EVERYTHING
According to WIRED, while most users can interact with large language models by simply typing any request, there are specific methods to elevate the quality of those responses. The publication highlights that small adjustments in phrasing or the addition of a few strategic instructions can significantly differentiate the output from standard results seen by the general public.
Adopting personas and creative constraints
One effective way to bypass the tendency of AI to be overly agreeable is to assign it a specific, inquisitive persona. For instance, asking the bot to act like a curious 10-year-old can force it to ask probing questions about a business plan or vacation itinerary. This tactic helps identify weak points in a user's logic that a standard "helpful" AI might overlook by simply agreeing with every suggestion.
Furthermore, users can leverage historical and cultural perspectives to gain new insights. By asking the AI to simulate what specific figures like Steve Jobs or Winston Churchill might think about a modern problem, users can generate diverse viewpoints based on the vast training data these models have processed. While these are educated guesses rather than true reflections of personality, they serve as powerful brainstorming tools.
Efficiency and multimodal capabilities
For those looking to streamline learning or utilize visual data, several other techniques stand out:
- The 80-20 Rule: Based on the Pareto principle, users can ask for "the 80-20" of a topic. This instructs the AI to identify the 20 percent of key information that yields 80 percent of the understanding, making it ideal for rapid learning.
- Visual Prompting: Mobile users can utilize their device cameras to include photos in prompts, allowing the AI to identify landmarks, insects, or translate signs in real-time.
- Image Remixing: Beyond generating new art, users can upload doodles or personal photos and ask the AI to transform them into different styles, such as claymation or photorealistic scenes.
Mastering these techniques moves the user from a passive recipient of information to an active director of the technology. By treating prompt engineering as a deliberate skill set rather than a casual conversation, individuals can extract significantly more utility from their AI subscriptions.