Prompt & Co.
by GenAIGeek

Generative AI Academies by Anthropic, NVIDIA, Google, IBM,...

Keep yourself up-to-date with top-notch GenAI trainings, be the crack you want to be and stay ahead !
These days, quite a variety of IT companies involved in GenAI provide up-to-date training material. Stay up-to-date !

Prompting: The ChatGPT Prompt Framework

Unlock full potential when prompting AI.

Act as a [Role], perform a [Task] in a [Format]

Role	Task	Format
Author	Essay	List
CEO	Article	PDF
Designer	Receipe	XML
Marketer	Ad Copy	HTML
Inventer	Analysis	Code
Therapist	Headline	Table
Journalist	Blog Post	Chart
Advertiser	Summary	Graph
Copywriter	Sales Copy	Project
Accountant	Video Script	Rich Text
Ghostwriter	Book Outline	Summary
Enterpreneur	SEO Keywords	Markdown
Mindset Coach	Course Outline	Word Cloud
Prompt Engineer	Email Sequence	Spreadsheet
Project Manager	Social Media Post	Presentation
Website Designer	Product Description	Plain Text File

R-T-F - Role-Task-Format
Act as Role, create a Task and show as Format.

Prompt example: Facebook/Meta Ad Marketer
- Role Design a compelling Facebook/Meta ad campaign to promote.
- Task A new line of fitness apparel for a sports brand.
- Format Create a storyboard that outlines the sequence of ad creatives including ad copy, visuals, and a targeting strategy.

T-A-G - Task-Action-Goal
Define a Task, state the Action and clarify the Goal.

Prompt example:
- Task The task is to evaluate the performance of team members.
- Action Act as a direct manager and assess the strengths and weaknesses of the team members.
- Goal The goal is to improve team performance so that the average user satisfaction score moves from 6 to 7.5 in the next quarter.

B-A-B - Before-After-Bridge
Explain the problem Before, state the outcome After and ask to the Bridge.

Prompt example:
- Before We are nowhere to be seen in SEO rankings.
- After We want to be in the TOP 10 SEO rankings in our niche within 90 days.
- Bridge Develop a detailed plan for mentioning all the measures we could also include a list of the TOP 20 keyboards.

C-A-R-E - Context-Action-Result-Example
Give the Context, describe the Action, clarify the Result and give an Example.

Prompt example:
- Context We are launching a new line of sustainable clothing.
- Action Can you assist us in creating a targeted advertising campaign that emphasizes our environmental committment?
- Result Our desired outcome is to drive product awareness and sales.
- Example A good example of a similar successful initiative is Patagonia's "Don't Buy This Jacket" campaign, which highlighted their committment to sustainability while enhancing their brand image.

R-I-S-E - Role-Input-Steps-Expectation
Specify the Role, describe the Input, ask for Steps and describe the Expectation.

Prompt example:
- Role Imagine you are a content strategist.
- Input I've gathered detailed information about our target audience, including their interests & common questions related to our industry.
- Steps Provide a step-by-step content strategy plan, identifying key topics based on our audience insights, creating an editorial calendar, and drafting engaging content that aligns with jour brand message.
- Expectation Our aim is to increase our blog's monthly visitors by 40% and enhance our brand's position as a thought leader in our industry.

Prompt Engineering, RAG, Fine Tuning

Which should I use? Prompt Engineering, RAG or Fine Tuning?

Prompt Engineering	RAG	Fine Tuning
The practice of designing and refining input prompts to elicit desired responses from AI models.	A technique that combines pre-trained language models with external knowledge retrieval systems.	The process of further training a pre-trained model on a specific dataset to adapt it to a particular task or domain.
Goal
To guide AI behaviour without changing the model.	To enhance AI responses with up-to-date and relevant external information.	To specialize a model for specific tasks or domains.
Use Cases
•  General purpose interactions •  Creative writing assistance •  Task instructions for AI •  Controlled content generation	•  Question answering systems •  Fact-checking applications •  Research assistants •  Customer support chatbots	•  Sentiment analysis •  Named entry recognition •  Text classification •  Domain language generation
Complexity
Understanding of model behaviour but no programming skills required	Retrieval system integration & query formularion skills	Deep learning, dataset preparation & training management
+ Pros
•  Quick to implement and interate •  No need for additional data/training •  Flexible and adaptable to various tasks •  Can improve model performance without changing underlying structure	•  Enhances responses with up-to-date external information •  Improves factual accuracy/relevance •  Can handle a wide range of queries •  Reduces hallucination in AI responses	•  Produces highly specialized models •  Significantly improves performance in target domains •  Allows model behaviour customization •  Learns task-specific patterns •  Requires less context & short prompts
- Cons
•  Limited by the base model's knowledge & capabilities •  May require extensive trial & error •  Results can be inconsistent/unpredictable •  Effectiveness depends on skills of the prompt engineer! ==> improve prompt engineering skills and use the ChatGPT Prompt Framework	•  Requires maintenance of external knowledge bases •  Can be computationally intensive •  May introduce latency due to retrieval process •  Effectiveness depends on the quality & relevance of retrieved information	•  Requires substantional computational resources and technical knowledge •  Needs large, high-quality datasets for effective training •  Risk of overfitting to the training data •  Less flexible, fine-tuned models may perform poorly on unrelated tasks

In case of questions, don't hesitate to contact us from GenAIGeek.ch !