Feature Request: "Restricted Words" Field for ChatGPT Customization

Hi OpenAI Community,

I have a feature request that I believe would greatly benefit the entire community, enhancing the experience for users of ChatGPT.

My suggestion revolves around the creation of a new field or tab within the Customize ChatGPT feature, aptly named ‘Restricted Words’. This feature would offer users the ability to populate a list of words they wish ChatGPT to avoid using in its responses.

Currently, ChatGPT tends to recycle certain words excessively, which can be frustrating for users. Despite providing instructions not to use specific terms, it occasionally often disregards these preferences. For instance, even after instructing ChatGPT not to utilise the word “embrace,” it may persist in doing so, leading to a little frustration.

Moreover, there’s a growing concern regarding the distinct vocabulary patterns exhibited by ChatGPT across various content types, such as articles, blogs, and technical documents. This predictability could potentially compromise the authenticity and uniqueness of generated content. While seasoned users might readily identify such patterns, there’s a risk that the general public will catch on, affecting the credibility of the content.

Introducing a ‘Restricted Words’ field would address these challenges on multiple fronts. Not only would it provide users with greater control over the language used by ChatGPT, but it would also incentivize the model to diversify its vocabulary and think more creatively. By encouraging lateral thinking, ChatGPT could produce more dynamic and contextually appropriate responses, while enhancing the overall utility of the platform.

I believe this feature addition could be seamlessly integrated into both free and premium subscription tiers, potentially offering varying capacities for word storage to differentiate between the two and develop into a more versatile and user-centric platform.

Thank you for considering my suggestion, and I’m confident that this feature would contribute positively to its development.

Vincent

There’s a growing concern that articles, blogs, and technical documents could be frivolous inaccurate content generated by AI, by those that don’t follow the prohibitions in the terms of use, to not:

  • Represent that Output was human-generated when it was not.

For example, when the AI has not a single word it would change except to fill back in the segment missing from a highly-predictive cut-and-paste.


screenshot-community.openai.com-2024.02.25-07_38_45


whereas the same AI “proofread” would blandicate and predicterize any nuance of actual human writing:

my-corrections

1 Like

The word I missed was ‘versatile’ (which is now rectified) nevertheless, the point of the post was to request a negative or restricted words field that users can populate to ensure that ChatGPT does not use these words in a document that is being created by the user.