Prompting AI
Creating topics and writing prompts
The topic dialog allows you to input a topic you wish to generate with a selection of options. These options modify the AI generation to produce different results.
Tone: Sets the writing tone the default option is Friendly.
Style: Sets the writing style, options include: Expositive, Descriptive, Narrative, and Persuasive.
Content-Length: Sets the content length from short-form to long-form. The default is a balance between the two.
Temperature: Sets the AI model's probability randomness (incorrectly marketed as a creative mode). The default setting is 0.3 which provides more deterministic responses with slight variations i.e. predictable with some nuance. High temperatures can be useful on some topics that don't necessarily need to be factually sound.
Models: Allows you to select models, we recommend sticking to DaVinci as it currently is the most stable. We have implemented ChatGPT and GPT-4, they have superior generation but are unpredictable, unstable and slow. We anticipate that these newer models will become stable and reliable in the near future.
Generation Time: Depending on the topic and OpenAI capacity, generations can take several minutes to complete.
Creating new Variations
Once you create a topic, it will be saved as a variation. If the generation seems a bit off or you would like to generate another variation, you can do so by selecting "Create new variation". This will allow you to configure options to produce different results.
AI Prompt Output Languages
The top-performing (80% and above) languages from OpenAI's Paper can currently be applied to when generating a prompt.
You can choose one of the following languages for output:
English German French Chinese / Mandarin Spanish Italian Japanese Afrikaans Russian Greek Indonesian
Selecting a language for your prompt will create generations/variations in that language. We recommend using a low temperature if you intend to create multiple languages for the same prompt. This will make the generation deterministic and is likely to create a direct translation instead of veering off.
Selecting a prompt output language will NOT translate existing variations, it only creates a new variation and the content might be slightly/drastically different in other languages.
How to output AI generations in different Languages
Open Authoring. Select the AI button from the toolbar (top right).
Select the + button (bottom right) to generate a new topic/prompt. Set the Topic, Style, Tone & Language. Write your prompt in English. Don't specify language itself in your topic/prompt input.
Settings. Adjust the Content Length, Temp and AI Model as desired.
Create. Select the Create button.
Check accuracy. Verify the content accuracy with an expert fluent in the language of your choosing.
Use the generated content. Create a Sotry, Chapter, Article or Questions.
Prompting Tips and FAQs
How do I get the best results from a prompt?
Run multiple short, specific prompts. Say for instance you want to create learning about spiders. It would be better to run multiple prompts with sub-topics about spiders i.e.:
Anatomy and Physiology of Spiders
The Diversity of Spider Species
Understanding Spider Lifecycles
Spider Behaviour: Web Building, Hunting and Courtship
Spiders and Ecosystem: Their Roles as Predators
Spider Bites and Human Health: Myths and Facts
...etc
Pro tip: If you have a broad topic and don't know what sub-topics create, you can always use Chat-GPT and ask it: "Suggest e-learning topics for [insert topic]". This will give suggested topics to use in the platform A.I.
How to formulate a prompt that references a specific source?
Provide a topic and the reference, for example:
Prompt Template Topic by/based on Reference/Author/Book
For example 5 Steps to Embrace Your Feelings based on the study Weaving colourful threads OR 5 Steps to Embrace Your Feelings based on Celia Kourie
Tip: Shorter, to-the-point prompts work better.
Can the A.I. produce longer generated text?
There are some options to produce longer text with A.I.
Increase the content length. When making a prompt you can increase the content length to long-form. This will attempt to get the A.I. to produce longer text. Just remember that GPT is tuned for summarising it still might not increase the output.
Create multiple prompts. To create more content on a single topic, try creating multiple prompts for the same topic except using sub-topics. This will allow us to use multiple generations to create individual chapters on the same story.
Use inline prompts. If a generation creates a section that is too brief, you can expand on the section by using an inline prompt. Simply type "/" forward slash and enter a specific prompt. With inline prompts you need to be specific to what you want.
Last updated
