Being “good” at prompting is a temporary advantage.
Current AI systems are already adept at understanding user intent, and their capabilities are continuously improving. GPT-4 and Bing have already moved on from heavily relying on prompts.
It is also possible to achieve impressive results simply by asking the AI for direct assistance.
For example, asking, “I want to write a novel, what do I need to know?” or you can ask for a prompt to be generated listing the context and goals related to whatever you are seeking, then the prompt generated by AI can be input on the same AI application or a different one.
That said, there are valid reasons to be proactively learning prompt engineering techniques.
AI based applications are poised to become the primary interface that we interact with and utilise technologies, making prompting an essential skill for everyone to some extent. At the most basic level, you have already been doing it for years whenever refining internet browser searches.
To prompt engineer effectively input needs to be as specific as possible, including the topic, associated keywords, the writing tone and/or the target audience. There are generally two parts to a prompt:
Instruction e.g. Summarize this email into 3 bullets, and;
Context e.g. "The email contains customer feedback from the past week."
The clearer the instructions and context are described and refined successively (in light of the outputs being generated), the better the output. Overall it’s just a matter of refinement and practice.
🔗 You can deep dive into prompting with DAIR’s Prompt engineering guide
Future ready?
Comentários