A lot of times you think if you are responsible for solving a problem, you have to make a solution that you understand or can conceive of. But with AI, it’s almost like you don’t have to understand it or conceive it or perceive it. I don’t understand how to put it into words, but right now. I have one dataset, and I’m trying to build prompts that can create something that resembles that database. Okay, so in the first run we generated something. I made a prompt that can compare this data with the existing data, and it made those fixes. But I was like, okay, wait. If we are using this prompt to generate something else, then we don’t we don’t have the dataset to compare against how are we going to win so I asked it to duplicate the first set of data we had generated and for one run, enrich it using the actual data. And for the second one, use a prompt and keep modifying the prompt until we get to where we want. And now I think reasonably we can generate using one prompt and use the other prompt to probe it until we get something high confidence.

I’m complicating it too much about, but basically for one day test we have a benchmark and in future we may not have the benchmark so we are creating a generation prompt and a probing prompt that can fluff up whatever the generation prompt missed out on. And this another prompt I don’t understand how it is going to work, what it is going to do. I just know it has a role to play that okay once we generate and there is of course most probably going to be gap between generated and the actual and we will have to bridge that gap and can we make a proxy type entity of sort that can always do it when we don’t even have the data? It’s very different than the first use case we saw, but still it is going to exist. So AI kind of enables you now to make things. That you don’t understand, but just know what you want done. So I think as long as you know what the bottleneck is or what is stopping you from getting you what you want, you just have to kind of describe a problem and ask it to come up with a solution of some sort and be able to have at least one yardstick to compare it against, and then maybe you can trust it to get the results you want on a novel run. So. What I understand now is I should be clear about what I want and the steps involved in what I want. Even if I don’t have the exact tool needed to complete all those steps, I can always ask it to infer something and make sure that it’s high confidence enough.

This is high strangeness. Getting solutions to problems because you took the initiative but dont know what exactly is helping you. Is it a ghoul or ghost or ai or singularity idk!!!!!!!!!!

genuinely different mode of problem-solving.

its like just knowing you have problem and being entitlted enough to ask that it be solved.

Door opens for the one who knocks, so for you problems to be solved. You will win . lesgooooooooooo.

Acknowledge there is a problem. Ask for it to be solved. Be clear about what winning looks like so that the high strungness can take the wheel and make something that can close the gap between what is and what winning is.