What are we doing here?

There are a number of things you can do with ChatGPT specifically to make it more usable, the same may be true in the future for things like Claude and the rest of the generative text models, but today we’re going to focus on ChatGPT.

I. Sign up for ChatGPT Plus

Yes, like most platforms that start out free, ChatGPT went to a freemium model, they have lots of GPU expenses to pay so we shouldn’t hold it against them. It is only $20/mo. and it unlocks all of the cool features, so it’s worth the 3-4 cups of coffee to at least try it for a month, if you don't like it, you can always downgrade.

II. Try Out GPT-4 (if you haven’t already)

Specifically use the same prompt on GPT-3.5 and GPT-4 and compare the results in both speed and output quality.

We’ll use a simple prompt that is specific enough to give the two different models something to work with:

Prompt: write an employee offer letter for a senior full-stack engineer for a company called UseRogue.com with a start date of 1/1/2024 and a salary of $500,000 per year plus benefits

You can see that the GPT-3.5 model moves a lot faster while the GPT-4 model gets more detailed and does a lot more formatting with bullets and numbers.

III. Turn on the Advanced (beta) Features

You may or may not need code interpreter, but plugins are worth trying out, but these beta features are not enabled by default, you have to go into the settings and turn them on.

IV. Set Your Custom Instructions

Custom instructions for GPT-4 refer to a feature that allows users to guide or influence the model's output by providing specific directives or context at the beginning of the interaction.

Here's a step-by-step breakdown:

1. Purpose: Custom instructions are used to tailor the responses of GPT-4 to specific needs, tones, or formats. This means you can ask GPT-4 to respond in a particular manner or style.
2. How It's Used: Typically, at the start of a conversation or prompt with GPT-4, you might provide an instruction, such as:
   • "Respond in a concise manner."
   • "Provide answers in the style of Shakespeare."
   • "Explain as if I'm a 5-year-old."
   • "Answer with a formal tone."
3. Why It's Useful: Without these instructions, GPT-4 uses its default behavior, which is to generate a response based on patterns it learned during training. By adding custom instructions, users can get more control over the type of output they receive.