In the mid-1980s, a revolutionary innovation emerged that forever changed the way people engaged with computers. It was called WYSIWYG, short for ‘What You See Is What You Get.’ Fast forward to the present era dominated by generative AI, a new concept has surfaced to redefine this dynamic: WYGIWYA, the acronym for What You Get Is What You Ask. In this post, we explore the elements that allow this paradigm shift in human-machine interaction.
WYSIWYG: A Game-Changing innovation
Before WYSIWYG, computer users had to rely on complex coding and text-based interfaces to create and modify documents. This process was demanding and required specialized knowledge. WYSIWYG changed everything by providing a visual representation of the final output, empowering users with an intuitive and familiar interface. Suddenly, people could see their documents take shape in real-time, allowing for quicker iterations, increased productivity, and a more immersive user experience.
Generative AI: Transforming Human-Machine Interaction
The rise of generative AI introduces a transformative era in human-machine interaction. It’s no longer just about the end result; it’s about skillfully designing the interaction to extract the specific information or output we desire.
Take, for example, Large Language Models (LLM) powering popular chatbots like Bard and ChatGPT. These models demonstrate an extraordinary ability to generate human-like text based on user-provided prompts. However, their output is directly influenced by the quality and relevance of input. This is true for both the training data and the user’s input. Ensuring accurate and useful output requires high-quality and comprehensive training data as well as clear and precise user input.
a. Prompt Design : Guiding the AI Model
When it comes to training data, its relevance to the questions the AI model is designed to answer is crucial for the model to learn the necessary information to produce accurate results. This is where prompt design becomes essential – ensuring that the training data is relevant and of high quality.
Prompt design involves providing instructions or guidelines to the AI model. It requires considering the specific tasks, objectives, and user requirements for the AI system and formulating prompts that align with those goals. For example, suppose the goal of the model is to provide detailed information about different car models, including features, specifications, and prices. In that case, the prompt designer might create multiple prompt variations similar to this one :
Prompt input:
“Provide detailed information about the [car model] from the year [year], including its features, specifications, and price.”
Example output:
“The [car model] from the year [year] has [feature 1], [feature 2], and [feature 3]. Its specifications include [specification 1], [specification 2], and [specification 3]. The price of the [car model] from the year [year] is [price].“
Prompt design is the responsibility of AI system developers, engineers, or designers who create and fine-tune the models. With a deep understanding of the system’s capabilities, architecture, and limitations, they skillfully craft prompts that align with the model’s specific goals and objectives.
b. The Role of User Input
Alongside prompt design, prompt crafting plays a vital role in the relevance of the AI tool’s response. It involves the user’s skill in formulating a clear, specific, and context-aware prompt that effectively communicates their intent or information needs to the AI system. Assuming good prompt design, a poorly crafted prompt leads to suboptimal responses, while a well-crafted prompt is key to achieving the desired outcome.
For instance, a poorly-crafted prompt would be:
“Tell me about AI.”
This request lacks specificity and context, providing no clear objective. Consequently, it may result in a generic or generalized response that fails to address the user’s specific information needs.
In contrast, consider the following well-crafted prompt:
“Could you provide an overview of the recent advancements in natural language processing within AI, specifically focusing on how language models that power AI tools like Bard and ChatGPT have improved understanding and generation of human-like text?”
This prompt provides clear context, a specific request, and guides the AI model to provide a detailed and relevant response that directly addresses the user’s interest in advancements in natural language processing.
To Summarize
WYSIWYG transformed human-computer interaction by offering us a visual representation of the final output. Now, with generative AI at the forefront, WYGIWYA encapsulates the significance of prompt design and crafting in maximizing the potential of AI models. By skillfully designing the interaction and formulating well-crafted prompts, we can unlock the full power of AI and ensure that what we ask is precisely what we get.
Resources
AI Talks 
Leave a comment