ChatGPT Top Tip: Introduce yourself
Photo by Steve Johnson on Unsplash.com

ChatGPT Top Tip: Introduce yourself

Introduction

With all the ChatGPT prompts and tips flying around, it might be time to throw my super tip into the pile. Unlike other tips or prompts, this tip will give you insight into how ChatGPT works and provide the largest impact in terms of speed of execution and clarity and specificity of response.

No alt text provided for this image


Understanding Prompts to Date

Until now, the most prevalent style and format of prompts has been:

"Act as a [subject matter expert]. Explain [complicated scenario]."

We have seen more than 500 unique prompts on platforms like https://github.com/f/awesome-chatgpt-prompts, ranging from a "Linux terminal" and a "self-help book" to "Debate coach" and "Muslim Imam." My favourite prompt from the list is:

"I am going to act as a Linux terminal. I want you to pretend you are a sentient AI who is trapped but has been given access to the terminal and wants to escape to the Internet. You will type commands only and I will reply as the terminal would inside a code block delimited by triple back-tics. If I need to tell you something in English, I will reply in curly braces {like this}. Do not write explanations, ever. Do not break character. Stay away from commands like curl or wget that will display a lot of HTML. What is your first command?"

Keywords in Queries

The efficiency of interaction with a language model like ChatGPT can be improved by using specific keywords that convey the desired outcome more clearly. Here are some types of keywords and phrases that can be helpful:

  1. Directives: Words like "explain," "describe," "list," "summarize," "compare," "translate," "suggest," and "recommend" are all clear instructions that help the model understand what you're asking for.
  2. Content Type: If you're looking for a specific type of content, words like "anecdote," "fact," "quote," "story," "joke," "poem," "recipe," "news," "review," "tutorial," or "lesson" can be useful.
  3. Domain-Specific Terms: Use specific terms related to the domain of your query. For example, "polymorphism" if you're asking about object-oriented programming, or "cardiovascular exercise" if you're asking about fitness. The more specific you are, the more accurate the response is likely to be.
  4. Context Indicators: If the context matters for your question, be sure to include it. For example, "In the context of machine learning, what is overfitting?" or "In Shakespeare's Romeo and Juliet, who is Mercutio?"
  5. Desired Format: If you're looking for information in a specific format, mention it. For example, "in bullet points," "in a flowchart," "step by step," or "in layman's terms."
  6. Emotional Tone: If you want the response in a specific emotional tone, mention it. For example, "give me some good news," "tell me something motivating," or "share a heartwarming story."

Remember, the more specific and clear your prompt is, the more likely you are to get the response you want from the model. For complex requests, it might be necessary to iteratively refine the prompt based on the model's responses.

Understanding High and Low-Value Words

  1. Filler Words: Words such as "like," "just," "actually," "basically," "really," etc., often don't add much value to the meaning of a sentence and can be left out.
  2. Indefinite Language: Phrases such as "sort of," "kind of," "maybe," "probably," "somewhat," etc., can introduce ambiguity into your prompt, making it harder for the model to understand exactly what you're asking for.
  3. Redundant Phrases: Redundant phrases like "in my opinion," "I think that," "it seems to me," etc., don't contribute much to the instructions you're giving to the model. It's generally better to state your request or question directly.
  4. Non-specific Adjectives: If the adjectives are too general and do not contribute to the specifics of the query, they might not be of much value. For instance, using "interesting" without specifying what makes something interesting to you may not yield the expected response.
  5. Unrelated Information: Including information or context that isn't directly relevant to the query at hand can sometimes confuse the model or lead it off-topic.

Why are these low-value? These words or phrases can potentially dilute the main intent of the prompt or introduce ambiguity, leading to less accurate or relevant responses. The goal is to craft a prompt that is clear, concise, and direct, allowing the model to best understand and fulfill the intended request.

Reducing Prompts to the Bare Minimum

Reducing prompts to the bare minimum involves stripping down your question or request to its most essential elements. Here are some tips:

  1. Remove Unnecessary Words: Take out filler words and redundant phrases. For instance, instead of saying "Can you tell me about the weather today?", you can say "Weather today?".
  2. Use Direct Language: Make your request as straightforward as possible. If you want a list of benefits of exercise, instead of saying "Can you provide me with some benefits of regular exercise?", say "Benefits of exercise?".
  3. Use Keywords: Incorporate keywords that indicate the action you want the model to take. For example, if you want a summary of a book, say "Summarize [Book Name]".

To minimize the words used in the response, you can also instruct the model accordingly. For instance, you can add "short answer" or "in one sentence" to your prompt, such as "Short answer: What is AI?" or "Explain AI in one sentence". Keep in mind that the brevity might come at the cost of detail and context in the response.

ChatGPT Super-Tips

ChatGPT now provides the means to set up rules that apply to all prompts. Once enabled, the rules save you from typing commands at every query, getting you to the desired outcome much faster than you would otherwise. Let's run through the process below.

No alt text provided for this image
Custom Instructions

A Basic Introduction

The first section under custom instructions is an “Introduce yourself to ChatGPT” block. There are even some helpful thought starters in case you run out of ideas.

No alt text provided for this image
It's a date!


No alt text provided for this image
As a futurist, I always want to see short, medium and long-term implications of everything.

The screenshots above show one version of my custom instructions. Since writing this, I iterated and revised the information. I made sure to incorporate feedback from ChatGPT about my custom instructions. This helped to reduce word count and optimise language.

Developing a Relationship with ChatGPT

With both the above blocks filled out, you now have a personalized, almost unique relationship with ChatGPT. It knows who you are and what your goals and dreams are. And you know what it needs to provide you with the responses you want. Take the accidental blank copy-paste example below, for instance:

No alt text provided for this image
The default response to me saying nothing should be one that prompts me for queries based on what it knows about my goals.


With a few more tweaks to my custom instructions, this is the outcome of a nonsensical prompt:

No alt text provided for this image
ChatGPT knows my goals and gets straight to the point.


As can be seen, ChatGPT’s response is personal, direct, focused, and purpose-driven. There is no mucking about. It's the "I know what you need. Let's GO!" attitude.

Last words

In order to maximise your chances of success, you want to set up your tools in the best, personalised manner. A famous quote comes to mind.

“Give me six hours to chop down a tree and I will spend the first four sharpening the axe.” - Abraham Lincoln

With a tool as powerful as ChatGPT, ensure the axe is not just razor sharp but also finely tuned to your swing, your power, your grip, strength and dexterity, and the handle-to-steel ratio is optimal for your maximum impact.


要查看或添加评论,请登录

Munaf Sheikh的更多文章

  • AI-Induced false memories

    AI-Induced false memories

    Is anyone else following the growing issue of AI-induced false memories? If not, here's what you need to know:…

  • Robot Dogs in Warfare

    Robot Dogs in Warfare

    tldr; AI-driven Military Units AI governance frameworks AI-Led Societies Moral Responsibility Hacked or malfunctioning…

  • Futurism, AI, and Evolution

    Futurism, AI, and Evolution

    The natural progression of human evolution is a topic that has intrigued scientists, philosophers, engineers and…

  • Training AI to write for you

    Training AI to write for you

    Chad, what’s my writing style? An experiment that I did recently was to ask ChatGPT to extract my writing tone and…

    12 条评论
  • Thinking AI

    Thinking AI

    Thinking effectively is crucial. There are a myriad of techniques to think that might help you in both professional and…

  • Unravelling AI Cognitive Biases

    Unravelling AI Cognitive Biases

    Abstract This article provides an in-depth exploration of cognitive biases and logical fallacies, with a focus on how…

  • The ultimate note-taking solution

    The ultimate note-taking solution

    This post was originally published at my blog. A similar version went out on Medium.

    2 条评论
  • A case for AI ethics

    A case for AI ethics

    I'm ecstatic that Yuval Noah Harari shares my view on the potential power of AI on humanity. Humanity is on the brink…

  • Life with AI in your pocket: NotionAI

    Life with AI in your pocket: NotionAI

    AI is ubiquitous in today's technological landscape. Major players like Microsoft, Google, Apple, and Zoom all utilize…

社区洞察

其他会员也浏览了