ChatGPT has revolutionized a wide range of industries, from programing to marketing. However, while fairly intuitive to use, results can vary drastically depending on the prompts given. In its most basic form, ChatGPT often delivers simplistic and monotonic answers, however as innovators and insights professionals, we are often in search of answers that are more complex and creative in nature.
In order to help you reach those deeper nuanced answers, we have compiled our top tips to creating super-charged prompts that get the best results from ChatGPT.
1. Develop a complex structure.
Chat GPT works best when given guardrails, and context is a vital aspect of this. We often wear many hats in the professional environment, sometimes we’re a creative, sometimes a project manager, sometimes an excel nerd. Context tells Chat GPT which lens it should look at the mind-blowing amount of data it has access to through.
Once context has been identified, we can then clearly explain what we would like to achieve. This should be to the point and in natural language – the way we would ask a colleague to do the task.
Finally, our prompt should outline the format we would like to see this in, or details around the output. ChatGPT is capable formatting data in tabular form, and can give detailed analysis or high level summaries.
The overall prompt should look something like this: You are ‘x’, you are trying to do ‘y’, the result should be shown as/ include ‘z’.
2. Give feedback.
Using the dialogue function is a necessity in order to get the output we are hoping for. Our brains don’t work in the same way as ChatGPT so it is not uncommon for the platform to interpret our prompts in a way different to how we imagined. Tone, format style and context are all easily adaptable elements that can lead to vast improvements.
E.g. Write this response as ‘x person’. Summarize this information in less than 200 words. You are a copy editor, how would you write this to match the tone of this blog ‘insert link’.
3. Work step by step.
Asking ChatGPT to write a blog from scratch using only a title as a prompt gives an incredibly broad range, and empowers the engine to go through multiple steps without any feedback. The result will be highly generic and lack creativity or nuance.
Instead, working in a step by step approach can yield much greater results through small pivots. An example would be: 1. Asking ChatGPT for blog title ideas. 2. Selecting an idea and asking for sections that should be covered. 3. Asking it to write a blog covering these sections. 4. Asking it to edit the response in the style of a person or tone of voice of an existing article or webpage.
The success of ChatGPT lies largely in its intuitive user experience. However, like many tools, the true value in it requires some training to unlock. Spending time getting familiar with writing and amending prompts requires upfront time but will yield huge time-saving results in the future.
It is important to also set expectations. In our experience ChatGPT can offset menial daily tasks such as stimulus creation or rough data synthesis, but as of yet does not stand up as a creative counterpart when it comes to more strategic tasks such as writing complex emotional-based discussion guides or analyzing emotional data.
For examples of the ChatGT prompts we have found most useful as insights and innovation professionals, check out our recent blog article here.
For more insight in to which AI tools are changing the face of emotional research by accessing the subconscious in ways previously only accessible in labs, get in touch today for a catch up with Sidi.