we want in a table do we want a fancy programming format like Json or just one giant wall of text here's where you put it let's take a look at an example here's a basic request for a Facebook post advertising a new flying Magic Carpet basic request basic response this is wordy ugly and most certainly won't resonate with the target audience okay let's try co-star first I give it the context that I run a magic carpet business then I tell it the objective is to write a Facebook post to get people to buy it I tell it the style I want basically to just copy successful companies I tell it to have an elegant persuasive tone I say the target audience are people in their 30s and lastly I tell the pat of the response four sentences no hashtags and put in some emojis ah much better basically using co-star guides you'd add Everything Chad gbt needs to answer exactly how you want and that's a w I'm coming out with lots more videos on fantastic ways to use AI so if you like this video you might be interested in this one

Channel: Patrick Storm

Views: 30,005

Keywords: prompt engineering, prompt techniques, chain of thought, chat gpt prompts, chatgpt prompt engineering, llm prompts, chatgpt, chat gpt

Id: 2djqKsRXt_Q

Length: 9min 31sec (571 seconds)

Published: Thu May 02 2024