Writing the right prompt : the importance of context

Computer with a screen displaying an introduction of Chat GPT

What is the right prompt?

The definition of “the right prompt” is very straightforward: the right prompt is the one that allows you to obtain the result that you want.

 

Writing the right prompt : the main parameters to work with

Among the main parameters you need to take into account, we have :

  • Temperature: with a scale of 0 to 1, this parameter allows you to control how creative you want the AI to be through the degree of randomness in token selection. A higher number leads to more flexibility to the model, while a lower temperature gives a more deterministic response.
  • Maximum length: allows you to control how long the combined prompt and response can be.
  • Model: when you work with text, there are 2 majors models today available the o1 suite as well as the gpt-4o suite. This contrasts with the 5 major base legacy GPT models that are available from OpenAI with their API: GPT-4, GPT-4 32k, GPT-3.5 Turbo, GPT-3.5 16k, Davinci, Curie, Babbage et Ada.

With GPT Workspace, you can use GPT-4o or GPT-4o-mini and choose the temperature though 4 options: Conservative < Inspired < Creative < High.

 

How to give the right context? Use cases

Context vs lack of context: type of content

Lack of context: how can I increase my productivity?

Context: write a blog post about how apps can be used to increase productivity within the food industry.

Just like humans, AI needs context to better perform. First, you need to explain what you want as thoroughly as possible in your prompt. What is this content used for? What is the kind of audience you want to target? Can you provide some examples of what you want?

Context vs lack of context: order

Screenshot of an example of "prompt" where more context is given versus less context

Source: Google Docs using GPT Workspace’s add-on

The first prompt is the better option since the context is provided first, whereas the LLM (Large Language Model) might simply create more context, such as: “November 15, 2024: Follow-up appointment to check the patient's condition”, if we use the second prompt.

Assigning roles

You can see in this clip that the following prompt has been used: "I have a business meeting tomorrow and I'm trying to pitch investors my brand of food supplement that boost your immune system. I want a compelling presentation about my products". In this prompt, the AI is asked to take on a specific role, the type of content – presentation – is also specified, as is the product category. Note that you can find the full video on our youtube channel.

 

How GPT Workspace integrates context?

GPT Workspace automatically considers the existing content of the page, sheet or slide you’re currently on, as well as the content you select in case you want to make edits. This specific feature allows you to automatically write a prompt that will be slightly more accurate as it feeds on more context.

Screenshot of Google Sheets with medical data

In this example, Chat GPT will consider the fact that the whole Sheet is full of content about medical equipment to answer any given prompt.

 

Conclusion

Using GPT Workspace increases your chances of getting the results you want, thanks to its advanced context integration. However, the more precisely you specify your expectations and the more context you provide, the greater your chances of success.

To have Chat GPT spit out the best possible result: write, test, refine, write, test, again and again, until you consistently get the outcome you had been hoping for. Good luck !