DailyPost 2387
CHATGPT & PROMPT ENGINEERING
Long-time back I had created a banner in my office, ‘Reply to what you are asked for, and not you want to say.’ Lots of times you would be finding people whose reply would remain the same for any question you ask. You would also have had the experience of getting bored by people who would run the same ppt over and over again, whatever might be the new dimension of the topic or the progress made in it. For them, it is of eternal value and to it all, they would end up reading it. Have we made the best of the knowledge resources made available by the digital world at the click of a mouse?
The answer has to be an emphatic NO. This is result of research on knowledge improvement in the last three decades in the US. Our general quest for knowledge is unfortunately limited to getting good grades in exams, clearing competitive exams and fulfilling the benchmarks of the professional word. In the scenario, we have the grand entry of ChatGPT, having digested all the internet knowledge repository and is ready to help mankind by conversing with them. A meaningful dialogue to take care of your professional and other requirements. Now it can be powered by GPT-4 which is a massive improvement of GPT-3.5, with input / output limit at 25,000 words and it being multimodal.
The capability to use this database of human knowledge in the best possible manner for your specific requirements using the right prompts is known as prompt engineering. This is the most recent of engineering professions and most relevant too, given the fact that most of engineering jobs related to coding and algorithms seems to be sorted out for the wider world. On the face of it, it does not seem to require any engineering degree, as of date there is no course being offered in it. It carries a hefty salary tag of $250,000 to $335,000 per annum. ChatGPT is a computer program that understands natural language than a code.
A prompt engineer first needs to understand the nuances of the ChatGPT and its features, for what all it can deliver and in which manner. The challenge is that ChatGPT can give unpredictable results, your job as a prompt engineer to deliver results which are most reliable and hence more valuable. So, the challenge is to give better inputs to ChatGPT. How to give better inputs? First is to learn how to give better instructions using right words (prompts). What does this process look like, input – tell ChatGPT what you want, explain how you want your output to be formatted and then get and evaluate the output, if it is aligned with your needs. If it is not, then you get back to the first step and tell what you want and there is whole process to reach you final result and even changed requirements. Intelligent tool and domain based iterative cycle is the key. Suffice to say, prompt engineering is the art of giving high quality input to get high quality output, through a well defined process.
BETTER PROMPTS IS EQUAL TO BETTER RESULTS IN THE WORLD OF CHATGPT, IS THE ART OF PROMPT ENGINEERING.
Sanjay Sahay