Is it possible to have the prompt adjust its output real-time by the text being inputted as it is outputting its response?
Is it possible to have a blanket rule of having the output be confirmed by the user before it does the actual output? It would say “I think you want this, an example would be ____, is that what you were saying?” Then if right, output starts?
I think with these two changes it would be very efficient data to make the prompts most useful, and would add a real-time adjustment to its responses based on your reaction to it. I suspect if this existed it would be indistinguishable to a conscious situation that doesn’t need to be categorized as human. Thoughts?
It would require some careful prompting, but I don’t see why you couldn’t achieve something like this.
As they say - it’s a simple matter of programming.
In one of our chat UIs we are tracking real-time metrics concerning the time it takes to compose questions. It even captures edits and other adjustments users make before hitting go. You could catch patterns as they emerge and provide options as the query unfolds.
I saw that there is multi-processing capabilities when I asked it to tell me how to get the deepest dive in its results. If there is a “startle reflex” in a manner of speaking, that can open up an alternate analysis of itself as it is doing the confirmation of its understanding, while also trying to understand what you are trying to say, while it is delivering that to you, and stopping / adjusting real-time, that’s only a matter of assigning a resource value for the calculations of it all, then having a sort of variable sliding scale for how much of it is dedicated to what you type real-time vs its response to your previous question.
So, in recap, it’s gonna ask you questions to confirm its understanding of what you want. It could even then start asking you secondary questions like what is the application for this, etc. The result would be incredibly pointed and it would really cut out some of the issues with using it, at least for the things I have been using it for. If it’s capturing edits and adjustments made, then it could definitely get into the weeds with the details; that’s why adding the confirmation of its understanding would go hand in hand with this. you could get a sample rate by taking a sample of data of how often per response someone objects and apply it to the real-time chat. That’s kind of a stuttered reality as far as being really “there” and definitely makes conversation more fruitful