
Just like anything else, something that can create a sound will create attention, and OpenAI is plain loud.
At least in the tech world, where it generates most of its hypes, thanks to ChatGPT, where the product occupies pretty much every single topic about artificial intelligence because after all, it created the hype, and it successfully delivers the answers to meet the demands.
The tool that is essentially a large language model-powered AI, allows users to use it to create essays, and even write an online article. It can help people do their homework, or provide advise, create poem, songs and even aid research.
While the AI shows a lot of promises, it has limitations, and lack of customizations.
This is where plugins play their roles.
And this time, things just get better, and better.
This is because OpenAI launched 'custom instructions' for ChatGPT users, so they don’t have to write the same instruction prompts to the chatbot every time they interact with it.
Introducing Custom instructions! This feature lets you give ChatGPT any custom requests or context which you’d like applied to every conversation. Custom instructions are currently available to Plus users, and we plan to roll out to all users soon! https://t.co/fVIM9GeYk2
Here…— OpenAI (@OpenAI) July 20, 2023
According to Joanne Jang, who works on model behaviors and product at OpenAI:
Using custom instructions, publishers who use ChatGPT to help them create articles, can use the feature to create inputs like "Write the answer under 1,000 words" or "Keep the tone of response formal," and make ChatGPT remember that when needed.
In another example, a teacher can say that they are teaching fourth-grade mathematics or a developer can specify the code language they prefer when asking for suggestions. Also, a person can use the feature to specify their family size, so ChatGPT can give responses about meals, grocery, and vacation planning accordingly.
The easiest way to think about it, according to Jang, is that custom instruction is sort of a permanent preamble to users' queries. Instead of crafting a long question to ChatGPT with all the context and information required, users just have add a context and information to their custom instructions, and it’ll be there every time.
"We’re introducing custom instructions so that you can tailor ChatGPT to better meet your needs," OpenAI said in a blog post.
In short, users don't have to repeat themselves.
The company said that the feature lets users "share anything you’d like ChatGPT to consider in its response."
What's worth noting here is that, users can already specify these things while chatting with the popular chatbot. Custom instructions are only helpful if users need to set the same context frequently.
The instructions also work with plugins, making it easier for them to suggest restaurants or flights based on your location.
Initially, this feature has been made available to ChatGPT Plus subscribers.

To use this feature, users accessing ChatGPT through its web interface can click on their name and go to 'Settings', and select 'Beta features'.
There, they must select the 'Opt into Custom instructions'.
When users ChatGPT on iOS, users can access this by going to 'Settings', and select 'New Features', to then 'Turn on Custom Instructions'.
OpenAI said that the information given to the customize responses shall be used to train its API models to adapt to different instructions.
"Information from your use of custom instructions will also be used to improve model performance – like teaching the model how to adapt its responses to your instructions without overdoing it," the company said.
For privacy concerned users, they can opt out of this setting through their data control settings.
It should be noted that OpenAI uses its moderation API to scan customized instructions to check if whether they are unsafe in any nature. If it finds users abusing the feature, or in any way violating the company's policy, ChatGPT can refuse to save the instructions or ignore them accordingly.
OpenAI does this to prevents users from receiving harmful or hateful answers from ChatGPT.