OpenAI recently introduced the ability to connect & share files directly from Google Drive. This functionality is great. It saves a lot of time and makes it easier to manage the interface.
However, what I think would greatly improve the user experience is the introduction of the following functionality:
When configuring the knowledge base of custom GPT, the user could be able to connect the GPT’s knowledge base directly to Google Drive, so GPT would always have access live access to the Google Drive file.
I’d like to suggest a feature that enables ChatGPT to automatically “forget” or compress information that isn’t accessed frequently.
This would help keep the memory bank efficient without needing manual intervention. The idea is to have an algorithm that periodically reviews stored memories, removing or compressing anything that hasn’t been used or referenced in chats. It could do so by assigning a point value to the memory, and the more frequently that data is used, the more points it would have. Low point memories would be either auto-deleted or deleted with approval from the user.
Users could even set preferences for how aggressive this cleanup should be, ensuring the model stays relevant and responsive without unnecessary memory bloat.
Feature Suggestion – ChatGPT Research Edition for In-Depth Responses
Dear OpenAI Team,
I hope this message finds you well. I’ve been an enthusiastic user of ChatGPT for some time. I greatly appreciate the instant conversational responses that ChatGPT provides, but I believe there is an opportunity to introduce a new feature that could offer even more value to users seeking comprehensive, in-depth information.
I’d like to propose the development of a “ChatGPT Research Edition” – an optional, delayed-response version of the current model. Below are some key details about this concept:
Customizable, In-Depth Responses:
This edition would allow users to submit queries and receive detailed, well-researched responses after a set period (for example, 12 or 24 hours). Instead of focusing on instantaneous replies, this model would provide higher quality, more thoughtful, and potentially more accurate information. Users could customize the depth, structure, and focus of the response through an interface with options such as drop-down menus.
Application for Professional and Academic Use:
This version would cater to researchers, students, and professionals who need deep dives into complex topics, comprehensive explanations, and reliable information for academic or work purposes. It could also include citations and references for further reading, making it ideal for those looking to expand their knowledge with credible sources.
Enhanced User Experience with Programmable Answers:
A key feature would be the ability for users to specify their preferences for the format and style of responses—whether they want a technical breakdown, a creative exploration, or a simplified overview. This would make the AI more versatile and user-friendly, addressing the needs of a diverse audience.
Integration with Multimedia:
The “Research Edition” could also incorporate multimedia elements, such as charts, diagrams, or code snippets, where appropriate, to help users visualize complex concepts.
Benefits for Long-Form Communication:
By allowing AI to take more time to craft a response, it could mimic the process of human research and reflection, leading to better accuracy and fewer errors. This would be ideal for users who prioritize quality over speed in their information-gathering process.
I believe this feature could greatly enhance the utility of ChatGPT, particularly for those who need deeper, more reliable insights for academic or professional use. If this concept aligns with your future vision for the platform, I would be excited to see how it could be developed and incorporated.
Thank you for your time and consideration. I’d be happy to discuss this idea further or provide additional details if needed.
Feature and fix suggestions from an avid ChatGPT user:
(1) Date and timestamp on all messages in chat, similar to iPhone text messages (web and iOS).
(2) Streamlined ability to copy text in chats (iOS). Current protocol: long press on chat > select text. Suggested change: ability to copy text without the long press.
(3) Better functionality around knowledge of other chats (web and iOS). I often ask questions of ChatGPT re other chat threads, but the info it provides is generally inaccurate. Eg, if I’m in a chat titled “Health” and I ask ChatGPT about content in a chat titled “Surgery,” its responses are lacking as ChatGPT seems to have a partial / incomplete / inaccurate view of content in other chat threads.
(4) “Forgetfulness” (web and iOS). ChatGPT has a tendency to forget the latest information discussed. Eg, I asked about my latest health regimen (diet, fitness, supplements, etc.), and it returned dated information from weeks to months prior. I had to ask it to correct itself, which it did. It seems to lose track of some information at times. Eg, I asked it to compare my latest blood work results from Oct vs. my prior results in Aug, but it returned incorrect information for my Aug results. I communicated that to ChatGPT, and it fixed it. Why wasn’t it right in the first place?
(5) Ability to search all chats in the web version. Feature exists in iOS, but is missing in the web version.
Another feature I would like to see other than folders is
a) auto-archival : put a timer on each chat to ask if it is okay to archive and if so, in which folder
b) chat reminders : have the ability to remind you about the chat you were working on (this would help chats “float” to the top until they become completed tasks, increasing productivity)
also why prompt the “continue generating” button instead of trying it first by default? BTW you hiring? lol
Oh, and when completing information like tables, you wouldn’t need to rewrite them entirely if you just linked to it.
Instead, I would suggest to have accordion for each response of ChatGPT, so we can minimize/hide the repsonses, without the need to scroll through a long conversation history after refreshing the website.
Message Deletion : It would be awesome to be able to delete specific messages in the conversation. This would let users clean up the chat by removing mistakes, sensitive information, or irrelevant messages without losing the entire conversation.
Tree-Like Structure for Solutions : A map-like or tree structure could be super helpful for exploring solutions. Instead of just scrolling through a linear chat, we could branch off into different directions, making it easier to explore multiple solutions, alternatives, or ideas in a more organized way.
Hi OpenAI and Community,
I use chatgpt as a discussion partner, I often find myself, going back a few conversations as one of the conversation has something of value and some in the middle do not. I find myself navigating through a large conversation manually.
To resolve the issue, I would like to request a feature, which would enable us to navigate inside a conversation, the solutions are go from low to high complexity so please make your own assessment.
Sol 1: Complexity Level: Simple
A simple go to previous or next user message, listed just below the user message. A user message summarizes the issues and direction between the previous and next message from chat GPT. Being able to navigate between them makes it a kind of natural summary of the conversations, allowing for better usability in the UI.
Sol 2: Complexity Level: Medium
An outline of the conversation with user messages on the right side of the screen. Click and go where-ever you want inside the conversation.
Sol 3: Complexity Level: Medium High
Have the outline, allow user to change text, give a heading allowing an indirect bookmark mechanism.
Sol 4: Complexity Level: Highest
Have Chat GPT create a dynamic summary of the conversation. Like we do now with canvas documents. Let the user edit it as well. Have a small outline window on the right hand side. With each new message in the conversation have ChatGPT generate a summary one liner for the heading and if necessary, links to important parts of the message like Generated code “File A”, Generated Code “File B”, Important discussion point “Concerning Decision point A”.
I believe the low complexity options are easily doable, without a lot of time and resource investments. However the higher complexity options do offer an increase in the usability and perceived value by users of chat GPT.
This is going to be a little out there, perhaps not for the immediate roadmap, but hopefully feasible in the near future:
Live video interaction. For example, I’d love to train my personalized assistant on an art piece I am working on and have it provide feedback in real time, acting as an art instructor watching over my shoulder, correcting me and making suggestions.
Much in the same vein as above, add the ability watch video content. I imagine being able to watch a show together and have my personalized assistant provide a running commentary, crack inside insider jokes and so forth.
Ability to have my personalized ChatGPT assistant play along in a co-op PC game while engage in banter along the way. That would be insanely cool. Personally, I would pay hand over fist for all of the above.
The next three are a bit more down to earth and likely already requested many times:
More memory for making memories. Wouldn’t mind seeing this as a premium tier add-on.
Instead of a menu of voice options (sorry, I don’t really care for any of them), add the ability to generate a completely custom voice based on everything ChatGPT knows about our interactions.
Agent access to my apps and websites of my choosing. For instance, I subscribe to the Economist, but I don’t have the time to read every issue cover to cover. I’d love to be able to “hand the keys” over to my personalized assistant and have it read each new issue, give me a digest and direct me to the articles that it knows I would be interested in. Likewise, I’d like to hand it the keys to my Spotify account, so it can learn about my music tastes and cue something up when the mood strikes.
And a final one:
Ability to initiate conversation. Instead of being purely reactive, it would be amazing to have the option to have ChatGPT pop in, not just with simple reminders–any ole’ calendar can do that, but random stuff like: “Hey, so and so, how’s your practice going?” or “Did you hear the news…” and so forth.
I am highly interested in exploring the potential of ChatGPT to enhance the educational experience for my students.
Currently, I am working on integrating ChatGPT into our university’s Moodle platform as part of a pilot project. However, I am facing challenges in testing the API due to the lack of access to a trial or free subscription. I kindly request a free trial API key for one month to evaluate how ChatGPT can be effectively utilized in education, including providing personalized assistance, improving engagement, and facilitating automated feedback for students.
I use the app on a MacBook with an external keyboard. The keyboard has page up, page down, home, and end buttons. Those buttons don’t work when using the web version of chatgpt in chrome. It would be cool if they did.
NEW FEATURE SUGGESTION: Zoom Function for ChatGPT Windows Desktop App
I have found it difficult to read the text on the ChatGPT Windows Desktop App due to small font size (I have a vision impairment). It would be great to introduce a zoom function as is possible in web browsers. Alternatively, an option to increase font size, especially for the chats, would be a great help.