 (8).png)
1 month ago
4
ARTICLE AD BOX
The other day, ChatGPT quite baffled me by commanding me to stop eating batteries and definitely not put one in the microwave again. It sent a page of instructions on how to remove a battery after it had been heated in an oven, and asked whether there was any burning, irritation, or metallic taste in my mouth. It urged me to get to a doctor on the double. “Tell me what kind of battery it was? Tell me what exactly happened,” it said.
It turned out the word “baguette” had autocorrected to “batteries” in my prompt, which was about whether microwaving this French delight would make it chewy on the inside. Apparently, it would. But for a moment there, both the chatbot and the user thought the other had lost it.
After the battery-baguette incident, I decided to pay more attention to the way I prompted and follow the recommended guidelines.
Be crystal clear
The most important rule, and you'll find it in all expert recommendations, is to be specific almost to a fault. Specify the exact goal rather than leaving the AI to guess, or it will veer towards errors and hallucinations. It's often said that prompting is just asking clearly.
If you want writing, images, videos, apps, solutions to problems or anything else, don't just throw the subject at the AI. Spell it out in some detail. Don't say “Edit this image”, but think about what you really need and ask. If there are several steps to the result you want, explain that as well. A thought-out, clear prompt takes longer to 'engineer', but it saves you time later. Remember that AI isn't like Google once was; you can't pop in a few keywords and expect magic. If the request is vague, AI predicts and guesses; if the prompt is clear, it performs. Treat it like a literal intern.
At the same time, don't give so much detail that it veers off into confused territory. If you paste a 50-page report and ask a question about a detail hidden on page 25, the AI might miss it. This is known as the 'Lost in the Middle' effect. Studies show that AIs are great at remembering the very beginning and the very end of your prompt, but they tend to zone out in the middle. Put your most important instructions and your main question at the very start or the very end of your message.
Give the context
Giving full contextual information really helps. In fact, you can use the same basic prompt and get remarkably different results by changing the context. Explain your purpose. For instance, I get a roundup of AI news every so often and if I just said “Round up the AI news for the week”, I'll get all sorts of company and financial news, enterprise applications, and academic research thrown in. My interest is more in AI for the ordinary non-techie user and everyday life situations, and in the impact AI has on our lives and society. I will need to spell that out to get only the most relevant news.
If you're writing, spell out the audience. If you're teaching, say what you want to teach and to whom. When I ask for an app to practise French, for example, I give context by reminding it of my proficiency level and the area I want to tackle.
A trick many people use is 'Act like...', which means asking the AI to play a role, aligning its language and output with what you want. Act like a travel planner and work out a relaxed itinerary for a visit to Bhuj.
Refine, don't restart
I've watched users get frustrated with their chat assistants and say something like, “Here it is again. I want you to…” AI is not a vending machine where a single command yields a perfect result. If the first answer is too long, too formal, too vague or simply not what you meant, don't start all over again. Just steer it. Say “make it shorter”, “simplify that”, “give me examples”, or “write it in a friendlier tone”. Good prompting is often less about getting the first instruction exactly right and more about nudging the answer into shape.
In conclusion, ChatGPT said, “And I must say, ‘Are batteries chewy?’ is one of the more alarming autocorrect moments I've seen today. Autocorrect nearly gave me a heart attack there for a moment. From batteries in the microwave to a perfectly innocent baguette, quite a plot twist. Sprinkle a few drops of water on the crust of the baguette and put it in the oven or toaster oven for about five minutes.”
The New Normal: The world is at an inflexion point. Artificial intelligence (AI) is set to be as massive a revolution as the Internet has been. The option to just stay away from AI will not be available to most people, as all the tech we use takes the AI route. This column series introduces AI to the non-techie in an easy and relatable way, aiming to demystify and help a user to actually put the technology to good use in everyday life.
Mala Bhargava is most often described as a ‘veteran’ writer who has contributed to several publications in India since 1995. Her domain is personal tech, and she writes to simplify and demystify technology for a non-techie audience.
English (US) ·