Let’s make ChatGPT even better: prompt archive management or how to improve the usability of our beloved AI tool
If you are like me, you must be familiar with that feeling of frustration when scrolling through the “infinite” list of your past conversations with ChatGPT, looking for the particular one you want to re-check again, just to realise that you can’t find it — at least, not quickly and without a significant effort. It’s almost easier to ask the question again, but then you get a slightly different answer, and that’s not what you want. It’s even more painful because you know the exchange of words you are looking for is in your prompt archive, but like a needle in a haystack, you cannot find it — not an ideal user experience.
So, I started ideating …
What if we had an option to organise these conversations somehow? Not just archive them but also group and tag them in meaningful-to-us ways. In other words, we need a better way to organise and manage our prompt archives.
A bit of a context
ChatGPT is essentially a conversational interface cleverly sticky-taped on top of a powerful language model. The standard chat-like design of this user interface enables an average person to converse with the language model in a very human-like fashion. I believe it was the simplicity and familiarity of the simple messenger-like design that greatly contributed to the mass adoption of ChatGPT. The platform grew from a few to 180 million active users in less than 12 months. Feast that has never been accomplished before by any other digital platform.
Over the past year, the company (OpenAI) introduced some UI changes and minor improvements to its simple yet widely adopted chat interface. I’m grateful for all these changes, as each improved my workflows by an inch. The big kudos to OpenAI team for at least partially diverting some of their attention to this “seemingly” superficial stuff.
I understand that in the grander scheme of things, the user interface doesn’t play as significant role in how LLMs work as the other under-the-hood core components. However, UI design does, in fact, impact how an average person interacts with the language model, and this alone, at least from the user perspective, renders UI design and UX issues worth discussing.
So here I go with the suggestion to introduce better prompt archiving and conversation management to ChatGPT.
Prompt archiving and better conversation management
The more I use ChatGPT, the more I wish there was a way to somehow organise all my past conversations (prompt queries and answers) that I had with it. Not just archive them but organise them in meaningful-to-me way.
In my opinion, the fact that we cannot at least tag or colour label the past conversations with ChatGPT is rather a big usability issue, which I find progressively more and more frustrating. I talked to a bunch of my friends and tech enthusiasts about this, as well as with people who are only marginally interested in technology. Most of them seem to share my view. Even my mom, who is not very tech-savvy, suggested, “It would be nice to have some way of organising the past conversations .. into folders perhaps.”
Tags, folders, stars and colours …
So, Mr Altman, if there are any UX designers at OpenAI, please “prompt” them to look into this. What if we gave people an ability to organise the past conversations? Tag them by keywords, favourite them or create folders and sub-folders and use them as prompt archive baskets? The user would be able to drag and drop their past conversations into folders like they already do with files in file managers on their computers.
Colour coding could be another way of “tagging” conversations. Labelling them by colours is the least we could do to bring some “tidiness” to the madness of having an infinite list of past searches.
Merging and hyperlinking …
Additionally, it would be nice if we could merge conversations or group them together. Hyperlinking or referencing between conversations would be another nice feature to have.
Export and share …
Exporting convos as PDFs or sharing them with others are features also worth considering. Not as critical as prompt archive management but nevertheless useful.
And of course Search …
The ability to search for keywords and collocations is almost given when trying to improve people’s access to large pools of information.
I believe that the team at OpenAI is working on some of these features already. I would be surprised if they didn’t. As a matter of fact, I asked ChatGPT about how it would improve itself in terms of addressing prompt archive usability issues. And here is the answer I received.
Not too shabby. If GPT knows this, then I’m 100% sure Mr Altman and his team of UX designers must know it, too.
In conclusion
Development of this kind of “superficial” features might be less important to a fast-growing company such as OpenAI as they have bigger fish to fry at this point. Q-Star or custom chip design are some of them — definitely more critical from the core functionality perspective. Yet, I’m gonna use this as a gentle reminder that if the purpose of these AI solutions is to be our tools then let’s make them of a more usable kind.
It is also possible that, even though the UI changes and features that I suggest in this article seem reasonably easy to implement, they are technically hard to pull off. Either way, I’m expecting better prompt archive management to be introduced sometime soon, hopefully. Please 🙏🏼 :)
