Mastering Prompt Engineering: Unlocking the Power of AI Technology
- Reporters: Seif Shaheen and Roba Hussein
- Editor: Sara Fahmy
Learning how to use AI and utilizing some of its main strategies, such as effectively writing prompts in day-to-day activities, enables users to harness its maximum potential. According to the latest data from Exploding Topics, a San Francisco-based company that identifies technology trends, around 40 percent of worldwide companies use AI.
Business Monthly, the journal of the American Chamber of Commerce in Egypt, says that one of the biggest challenges AI faces is having good “amounts of structured and labeled data to effectively learn and identify patterns and build accurate predictions.”
Because AI is always evolving, in essence, teaching and learning on its own, and gaining immense amounts of information in the process, users will need to precisely and accurately formulate questions in order to achieve their desired results.
Sharaf Mofeed, an expert in artificial intelligence and founder and editor-in-chief of Bawabet Masr, believes that AI is now a critical component of daily life, and plays a role in many jobs.
“AI will improve our lives,” Mofeed says, pointing to its role in helping doctors identify early signs of a number of diseases as one of the many advantages it provides.
He added that AI can be integrated into education and curriculum design based on student needs.
But there is one particular skill he says should be learned to increase the efficiency of AI use.
“The ability to write well-written [command] prompts in order to achieve the best and most detailed answers [is essential].”
He added that incorporating command engineering into AI-driven searches for optimal results relies on six essential principles. Foremost is clarity, which he explained ensures that “the system understands exactly what you need without any ambiguity.”
The other principles include specificity, which avoids broad or general responses by keeping requests precise; context, which plays a crucial role by providing background and helping the AI deliver more accurate answers to users; the language used to shape the quality of the responses; the order of information to help the AI process the users’ requests; and feedback to help refine future interactions, making the system smarter and more responsive over time for users.
“The choice of words significantly impacts the answers you receive,” Mofeed explained, highlighting the language factor.
Ahmed Mokhtar, an expert in prompt engineering, agrees that best results from an AI system depend on the pinpoint accuracy inherent in prompt design.
During his recent lecture at AUC titled “Prompt Engineering 101: How I Stopped Worrying and Let AI Do the Heavy Lifting,” Mokhtar pushed the dialog further to real-world implementation of these principles as well.
“You are trying to enhance the association, emphasizing this word, between your words. This means you need to design a solution for your problem and refine your prompt accordingly,” Mokhtar.
He added that the answers sought by a user for a specific problem should also be included in a prompt, even if slightly, so that the results provide the most precise solution.
He further described many of the prompts currently in use as open ended, specific, and conditional.
Mokhtar also provided a list of do’s and dont’s of writing prompts. Some of these include being clear and specific and using examples. However, he advised that AI users shouldn’t have validation bias, where they tend to seek out information that supports their own belief systems.
He also said that users should take AI’s feedback into consideration. In fact, AI can offer hints, ask probing questions, or provide general guidance.
Some of the daily uses of AI include writing assistance by generating new ideas, drafting documents and emails, social media posts and their captions as well; it can also be used for coding support to debug, generate and document codes.
The Center for Learning and Teaching Professor of Practice Maha Bali says that clarity and purposeful use of AI makes all the difference.
She says that the more direct the prompt, the more likely the user can avoid biases. For instance, writing “compare the following” rather than “why is this better” can help you avoid inaccuracies.
However, she does offer a word of caution.
“AI does make things up and creates things from various or different sources, so using information from it can sometimes break ethical barriers of having the right to use something.”
ICOM.AI, a conversational AI, which can mimic human conversation, gets its data from various sources, including websites and databases, and structured knowledge bases.
The array of different sources make it difficult to attribute writing to a specific person.
Bali further cautioned that AI use comes with ethical considerations. If work is being submitted, the recipient should be made aware whether AI has been used.
If users are using AI to “create data that does not exist” then that becomes an issue whereas if you “analyze data”, then that’s fine, she added.
Mechanical Engineering senior Youssef Sharaf says that “AI tools have been instrumental in assisting me with some projects and research, particularly by providing valuable resources and references.”
While Sharaf acknowledged that AI sometimes produces inaccurate information, he emphasized that, in most cases, the information it provides is accurate and greatly enhances his learning process in general.
