The Prompt is the text or instruction you provide to generative AI tools to obtain the desired response. To achieve a good result, it is necessary to be precise and clear. Here are some tips for creating effective prompts:
- Be specific: Clearly specify the type of response you expect.
- Think of the AI tool as a colleague: If you were to ask for help from a human colleague, how would you phrase the question or describe the task you need assistance with?
- Generative AI is not just a search engine but also a conversational partner, as evidenced by the name ChatGPT. If you are not satisfied with the initial response, provide further clarification of the task. Ask for elaboration, clarification, adjustment, or rewriting. For example:
- Clarify the context and target audience: What level should the text be written at, what style, what tone, who is the audience, etc.
- Narrow down the context: Provide relevant details about the subject or the desired action.
- Assume that AI draws from sources of good quality and thus constructs something that is at least at your own level.
- Use anonymized data: Before inputting data, ensure that personal or identifiable information and business-critical details are anonymized or removed from the dataset.
Use your knowledge and common sense - monitor and evaluate the output
Generative AI is not flawless and can produce unexpected or outright false results. It's important to monitor and evaluate the output from generative AI models to ensure accuracy and compliance with necessary standards.
Key considerations:
You do not know the source.
You do not know the scope of the material the service uses as a source. It is often based on text from news sources, publicly available research, social media, discussion forums, etc. Likewise, there is often a cutoff date for when the model was last trained.
AI services hallucinate:
AI models can hallucinate, meaning they invent sources or facts that are not accurate. This is due to the way these models operate, with the algorithm essentially attempting to guess an answer. Language models will go to great lengths to provide what appears to be a convincing response. Always use other tools to verify statements and sources, including referencing the original source rather than the generative AI.
Be aware of ethical considerations
Generative AI can raise ethical questions, especially when it comes to personal data and business-sensitive information. The service you use will have a code of conduct for assessing what is right and wrong. This shapes the responses, and therefore, perspectives existing in the source material may be suppressed or omitted in the response. This is largely an ethical consideration regarding, for example, the inclusion of viewpoints.
Ensure to be mindful of the following:
- Transparency: Clearly communicate with users and stakeholders about how generative AI is used and what data is used. For example, if you use generative AI to gain insights into an academic area and it affects your further processing or assumptions about a topic.
- Accountability: Take responsibility for any errors and biases caused by generative AI and make necessary corrections or compensations."