A notable example can be found in Ancient Greece, where Socrates used his method of maieutics, based on dialogue and strategic questions, to guide his disciples towards the truth. Centuries later, the importance of knowing how to ask questions continued to gain strength even in fiction: in the novel 2010: Odyssey Two by Arthur C. Clarke, Dr Chandra, creator and supervisor of the supercomputer HAL 9000, is sent as a psychologist to unravel the reasons why the artificial intelligence had become paranoid. This episode illustrates how the art of asking the right question can make the difference between understanding and misinterpreting the behaviour of an intelligent system.
In the current field of artificial intelligence (AI), this same ability to formulate clear, specific and contextually relevant questions is key to obtaining useful and reliable answers from AI models. Throughout this article, we will explore the fundamentals of prompt engineering and present practical techniques for improving communication with these models.
Introduction to Prompt Engineering
Prompt engineering is a fundamental discipline in the field of artificial intelligence. It focuses on designing and optimising the instructions provided to AI models to obtain accurate, useful responses aligned with the desired objectives. The quality of the prompt not only influences the effectiveness of the responses generated, but also impacts the overall efficiency of the interaction with the model.
As AI systems become more sophisticated, so too does the importance of learning how to communicate with them effectively. To achieve this, there are techniques and rules that facilitate this task, such as the mnemonic SMART and KISS, as an approach to this field, and without wanting to be exhaustive.
The SMART technique for effective prompts
The SMART technique is used to structure prompts in a way that maximises their effectiveness. Each letter of SMART represents a key principle:
- Specific: A clear and detailed prompt reduces the possibility of misunderstanding by the AI model.
- Example: Instead of asking ‘What’s going on in the world?’, it is better to ask ‘What are the most relevant economic news items today?’.
- Measurable: It is essential to be able to evaluate the quality of the generated response, establishing criteria that allow verifying if the result meets expectations.
- Achievable: The request must be within the known capabilities of the AI model. Asking something that exceeds its knowledge or abilities can result in inaccurate responses.
- Example: Instead of ‘Explain a person’s thoughts to me’, an achievable request could be ‘Explain the possible reasons behind a specific decision’.
- Relevant: The prompt question should be aligned with the desired objective. Irrelevant requests divert the model’s focus.
- Temporal: Including a time frame, if applicable, helps to contextualise the answer.
- Example: ‘What was the economic situation in Europe in 2008?’ is more precise than simply asking ‘What happened in Europe?’.
The KISS rule: Keep It Short and Simple
Another key technique is the KISS rule (Keep It Short and Simple), which emphasises the importance of keeping prompts concise and direct. A short and clear prompt minimises ambiguity and improves the model’s understanding.
Example: Instead of asking ‘Could you help me understand more about that topic that is related to technology and science and that is important for companies?’, use ‘What is the impact of artificial intelligence on technology companies?’
Importance of context
Providing enough context in the prompt is essential for the AI to understand the intent of the request.
- Ineffective: ‘What is the capital?’
- Effective: ‘What is the capital of France?’
The right context guides the model and increases the likelihood of obtaining relevant and useful answers.
Practice makes perfect
Like any discipline, prompt engineering requires constant practice and experimentation. Through trial and error, users can refine their approaches and discover the most effective strategies for interacting with AI models. Mastering this ‘language’ is learning to converse with AI: don’t be afraid to try different approaches and analyse the results.
Applications for creating prompts
There are several applications that facilitate the creation of effective prompts:
- PromptPerfect: Considered one of the best. It is compatible with several AI models and helps you refine your prompts to obtain exceptional results. It offers suggestions, variations and detailed analysis.
- PromptoMania: Specialised in image generation. It allows you to create detailed prompts for text-image diffusion models such as DALL·E, Stable Diffusion and Midjourney.
Designing clear and effective prompts is the key to unlocking the true potential of AI. With practice, patience and the formulation of precise questions, like Dr Chandra investigating HAL 9000 or Socrates conducting dialogue, we can better unravel the mysteries of this powerful technology.
