Independent Online

Monday, December 11, 2023

View 0 recent articles pushed to you.Like us on FacebookFollow us on TwitterView weather by location

Tech News: Do you know the ‘language’ of prompting?

This picture taken on January 23, 2023 in Toulouse, southwestern France, shows screens displaying the logos of OpenAI and ChatGPT. Photo: AFP

This picture taken on January 23, 2023 in Toulouse, southwestern France, shows screens displaying the logos of OpenAI and ChatGPT. Photo: AFP

Published Jul 4, 2023


If you do not know the “language” of prompting, you will find it increasingly difficult to communicate with the numerous artificial intelligence (AI) systems that are now part of our everyday life.

Suddenly, after the popularisation of AI tools, AI has become an integral part of our lives and businesses. Generative AI, virtual assistants and chatbots are currently present in many things we touch. And the most impressive is that AI keeps getting smarter and more capable and has changed the way we interact with humans and, in particular, machines.

These dramatic changes gave rise to the need to improve our communication with machines. Therefore, understanding prompting is crucial for effective interaction with AI systems because it forms the basis for communicating our intentions and obtaining the desired outputs. A knowledge of prompting is the only way to unlock the full potential of AI systems.

Prompting refers to the input or instructions provided to an AI system (eg text or voice) to elicit specific responses or perform particular tasks. Prompt engineering serves as a means of guiding the AI language model’s behaviour and shaping its output through a step-by-step process of creating inputs, whether it is a chatbot, virtual assistant, language translation tool or a content generation tool.

Prompt engineering will thus enable us to extract relevant information and gain new insights in a variety of fields. AI has numbered the days of browsing through numerous webpages on the internet to find the information that is needed. New AI technologies, such as ChatGPT and Bard, makes it much easier to find the appropriate information, as long as the right questions are asked. How to structure and phrase the prompts are important in obtaining the quality, accuracy and relevancy of results.

Language models

AI language models (eg ChatGPT) depend on deep learning algorithms, natural language processing (NLP) and large datasets of articles, books, journals, reports and webpages to comprehend human language. Unsupervised learning, where the model analyse unlabelled datasets for appropriate and accurate responses is used to enable the AI model to generate text based on a specific prompt. Huge datasets from a variety of sources increases the model’s understanding of human language, including grammar, syntax, and semantics.

Effective prompting

Effective communication with AI language models entails the following of certain guidelines to get quality, accurate and useful answers:

Be clear and specific: Pose well-defined and specific questions to receive clear and concise answers.

Be concise: Keep questions brief and focused, avoiding multiple inquiries in a single prompt.

Provide context: If the question pertains to a particular topic, offer some context to aid the AI model in comprehending the query accurately.

Use proper grammar: Formulate queries in complete sentences with correct grammar to enhance understanding.

Use keywords: Incorporate relevant keywords in the questions to assist the AI model in grasping the essence of the inquiry and providing suitable responses.

Proofread: Before submitting a prompt, review it for grammatical correctness and overall clarity.

Using verbs in AI prompts

Since prompting entails giving instructions, structure and direction to the AI, it is important to use verbs as action words to guide the AI in the actions it should take and in what information to provide. Appropriate verbs in a prompt enable the AI to perform very specific tasks, such as analysing data, providing a definition, explaining concepts or brainstorming ideas. The more specific and targeted the verbs are, the more precise and relevant the response of the AI will be.

Some examples of verbs that can be used in prompts are: generate, illustrate, change, clarify, list, compare, predict, criticise, rephrase, suggest, summarise, translate, and many more. A typical example would be: “Analyse this dataset and provide insights on the patterns and trends observed.”

Tone of voice in AI prompts

When formulating prompts for AI, it is crucial to contemplate the desired tone of voice that a business wish to convey to its audience. The tone of voice the business chooses for its AI prompts has a profound impact on the way the audience interprets and connects with the information delivered. It can evoke specific emotions, establish a sense of familiarity or create an atmosphere of professionalism.

By consciously selecting a consistent and appropriate tone of voice in AI prompts, a business can reinforce its brand’s identity, as well as establish trust and credibility with an audience. By adopting a tone that aligns with the target audience’s preferences, a more engaging and enjoyable user experience can be created. For instance, a conversational and friendly tone may work well for a casual chatbot, while a professional and concise tone may be more appropriate for a customer support AI.

Some suggestions to consider when choosing a tone of voice for AI prompts are:

Define the brand’s personality: Determine the key traits and values that define the organisation’s brand, and reflect them through the tone of voice. Is the aim warmth and empathy or professionalism and expertise? Outline the attributes that best represent the brand’s character.

Understand the audience: Gain insights into the target audience’s preferences, demographics, and communication-style. Tailor the tone to resonate with their expectations and build a connection. Consider factors such as age, cultural background, and the context in which they will be interacting with the AI system.

Be consistent across platforms: Maintain a consistent tone of voice across all touchpoints and platforms where the AI system interacts with users. This consistency reinforces the brand’s identity and ensures a seamless experience for the audience, regardless of the channel they choose to engage with.

Adapt to the context: While consistency is essential, also consider the specific context in which the prompts will be used. Adjust the tone accordingly to align with the purpose, subject matter or desired outcome of the interaction. Flexibility in tone allows for a more tailored and effective communication approach.

Test and iterate: Continuously evaluate the effectiveness of the chosen tone of voice through user feedback and testing. Monitor user satisfaction, engagement-levels, and perception of the brand to refine and improve the AI prompts over time.

By consistently conveying a desired personality and values, a business can foster deeper connections with its audience and create compelling interactions that leave a lasting impact.

Some suggestions for tone of voice that can assist in crafting improved AI prompts are: excited, assertive, formal, poetic, friendly, powerful, gentle, professional, humorous, informative, scientific, urgent, encouraging, persuasive and many more. A typical prompt example of tone of voice is: “Craft a short article on the benefits of exercise for a person’s health, written in a scientific and informative tone that encourages readers to adopt a healthier lifestyle.”

Unlocking AI’s potential

By fully understanding how to effectively communicate with AI through prompts, it can take us a step closer to unlocking AI’s full potential. This will not only enable us to extract relevant information from the AI system, but also allow us to gain new insights, making us more informed on different fields of interest and generate material aligned with an organisation’s brand and audience.

It is urgent that we start “speaking” prompting.

Professor Louis C H Fourie is an extraordinary professor in information systems at the University of the Western Cape