OpenAI just launched a way to improve your ChatGPT Plus prompts. You can now save a bunch of keystrokes (tokens) using a new beta feature called “Custom Instructions.” Unlike our forgetful ChatGPT of yesterweeks, the improved interface can “remember” your persona, your preferences, your style, etc., to save you from having to constantly tell it who you are, who it is acting as, and what constraints it should respect.
For example, a teacher who uses ChatGPT Plus to assist with the creation of lesson plans might use Custom Instructions to constrain answers that are appropriate for a third-grade classroom. Family chefs can specify the size of their household for tailored recipe suggestions. Coders can get responses in their favorite language. It’s a context that stays put, a preface to your questions that doesn’t need a reintroduction. Importantly, there’s a toggle switch to turn it off when you want ChatGPT Plus to act as something or someone else.
While there are teething issues – as is the norm with any technological innovation – OpenAI has already demonstrated a responsible approach. Safety is paramount; it will not permit harmful instructions or divulge personally identifiable information. This feature is not about knowing your darkest secrets, but to tailor-make interactions that are uniquely yours. You type less, it understands more.
In practice, this may really not be all that helpful. The beauty of ChatGPT Plus is that it can be easily prompted to constrain its answers to specific questions by properly crafting your prompts and coaching the AI model to give you what you want. Having it “remember” who you are sounds really helpful, but you’ll need to spend some time tuning Custom Instructions to ensure that the feature’s outputs are aligned with your goals.
Author’s note: This is not a sponsored post. I am the author of this article and it expresses my own opinions. I am not, nor is my company, receiving compensation for it.