The Business Thinker's Guide to ChatGPT
Jun 30, 2023
There are many people using ChatGPT casually for help with certain professional and personal tasks. While ChatGPT is an amazing tool, and there are infinite posts about how to make ChatGPT work better, there still seems to be a gap in most people’s understanding of how it works.
This is not another “top 10 best prompts to change your life” post (no mind-blown emojis here), but rather a guide to help the non-technical user understand and optimize their use of ChatGPT.
Find an example of our master prompt in this post. We use versions of this master prompt in ChatGPT before generating any content (decks, posts, emails, website copy, code, etc.).
GPT is the actual LLM (Large Language Model).
GPT stands for Generative Pretrained Transformer.
ChatGPT is the application/interface on top of GPT to interact with the model without an. API. Note this guide only focuses on the ChatGPT application.
It is our recommendation that everyone who uses ChatGPT creates a “master prompt,” which provides better background and context on you and/or your company.
Find the paster prompt for our 🧬🌍 GenWorlds platform posted here
Prior to anything we do in ChatGPT, we use this master prompt to set the context. See a quick demo below where we use our master prompt to quickly create content around a product launch:
Feel free to use this ChatGPT session as your own initial master prompt to help you learn more about ChatGPT and Generative AI: https://chat.openai.com/share/325917c7-fde4-4c10-a03b-67fa779f5fd7
Context windows generally represent how many words GPT can effectively remember and effectively generate responses from.
Context windows = inputs (user prompt) + outputs (ChatGPT response) for the entire chat instance.
Context windows are made up of tokens. GPT models process text using tokens, which are common sequences of characters found in text. The models understand the statistical relationships between these tokens and excel at producing the next token in a sequence of tokens.
Word count ~= Tokens * 75%
GPT 4 Context window is ~6k words (~8k tokens).
GPT 3.5 (free version) Context window is ~3k words (~4k tokens).
As you go over the token limit, you will mostly receive output based on your latest inputs.
While it’s possible to exceed the token limits in a conversation with ChatGPT, doing so might lead to a loss of context from earlier prompts. This is because ChatGPT begins to summarize previous parts of the conversation to manage its memory capacity, resulting in a reduction of detail from the earlier portions of the chat instance.
Best Practices, Tips, and Tricks
Provide the most context with the least amount of words. No need to be polite; save your words, use them deliberately.
Be mindful of the context window and think about editing earlier prompts rather than doing a new prompt if you want to stay in the same chat session.
Spaces and tabs do take up tokens, so avoid unnecessary spaces (e.g., two spaces after a period or a space after a bullet before the first word)
ChatGPT is good at making content more or less concise
ChatGPT is good at providing easy-to-digest examples of complex material
GPT is not good for calculations; it’s typically quite bad with numbers. If you want to do calculations, connect the Wolfram|Alpha plugin.
ChatGPT makes mistakes (“hallucinations”); always review and verify its outputs. It’s not a search engine.
ChatGPT is good at providing outputs in specific formats (e.g., Twitter posts, LinkedIn posts, sales emails, CSV, etc.); all you have to do is ask it to provide the answer in a specific format.
Using all caps could change GPT’s understanding of the input
Use different chat sessions for different topics so the context is not so broad. The more narrow the context, the better your results will be.
For every new conversation, add a master prompt and end it with: Respond “Acknowledged.” Once ChatGPT responds “Acknowledged,” then ask it to do whatever you want. This will help avoid ChatGPT responding with a summary of what you just told it, thus wasting words.
GPT understands markdown syntax well
# for title
## for subtitle
- (or any Word format) for bullets
You can now share links from your ChatGPT instances with a team
This is valuable so you don’t have to waste time re-prompting and waiting for ChatGPT to re-execute; the convo links allow you to leverage what’s already been stored/pre-computed in GPT
We recommend keeping a master convo link for each function/internal use case
Note the shared chat instance is not like a shared doc; another user will not see any additional prompts in your ChatGPT instance
Get in Touch
We hope you found this helpful. We have been working with businesses across many verticals uncovering use cases for our AI Agents to execute (VC funds, pharma, media, aviation, research, marketing agency, dev shops, legal, insurance, etc.). For now, think of our AI Agents as entry-level employees which can coordinate to execute tasks and achieve common goals.
If you are currently thinking about how to use Generative AI in your business or if you have any questions on the topic, please reach out to email@example.com.
Be aware that all data is accessible by OpenAI, the company which owns ChatGPT.
OpenAI will likely be using all of the data it aggregates to train and retrain its models.
There is still plenty that we (and even OpenAI) do not understand about ChatGPT and LLMs in general. Additionally, OpenAI is rapidly updating its products and policies. As such, please understand that certain points in this guide could become out of date, but we will do our best to refresh as often as possible.
Learn More About 🧬🌍GenWorlds
At Yeager.ai, we are on a mission to enhance the quality of life through the power of Generative AI. Our goal is to eliminate the burdensome aspects of work by making GenAI reliable and easily accessible. By doing so, we foster a conducive environment for learning, innovation, and decision-making, propelling technological advancement.