✍️ Train ChatGPT to write like you

PLUS: tell me the customer painpoints from this call

Author

Martin Slaney
March 1st, 2023

Hey there 👋

In today’s Product Prompts you’ll learn:

  • How to train ChatGPT to write like you

  • How to find any company-documented info you need using Slack

  • How to quickly ask questions based on a customer research call

  • What the heck embeddings are

  • and more 👇

💬 Prompt Deep-dive: Train ChatGPT to write like you

I come across a ton of use cases in my day job where I need to write something, which ChatGPT can help me with, but I have had to spend a fair chunk of time making it sound more like me. Luckily it’s possible to train ChatGPT to do this. Whilst I haven’t actuallydone this for this newsletter yet (honest!), here’s how you would go about it 👇

Please learn my writing style from the following content:

And then I pasted in my recent AI Audit on Monzo - or most of it. ChatGPT has a limit of 4,096 “tokens” that can be used in total between a prompt and the output. 1,000 tokens is roughly equal to 700 words in my experience. So your max prompt size is around 1400 words.

Here’s the output, whereby I’m given an interpretation of my style:

Based on the provided content, your writing style appears to be conversational, informative, and easy to read. You use casual language, such as "Hey there" and "🤪", to engage with the reader and create a friendly tone. Additionally, you provide clear headings and bullet points to organize information and make it easy to follow.

You also have a clear understanding of the product strategy and goals of the company you are discussing. You use this knowledge to propose potential AI features that align with the company's mission and values. You provide detailed explanations of these features, including their benefits, risks, and potential implementation strategies.

Overall, your writing style is well-suited for discussing complex topics in an approachable manner, and you effectively convey your ideas and insights to the reader.

Which I think is pretty fair, if rather complimentary. I’ll take it 😊

You can follow this up with additional content, to further train the model and refine its understanding of your style.

OK so now to see if we can generate a blog in my style:

In the same writing style, write a blog article on the benefits of using OpenAI's GPT Playground versus ChatGPT, the conversational interface which uses GPT3.5. Use emojis and analogies.

Hey there! 👋

Have you ever wanted to get a taste of what it's like to work with a language model like GPT-3? Well, look no further! OpenAI's GPT Playground and ChatGPT are two tools that let you interact with GPT models, but they each have their own unique benefits.

Think of GPT Playground as a sandbox for playing with GPT-3, where you can experiment with different prompts and see how the model responds. It's like a playground where you can try out different toys and games to see which ones you like best. 🎉

On the other hand, ChatGPT is more like a friend who you can have a conversation with. It's like having a chatbot that uses GPT-3.5 to respond to your messages. You can ask it questions, tell it jokes, or just chat about your day. It's like having a virtual buddy to hang out with. 🤖

So what are the benefits of using GPT Playground versus ChatGPT? Let's take a closer look.

GPT Playground allows you to experiment with different prompts and see how GPT-3 responds. This can be helpful if you're trying to generate text for a specific purpose, such as writing an article or crafting a marketing message. You can try out different prompts and see which ones yield the best results. It's like trying out different ingredients in a recipe to see which ones taste the best. 🍲

ChatGPT, on the other hand, allows you to have a conversation with GPT-3.5. This can be helpful if you want to see how the model responds to specific questions or topics. You can ask it questions and get answers in real-time, which can be a great way to explore different ideas and topics. It's like having a personal assistant who can help you with anything you need. 💼

Another benefit of GPT Playground is that it allows you to see how GPT-3 responds to different parameters and settings. You can adjust the settings and see how the model's output changes. This can be helpful if you're trying to fine-tune the model for a specific use case. It's like adjusting the settings on a camera to get the perfect shot. 📷

ChatGPT, on the other hand, allows you to have a more personal and interactive experience with GPT-3.5. You can ask it questions and get answers in real-time, which can be a great way to learn and explore different topics. It's like having a tutor who can help you with your homework. 📚

In conclusion, both GPT Playground and ChatGPT are great tools for interacting with GPT models, but they each have their own unique benefits. GPT Playground is great for experimenting with different prompts and settings, while ChatGPT is great for having a conversation and getting real-time answers. So why not try both and see which one you like best? Happy experimenting! 🤗

Not bad at all! If you want specific sections, this could be included in the prompt, and if you want a longer article you can get it to write one section at a time.

Pro tip: you’ve got the writing style worked out, you can just return to the same chat every time, and use your prompts to write whatever content or format you need.

🛠️ Cool new tools

Metal - AI Developer tools
A developer platform that makes it easy for engineers to use embeddings* in their applications. It provides APIs, models, infrastructure, and more for embedding generation and indexing. It was launched by Y Combinator in 2022.

*Embeddings are like lists of numbers that show how similar or different things are from each other. For example, if you have two words like “cat” and “dog”, you can use embeddings to see how close they are in meaning. The closer the numbers are, the more related the words are. The farther the numbers are, the less related the words are.

Embeddings are helpful for machines to understand and work with different kinds of data, such as text, images, sound, etc. They can help machines find patterns and meanings in data that humans can easily see or hear but machines cannot. For example, embeddings can help machines recognize faces in photos or emotions in voices.

Reword - Writing
I’ve tried a lot of GPT-powered writing tools, generally with at best moderate results, but Reword looks to be right up there in terms of helping you write well. The really cool bit is using a Google Search Console connector to detect and track your existing articles - this means your AI never stops learning, as it’s progressively trained on your detected articles.

Dashworks - Internal Knowledge Management
An AI search assistant for work. Find any file you need from different apps and places. Dashworks is like a smart assistant that can answer your questions about anything related to your work, like your product, customers, policies, and so on. Works with all kinds of tools like Jira, GitHub, Confluence, Dropbox, and today they’ve announced a Slack integration 🤩

Gong - Customer Research and Sales Intel
Call summariser & QA for Sales/Product Research call. Get a one-click summary, and quiz the call using natural language. Saves a ton of time when referencing this information in follow up emails or getting context on a call.

👀 Look out for Sunday’s AI Audit.

❝

AI Audit - an analysis of AI growth opportunities for a well-known product.

Each week, in addition to the weekday newsletter like this one, I put myself in the shoes of a product team, ideate the generative AI opportunities and call out what I think would add the most value to the customers.

Thanks for reading, and happy prompting,

Martin, Chief Prompt Officer

Reply

or to participate.